S A A R F S A A R F T A M S JUNE TECHNICAL REPORT Revision: 5. Television Audience Measurement Survey

Transcription

1 S A A R F Television Audience Measurement Survey S A A R F T A M S JUNE 2009 TECHNICAL REPORT Revision: June 2010

2 TABLE OF CONTENTS COPYRIGHT... 1 EXCLUSION OF CLAIMS... 1 SAARF OBJECTIVES AND BUSINESS... 1 SAARF MANAGEMENT... 2 SAARF COUNCILS... 2 SAARF WEBSITE ADDRESS BRIEF HISTORY OF SAARF TAMS Before PeopleMeters st SAARF Tender and AC Nielsen Dis-Invests Mark 1 PeopleMeter Industry Acceptance Mark 2 PeopleMeter Black Market Meter Panel Limitation of Coverage The Next Generation of PeopleMeters Eurometer PC Tips and Eurometer Panel Measuring Digital Satellite Channels Measuring DSTV Satellite in Total Measuring DSTV Satellite by Channel OATS Augmented Satellite Panel Nielsen Media Research Established Only Multichoice DSTV Measured SAARF Universal SU-LSM Merger between AGB Italia and Nielsen Media Research and the New SAARF TAMS Contract Recent Changes to SAARF TAMS Panel CURRENT METHODOLOGY AND PROCESSES Universe Sample Design SAARF AMPS : the Establishment Survey Coverage of Satellite Households Panel Turnover and Renewal Growth of the SAARF TAMS Panel DEFINITION OF TERMS Adult AQH ASO ASR ATS Audience Share Avail BARB (UK) Bouquet Break Position... 12

3 3.11 Cable TV System CAPI CATI Children Churn (Rate) Commercial and Programme Logs Concurrent Viewing Consolidated Audience Consolidated Viewing Content Provider / Supplier Cost Efficiency Cover CPP/CPRP CPT/CPM Criterion of Viewing Daypart Decoder DSO Disaggregated Viewing Data Disproportional Sampling Effective Frequency Effective Reach Enforced (panel) Turnover Engagement Event Types Exclusive Reach Fragmentation (Audience) Frequency Gold Standard Good Households or Respondents GRP Home Language Household Purchaser Housewife HUT Individual Viewing Statements Lead-in / Lead-out Linear TV Live Audience Main Purchaser Makegood Media Dependent Media Independent Net Daily Reporting Sample (in-tab sample) OTS Out of Home Viewing Overnight Viewing Data Panel Balance Panel Fatigue Panel Maintenance Patronage Penetration (medium / channel) PeopleMeter Personal Meters Person Viewing Picture Matching... 19

4 3.67 Podcast Polled Sample Polling Pollux Position in Break Post (campaign) Evaluation PPM Primetime Processed Viewing Statements Product Placement Profile (adhesion) Programme Genre Programme Loyalty PUT PVR Random Sample Rating [TV] Reach (or Cover / Cume) Reporting Homes Response Bias Return Path Return Path Data (or Click Stream Data) RIM Weighting SAARF Universal LSM s (Living Standards Measure) Sales House Sample Sample Error Sample Size Sample Frame SI Slave Meter Solo-Viewing Standard Error SAP STB Stratification Target Market (or Target Audience) TSV Trick Mode TVEvents System Uncovered Viewing / Set Use Unidentified Channel Viewing Visitors (also known as Guest Viewing) VOD VOSDAL Wearout Weight Factor Within Programme Break Zapping Zipping HARDWARE DESCRIPTION AND OPERATION SAARF TAMS Panel Uses 2 Different PeopleMeters Eurometer An Intrusive Installation... 27

5 4.4 Eurometer Function TVM 5 System Description TVM 5 Principles of Measurement SOFTWARE Pollux Processing Software: Overview Pollux Processing Software: Architecture Pollux Processing Software: Production Process Pollux Processing Software: Production Process Flow Chart DAILY PROCESSING CYCLE Data Collection Daily Polling Record Validation Report Log TVEvents and Independent Broadcast Log DAILY WEIGHTING AND PANEL BALANCING Daily RIM Weighting: Household Pre-Weight Daily RIM Weighting: Individuals RIM Weights and the Disproportionate Sample Universe Updates Role of Phone Counselors and the Monthly Update DAILY REPORTING CYCLE Overview THE EDIT RULES Current Edit Rules REPORTS TO INDUSTRY Weekly Report Monthly Cume Channel Reach Report COSTS REFERENCE FOR TV BROADCASTERS AND 3 RD PARTY PROCESSORS Overview LCA Levy Collection TV Broadcasters Requiring Additional Services Current TVEvents Channel List which the system is Identifying June rd Party TV Analysis Software Companies (incl. media Auditing Companies) GOLD STANDARD METHODOLOGY ALGORITHMS Overview General Principles in Regard to the Reach Algorithm... 44

6 12.3 TV Rating (also known as AMR, Average Minute Rating) ATV (Average Time Viewed) ATS (Average Time Spent) TVR (Total Television Rating) SHR (Share) RCH (Reach) Coverage [Synonyms: Reach 1+, Cumulated audiences] Effective Reach (n+, n-m) [Synonyms: effective reach, reach n+] POST EVALUATION GRP (Gross Rating Point) GRP per Spot Reach [Synonyms: reach 1+, cume reach] OTS (Opportunity to see) Cost / GRP Cost per CPP (Equivalent Cost per Point) ANNEXURES Annexure 1 Annexure 2 Annexure 3 Annexure 4 Recruitment Questionnaire Monthly Updates Questionnaire Recruitment Show Cards Monthly Cume Channel Reach Report Prepared by AGB Nielsen Media Research South Africa

7 Copyright The copyright of this report is reserved under the Copyright Act of the Republic of South Africa. No part of this publication may be reproduced by any means or used in any form by persons or organisations other than members of the South African Advertising Research Foundation (SAARF) without the prior written permission of the Chief Executive Officer of the Foundation. Any person who reproduces or publishes information from any SAARF report should make clear reference to the source of such information and should lodge a copy or details of the material reproduced or published with the Chief Executive Officer of the Foundation at the time of its release. No use or attempted use of data published in or derived from any SAARF report in a court of law is permissible. SAARF data may not be used in governmental proceedings except with the explicit permission of SAARF; such permission shall be sought in advance for clearly defined purposes separately in every instance. Exclusion of claims Despite all reasonable efforts to ensure accuracy in the assembling of information and data or the compilation thereof, SAARF is unable to warrant the accuracy of the information, data and compilations as contained on its website, as well as contained in any database, electronic report or publication for which it is responsible. Accordingly, no guarantee or representation is given by SAARF as to the accuracy or validity of such information, data or compilations. Users of its website or any of its databases or electronic reports as well as readers of any SAARF publication are deemed to have waived and renounced all rights to any claim against SAARF, its officers, servants or agents for any loss or damage of any nature whatsoever arising from any use or reliance upon such information, data or compilations. SAARF Objectives and Business The Main Business of the Company: SAARF is tasked with facilitating, coordinating and determining the joint industry research needs of its stake holders and to ensure that these needs are met within the limitations of the funding available for this activity. SAARF is an independent authority, acting as the custodian of the currency and a clearing house for industry research. To carry out this mandate, SAARF must keep track of local and international developments, ensure that new approaches are explored and developed, and that the necessary joint industry research programmes, which accurately reflect South African society, are implemented, managed and monitored. The Main Object of the Company: To provide tools for targeting and segmentation of markets as well as to establish, commission and manage comprehensive, valid, reliable, continuous media audience and product usage research, surveys, investigations and reports, that provide comparable multi-media and multi-product / brand usage information that reflect the totality and complexity of the South African society. The Ancillary Objects of the Company: 1. To co-ordinate joint industry research amongst the advertising, marketing and media industries. 2. To investigate any research techniques whether in practice or proposed and to establish the degree of validity and reliability of the results obtained thereby ; to seek improved methods in media audience and product usage research and to provide improved tools for targeting and segmentation of markets. 3. To act as a liaison between the advertising, marketing and media industry and universities, government and other official bodies involved in any form of education or research related to media audience, demographic and product usage research as well as tools for targeting and segmentation of markets.

8 4. To arrange seminars and courses directly or indirectly sponsored by the South African Advertising Research Foundation on any or all aspects of advertising, media or market research as well as tools for targeting and segmentation of markets. 5. To act as mouth piece of the industry on matters pertaining to media audience and product usage research as well as tools for targeting and segmentation of markets. 6. To promote and maintain fair, reasonable and proper standards of media audience and product usage research as well as for tools for targeting and segmentation of markets. 7. To maintain and augment a library containing information concerning media audience, product usage, and related research as well as on tools for targeting and segmentation of markets, and to make it assessable to members, students and the general public. 8. To do all such other acts, including the publication in print or electronic format, of books, memoranda journals, magazines, circulars, reports and any documents or databases as the South African Advertising Research Foundation may consider expedient to promote the interests of members. 9. Likewise to do all things and carry on any activity related, connected or associated with any of the above objects and purposes. 10. To finance the operations of the company by engaging in any lawful activity, which may generate funding for the company. SAARF Management Responsibility for the management of SAARF affairs rests with a Board of Directors, representing the members of the Foundation - The Marketing Association of South Africa; Print Media SA; the National Association of Broadcasters; Cinemark; Out of Home Media South Africa; the Association for Communication and Advertising and the Advertising Media Forum; OPA (Online Publishers Association) - together with the Chief Executive Officer, Paul Haupt, under the independent chairmanship of Sarel du Plessis. SAARF Councils The SAARF board of directors is the highest SAARF authority. It consists of directors nominated by all SAARF stake holders namely media owners, advertisers and advertising agencies. Much of the work done by SAARF is guided by a series of councils. The SAARF advisory council is involved with all aspects of the operation of SAARF. With the exception of the permanent staff members of SAARF who serve on the council, it also consists of representatives of the marketing, media, advertising industry and SAARF s contractors. The councils mandate is to advise the SAARF board on what research should be undertaken and in instances where the necessary authority has been delegated to it, to decide on details. In addition, a number of research experts serve on this council to advise on how research should be carried out. Because of the complexity of media audience and the product purchasing and usage research, a series of councils for each medium and for product research exists to advise the SAARF management on details of research in the relevant area. The SAARF Demographic Research Committee of the Bureau of Marketing Research (BMR) of the University of South Africa (UNISA) advises on demographic aspects of the population and annually adjusts the estimates of the size and composition of the South African Population, which are used in the SAARF surveys. The SAARF board and all the SAARF councils operate on a voluntary basis. SAARF Website address To view further information regarding SAARF, please log on to our official website:

9 1. Brief History of SAARF TAMS 1.1 Before PeopleMeters Prior to TV audiences being measured via meters in South Africa, quarterly TV Diaries were used in the same way as radio diaries are used to measure radio today. During June/July 1986 SAARF conducted discussions with the principle hardware manufacturers and also with suppliers of Peoplemeter research in the USA, the UK and Europe. SAARF decided that existing systems could cope with the requirements of estimating the size of TV audiences rapidly, reliably and validly using electronic meters. Significantly, SAARF also came to the conclusion that further technical improvements of a revolutionary nature were unlikely in the short to medium term st SAARF Tender and ACNielsen Dis-Invests 1986 SAARF prepared a comprehensive specification for the AMPS Meter System [now the SAARF Television Audience Measurement Survey: SAARF TAMS ] and the contract was awarded to ACNielsen during the mid 1980 s. However, after political unrest during 1986, ACNielsen dis-invested and IBIS, the company that reverted back to ACNielsen a few years later, took over their local interests, including the television metering contract. Because of international sanctions, SAARF and IBIS had no other choice but to develop a meter locally, and the software to operate it. 1.3 Mark 1 PeopleMeter 1989 The initial hardware could only be used in households with mains electricity and a telephone served by an automatic exchange: because of a relatively low penetration of telephones in black households, this Mark 1 meter panel therefore consisted of White, Coloured and Indian TV households with mains electricity and automatic telephones. As a result the total TV coverage was as follows:- % Whites 81,5 Coloureds 62,8 Indians 75,0 Blacks 0,0 The first weekly report was cleared for release in July Industry Acceptance This new TV currency was quickly grasped by the media planning industry and the broadcasters. Soon SAARF was asked to look at extending the AMPS Meter System to black TV households. 1.5 Mark 2 PeopleMeter 1991 In 1991 the Mark 2 meter was locally developed. This could operate not only with automatic telephone exchanges but also via radio transmitters in households without telephones. The daily viewing data in non-telephone households was transmitted by radio to a host meter in an automatic telephone household. The host telephone line would then send the viewing data for both the host and any linked non-telephone household. The host meter could be as far as 25 km away, depending on the terrain.

10 1.6 Black Market Meter Panel 1991 The Mark 2 meter was an important breakthrough as it enabled SAARF to go ahead with the establishment of a Black meter panel. In addition, SAARF was able to extend the Coloured, Indian and White coverage to include households without automatic telephones. 1.7 Limitation of Coverage However, in the case of Blacks, a decision was made to limit coverage to metropolitan areas, cities, large towns and small towns. This was due to the low incidence of Black households with TV sets and mains electricity in villages and rural areas, as well as the practical difficulties of finding suitable automatic telephone households to act as hosts. 1.8 The Next Generation of PeopleMeters 1998 By 1998 it had become clear that the Mark 1 and 2 meters had reached the end of their usable lifespan. Age, as well as damaging electrical storms, had taken its toll and the declining number of usable meters was inhibiting the number of reporting households. The problem was made worse by ACNielsen s inability to source spares for the old technology meters. Also, Y2K was looming, and this was in any event going to force ACNielsen to move off the mainframe software to Y2K-compliant PC-based software. A decision was taken to switch to more modern meters in the awareness that they would probably have to be sourced overseas. The whole process of moving off the old meters and changing to PC based software was helped by the new dispensation in South Africa, the dropping of sanctions and the return of ACNielsen, who bought back the IBIS operation. 1.9 Eurometer 1999 By early 1999 the necessary desk research had been completed and the decision taken to use the Finnish Eurometer from Mitron and PCTips software from ACNielsen s Systems Asia Pacific to collect and process the viewership data from the SAARF TAMS panel PCTips and Eurometer Panel 2000 By September 1999 ACNielsen had moved off the mainframe software and migrated to PCTips at this time the old Mark 1 and 2 meters were still in use. At the beginning of 2000 a separate panel of Eurometer households was established for monitoring purposes. By mid March 2000 this panel of 200 reporting Eurometer households was merged into the panel of existing Mark 1 and 2 households. At this time the old meters were failing quickly and by September of 2000 the decision was taken to remove all the old meters and to equip the whole panel with Eurometers By early 2001, there were no old meters left in the system Measuring Digital Satellite Channels 2001 The advent of digital satellite TV posed worldwide measurement problems. This was new technology, and the developed world was struggling with the challenge of measuring this service. SAARF took an early decision that SAARF TAMS would not measure satellite households in South Africa until a technical solution had been found. Consequently, the universe definition specifically excluded satellite households up to the end of The problem was that as DSTV became more popular, the SAARF TAMS universe was becoming more and more skewed away from the real TV universe in South Africa.

11 1.13 Measuring DSTV Satellite in Total During 2000 it was realized that DSTV could be measured as a single channel by recognizing the DSTV frequency in a satellite household. Establishing this undifferentiated data from SAARF TAMS households with satellite reception was termed Phase I, and was completed by week 01 of Phase 2 was achieved in July 2001 when ACNielsen was able to separate out the various component channels making up DSTV, in order to merge the satellite ratings of all of the terrestrial channels with their terrestrial ratings. Thus the MNET rating became a composite of the satellite rating added to the terrestrial rating, and so on for SABC 1, SABC 2, etc Measuring DSTV Satellite by Channel 2002 Phase 3 was achieved when the more robust commercial DSTV channels (those with larger audiences) could be reported on separately. This took place from and including week 13 of OATS Augmented Satellite Panel 2002 Since satellite households were a small percentage of the total SAARF TAMS panel [11,4% January 2002] few channels were sufficiently robust to be separately reported. Hence, in July 2001, Oracle Airtime Sales implemented ACNielsen s proposal to augment naturally occurring satellite households on the panel, with a booster panel of randomly located satellite households to create a satellite sub-panel of 300 households. During processing the enlarged DSTV sample was down-weighted to its natural incidence Nielsen Media Research Established By May 2002 ACNielsen and Nielsen Media Research had merged into a single global media measurement company, under the original name of Nielsen Media Research Only Multichoice DSTV Measured The current SAARF TAMS Satellite panel refers to Multichoice DSTV households only [this is now in the process of changing to all satellite households, as per SAARF TAMS Council decision]. The universe excludes other satellite delivery systems [such as Sentech s Vivid system]. The SAARF AMPS survey is the establishment survey for SAARF TAMS, and is used to define the SAARF TAMS universe, and when this is done only Multichoice DSTV households are allowed into the SAARF TAMS universe at this stage SAARF Universal SU-LSM Finally, the old 8 point LSM scale from 1993 has given way to the new 10 point SAARF 2001 Universal LSM scale. Over the years SAARF TAMS and SAARF AMPS had diverged in terms of their LSM definitions. SAARF TAMS is based on a survey of households; therefore only LSM variables based on household and not personal variables could be used. The last time that the SAARF AMPS LSM definitions had been defined in terms of household variables was in 1993, for this reason the SAARF TAMS LSM definitions remained based on SAARF AMPS Now, the SU-LSM definitions have reverted back to household criteria only, and therefore there is again complete correspondence between SAARF AMPS /RAMS and TAMS in terms of SU-LSM s. The 10 point SU-LSM scale was launched with the release of SAARF AMPS 2001A Merger between AGB Italia and Nielsen Media Research and the SAARF TAMS Contract of 2005 The previous SAARF TAMS contract was to terminate in March Prior to that happening SAARF issued a 6 year tender both locally and internationally. During this period, Nielsen Media Research and its major international TAMS competitor, AGB Italia, concluded a deal to merge their TAMS interests, outside of North America.

12 The new company, AGB Nielsen Media Research is the major international research company specializing in and focusing on measuring TV audiences. AGB Nielsen Media Research operates in 30 countries around the world, South Africa being one of these. The new company entered the race for the new SAARF TAMS contract and this was awarded to AGB Nielsen Media Research during This contract will terminate in March Recent Changes to SAARF TAMS Panel The new SAARF TAMS contract and the availability of world class TAMS hardware and software made significant changes possible to the way that the panel is managed and the way the data is reported Metering System During the last half of 2005, the AGB TVM5 meter was installed in all DSTV satellite households on the panel. This meter was probably the most advanced PeopleMeter in the world at the time, and was specifically chosen to meter the more advanced households in terms of their TV viewing equipment, the availability of digital satellite and also the myriad of ancillary equipment such as PC games, VCR s and Video recorders Independent Broadcast Log During 2005 the proprietary TVEvents system was installed. This system manages the download and capture onto computer of all commercial TV broadcasts the system runs 24x7, as all programmes and advertising spots must be captured into the integrated database for a complete independent broadcast log. Once the transmissions are captured, an independent broadcast log for each commercial channel is created, categorising all programmes and the contents of each advertising break, along with the material flighting code of each spot. A major advantage of this system is that it contains Automatic Spot Recognition. As every new advertising spot is picked up by the system, a digital audio and visual signature is created for the spot. Every time this spot is flighted on any TV channel, the ASR recognizes the spot and the system then includes the new flighting of the spot into the database of transmitted spots, automatically. The end result is an independent broadcast log which is the guarantee that each spot has either been broadcast correctly, not broadcast or not properly broadcast [interrupted transmission etc.]. TVEvents went live in March Move to Random Iterative Method (RIM) Weights Also during 2005, at the urging of SAARF, SAARF TAMS moved off the cell weight methodology to the RIM methodology, for the daily weighting of the SAARF TAMS panel data. Cell weights had traditionally been used in the research industry as this methodology was sound and had the advantage of requiring little computing power at a time when computers were not powerful. The major disadvantage of the Cell system is that it cannot re-weight a substantial set of variables since sample sizes of the individual cells reduce very quickly as the number of variables increases. With the coming of powerful PC s and the need to re-weight samples by a substantial list of variables, the RIM methodology recommended itself. The methodology sequentially iterates through the various variables and re-weights these, coming ever closer to the best weights for the sample for that day. The iterative process stops when the fit is closest to perfect.

13 By early 2005 and after substantial testing, the SAARF TAMS data was weighted daily by RIM methodology Move to Overnight Delivery of Ratings During the 1970 s and 1980 s internationally, the switch was made from the 7 day Diary to measure TV, to a PeopleMeter system. Almost all countries moved directly from the 7 day Diary to the delivery of Overnight TV ratings. South Africa did not, we moved to a PeopleMeter system in 1989, but issued the data in weekly tranches, not Overnight. The reasons were mainly due to indifferent telecommunications infra-structure [requiring more than 1 day to recover the data from panel households] and the problems in maintaining panel households in a large country such as South Africa [i.e. distance and time required to visit households]. However, it s important to realise that whilst the data was issued weekly, in all other respects it was Overnight data i.e. the data was processed daily and each day s data was kept separate and not averaged into a week. In other words we were issuing Overnight data, released a week at a time. The new TAMS contract changed all that: Advances in hardware and software made Overnights possible. GSM polling technology was much better than copper landline (GPRS proving even better than GSM). Broadcasters wanted Overnight data to track and respond to competitive programme lineup changes. Ad agencies wanted to track the progress of their campaigns and adjust advertising weight as the campaign progressed. The market wanted to respond to competitive advertising activity much quicker than possible under the weekly data regime. Since July 2006, the TAMS data has been issued on an Overnight basis. It s important to understand that Overnight delivery is not guaranteed: there have been instances where data has been withheld for 1 day, because polling was not entirely successful [because of something out of the control of the contractor] resulting in a low sample and possibly skewed data. Leaving 1 day extra to poll late reporters brings the sample back up to acceptable levels, therefore no data is lost. (See 3.55 Overnight Viewing Data on page 18) TV Broadcast Logs The move to and availability of Overnight ratings lead to another improvement in the TAMS service to the industry. The post transmission as run broadcast logs were available too late for Overnight ratings delivery. Two major things are required in delivering a TAMS service: the audience ratings of the various TV channels, and the database of events that the channels broadcast, i.e. spots, programmes, sponsorships etc. When these two things are merged, the data user then knows what the channel audience was at a point in time, but crucially also what the audience was watching at that time. Since 1989 the Event database was provided by the broadcasters and this was manually captured by the contractor. In other words there was no independent audit of the veracity of the Event log or how accurate in time, it was. More recently, it became clear that the major advertisers [and their media auditors] wanted an independent log of events, so that they knew that spots were in fact broadcast, and that the independent broadcast log verified that the spot had in fact been transmitted. The output of TVEvents is therefore an independent and verified broadcast log, correct to the milli-second.

14 Introduction of the Rural Component to the TAMS Universe There had always been one omission from the SAARF TAMS Universe, rural Black people. The reasons were that the telecommunications infra-structure was neither robust enough nor widespread enough to support daily polling of the market. Also, initially, the incidence of in-home TV sets was not widespread enough, nor was the provision of mains electricity, which PeopleMeters require to operate. By 2005 it was recognized that all of the problems had now disappeared. GSM mobile technology made daily polling possible, Eskom had done a superb job in electrifying the rural areas, and therefore the incidence of in-home TV sets was strongly increasing. 212 rural Black households were therefore recruited and installed, with the objective of having 180 good reporting households, on a weekly basis. The viewing data from this new market was extensively tested, found to be good, and the data was released to the industry from early April Because the rural black households are under sampled [lack of choice of TV channel justifies a smaller sample] the data are weighted back up to represent the universe accurately Inclusion of 4-6 Year Panel Members Since 1989 the SAARF TAMS panel had included all members in panel households from 7 years of age. Children younger than 7 were captured in the processing software, but not for their viewing data but rather to identify their mother. Various SAARF stakeholders indicated that viewership 4-6 year olds was also needed. On investigation it was found that the international protocol was in fact to measure all household members from 4 years and up [The contractor supplied data to support this]. The decision was taken to also include the viewership of these children in panel households in the viewing data. Training and counseling of all panel households who have young children commenced, their demographic data was verified on the household master database and demographic member data was updated in each PeopleMeter in the affected households. Viewing data from the 4-6 year olds was thoroughly tested, and over time found to be good and representative. As from early June 2008, the SAARF TAMS Universe was updated to include 4-6 year olds and the viewing data from the panel included all viewing data for members 4+ years of age.

15 2. Current Methodology and Processes Universe The daily SAARF TAMS data comprises adults 16 years and older, and children in the panel households between the ages All panel households must be private with Eskom mains electricity and a working TV set/s. The panel is fully national and reflects the TV viewership of all people in private households in metro areas, cities, towns, villages and rural areas. Out-of-home TV viewing is specifically not covered since the meter is not portable, as is the case in all international TAMS panels at this time. Both telephone and non-telephone households are covered, each household is polled for the daily TV data either by Telkom landline, FastNet or GSM / GPRS cellular. 2.2 Sample Design The SAARF TAMS panel sample is drawn via an area stratified probability design, names and addresses as well as demographic data are sourced from AMPS which is the establishment survey. Substitution (with alternate addresses) is necessary in the following instances:- Refusals TV equipment beyond repair Dwelling structure precludes the installation of an outside aerial Electricity supply erratic or below required voltage In these instances the nearest qualifying household in the area is recruited. When recruiting a household, certain household information and demographics of all the household members is collected. A suitable adult household member is interviewed for this purpose. Constant updating of the Pollux household master database takes place through collection of required personal information during servicing visits or phone counseling. In addition, systematic updates of the household data on one twelfth of the panel households takes place monthly on a rotating basis, so that at any point in time household information is never older than one year. The Recruitment Questionnaire 1 and the Monthly Update Questionnaire 2, together with the Recruitment Show Cards 3 [the Update Show Cards being essentially the same], and the Counseling Report 4 are attached as annexures 1, 2, 3 and 4. During June 2009, it became apparent that panel household demographics were not being updated quickly enough, particularly in times of economic change. Consequently, the contractor has recommended interviewing the entire panel every six months. This process is being piloted for CATI, the assumption being that ± 20% of households will still require a personal, face to face interview. 2.3 SAARF AMPS : the Establishment Survey Every television panel worldwide requires an establishment survey which estimates the universe and sets down the parameters under which the sample is designed. SAARF TAMS uses SAARF AMPS as its establishment survey. This is released every 6 months and the SAARF TAMS Universe is adjusted each time, so remaining in sync with SAARF AMPS [i.e. the TAMS Universe estimate from SAARF AMPS ].

16 2.4 Coverage of Satellite Households 2009 Currently, the satellite TV viewership data reflects the DSTV Multichoice platform only. All other satellite viewership is specifically excluded, and is also not included in the universe i.e. the satellite universe is slightly smaller and the nonsatellite universe slightly larger than it would be if all satellite broadcasters were included. Currently, there are satellite households installed on the panel, of which over 90% are good reporting households on a weekly basis. The above estimates include the SAARF DSTV panel of 173 installed DSTV households. SAARF has taken the principal decision that the SAARF TAMS panel universe will be expanded to include all satellite broadcasters, thereby meeting its remit to establish and maintain the buying and selling currency for all TV broadcasters. The satellite universe will therefore be expanded to include the Sentech Vivid platform and the other new satellite broadcasters as they begin broadcasting [Telkom Media, On Digital media and Walking on Water]. This means, in effect, that the universe estimates will be changed to reflect those universes of households that have access to any satellite broadcaster in South Africa. Whilst it has been the practice to use audited sales figures from Multichoice to update and adjust the SAARF AMPS satellite universe estimates, this practice has now ceased: firstly because SAARF AMPS is released twice a year [meaning that more regular SAARF TAMS universe updates are done], and secondly that having to use audited sales figures from the other satellite broadcasters simultaneously, would soon degenerate into an unworkable situation in the calculation of the total satellite universe. Currently, digital satellite is measured in the following two ways: Service Information [SI] protocols from the contractor are integrated into the firmware running on the satellite decoders. The SI outputs the channel being viewed to the PeopleMeter [explaining why the decoder must be connected to the PeopleMeter]. Secondly, Banner Recognition enables the meter to detect which DSTV channel/s the decoder is tuned to. The following Multichoice decoder types have had SI software written and are therefore able to be measured on the SAARF TAMS panel: 720i DualView 800 series Various 9 series decoders Banner Recognition will work on any decoder type. Currently there is no SI integrated into the PVR decoder, although this has been requested from the media owner, therefore PVR s cannot currently be measured. The plan is to include live viewing in PVR households and then to move to the UK VOSDAL methodology. This is Viewing on Same Day as Live, meaning that all viewership in the PVR households will be captured every day [including all live or delayed viewing up to the beginning of daily polling starting at 02H00 daily] and this viewing will be uploaded for processing into the daily viewing statistics for every day. 2.5 Panel Turnover and Renewal Panel members can gradually become dissatisfied with having a PeopleMeter in-home, also repeated counseling sometimes cannot change household behaviour to make them into a compliant household their TV viewing patterns and/or use of the remote control unit drift away from their real behaviour and they have to be replaced. Forcible replacement is not often required; most countries running a television panel rely on natural attrition to achieve renewal, stemming from:-

17 - Panelists moving residential address [possibly to a location where the household is not required]. - Removal and replacement of households demonstrating unsatisfactory button pushing compliance, even after counseling. - Removal of households showing generally poor co-operation. - Removal of households with a suspended electricity service of more than 6 weeks duration [households with suspended land line service can be provided with a GSM meter connection]. - Removal of household with chronically broken/damaged TV set/s. The SAARF TAMS panel has, in addition, an element of forced rotation, in that approximately 100 of the oldest households are removed from the panel per annum, and replaced by newly recruited households. 2.6 Growth of the SAARF TAMS Panel During early 1998, Tony Twyman, technical director of the Broadcasters Audience Research Board (BARB) in the UK, visited South Africa at the invitation of SAARF to scrutinise the SAARF TAMS operation. One of his findings was that the number of households actually reporting each week was too low to give consistently stable ratings, particularly when precise target market definitions were used. As a result plans were put in place to grow the SAARF TAMS weekly reporting panel from an average 550 to households. This was achieved as mentioned under Coverage of Satellite Households with the DSTV panel contributing a further 156 reporting DSTV households. With the inclusion of the rural panel, the current ideal panel as at June 2009 is made up as follows. These estimates are all weekly good reporting estimates. Households Representative Panel DSTV TOTAL Non-Satellite Satellite Totals

18 3. Definition of Terms Users of SAARF TAMS data should have a thorough knowledge of the following definitions, which will help them in analysis of the data. 3.1 Adult Currently an adult is a panel member of age 16 years or older. From AMPS 2009A an adult will be 15 years or older, and SAARF TAMS will follow suit. The ages of all persons in the panel are automatically updated by the Pollux processing software on the anniversary of their recorded birth dates. This is why all babies on the panel are routinely recorded on the household master database, so as they age and eventually turn 4 years of age, they automatically come onto the panel. 3.2 AQH Average Quarter Hour denotes the average audience across quarter hour unit intervals. 3.3 ASO Analogue Switch Off denotes the cessation of analogue broadcasts, to be replaced by digital broadcasts. In Europe, the most usual ASO has been set to be between 2008 and In South Africa it is currently set for ASR Automatic Spot Recognition, a proprietary solution within the contractors TVEvents system. The ASR engine scans real time digitized TV transmissions for commercials and stores these in a database of recognitions. The ASR works with both video and audio algorithms, to maximize the level of recognition. 3.5 ATS Average Time Spent viewing, the total sum of all recorded time spent viewing [e.g. minutes] across a given period [e.g. day, week] divided by the number of individuals in the universe / population being measured. Also known as ATV [Average Time Viewed]. 3.6 Audience Share Percentage of total TV viewing across a specified time interval of a given channel or programme. 3.7 Avail Availability of a commercial position / time slot in a scheduled commercial break on a given TV channel / network that is available for purchase by an advertiser 3.8 BARB (UK) Broadcasters Audience Research Board, a Joint Industry Committee responsible for TAMS data supply in the UK.

19 3.9 Bouquet A French term referring to the selection of general and thematic TV channels that are offered as a package to subscribing households Break Position Position of a commercial spot within a break e.g. first in break / last in break Cable TV System Wired transmission system serviced by a cable operator, who receives TV transmission centrally and relays them to subscribers via a cable headend across a cable network CAPI Computer Assisted Personal Interview, a method of conducting face to face interviews with the use of a PC for prompting questions and recording answers CATI Computer Assisted Telephone Interview, a method of conducting interviews over the phone with the use of a PC for prompting questions and recording answers Children A child is a person of age 4-15 [from 2009 this changes to 4-14], living in a private households as defined in Overall Methodology Universe. The ages of children on the panel are automatically updated daily, as with adults on the panel Churn (Rate) Index of turnover applied to commercial pay-tv systems as a whole or to channel packages, especially premium pay-tv services. An important indicator of pay-tv service performance, which is defined as the number of subscriber disconnections during a set period divided by the average number of subscribers during that period. This produces an annualized percentage Commercial and Programme Logs A record of all commercials and programmes transmitted by TV channels, this information may be provided by the TV channel itself or by an independent source; and is matched against the processed individual viewing statements so as to permit viewing figures for specific commercials and programmes. The logs may contain additional information such as typology codes [genres] and commercial flighting codes Concurrent Viewing Same individual registered by a TAMS meter as a viewer for two or more TV sets at the same time Consolidated Audience The consolidated audience is the sum of the live and time shift audience.

20 3.19 Consolidated Viewing Sum of live and all time shift viewing of television transmissions within a set time interval after the transmissions. (Note: definition allows for the same viewer to be counted more than once as a viewer of a transmission) Content Provider / Supplier A Company or organization responsible for creating content, whether movies, TV programming or interactive games applications, etc Cost Efficiency Financial performance measure of a schedule of advertising spots that is calculated by dividing the price paid, by the audience delivery with reference to the target audience(s) of the advertising campaign. The principal measures of cost efficiency are Cost Per Rating point (CPR) and Cost Per Thousand (CPT or CPM) Cover Measure of advertising reach, derivative terms like 1+ cover and 4+ cover denote the percentage of the target audience that has been exposed to a schedule of advertising spots at least a certain number of times (e.g. at least once, or at least four times, etc). (See also 3.83 Reach) 3.23 CPP / CPRP Cost Per Rating Point, the average cost of achieving one commercial rating point (i.e. advertising GRP) with a 30 second advertising spot (or other standard unit of airtime) for a given target audience. CPP s are widely used as a measure of the cost efficiency of advertising campaigns or for comparing price differences across different TV channels. The alternative widely used measure of cost efficiency is Cost Per Thousand (CPT or CPM) CPT / CPM Cost Per Thousand or Cost Per Mille: Television: The average cost of achieving 1,000 commercial impacts against a specified target audience, and usually adjusted to a 30 second advertising spot length. CPTs / CPMs are widely used as a measure of the cost efficiency of advertising campaigns or for comparing price differences across different TV channels. The alternative widely used measure of cost efficiency is to advertising Cost Per Rating point (CPR or CPRP). Internet: Cost of achieving 1,000 ad impressions Criterion of Viewing Panel members are requested to use the following definition of viewing to determine when they should log in or out: TV viewing is being in a position where you are able to see the screen of a TV set which is switched. (See also 3.62 PeopleMeter and TSV) 3.26 Daypart Division of the broadcast day constitution a single timeband (e.g. early morning: ; peak or prime time: , etc.). Most TAMS systems divide the day into eight dayparts (e.g. Early morning; Mid to late morning; Lunchtime; Early to mid afternoon; Late afternoon; Peak / Prime time; Late evening; Night). The dayparts correspond with broad variations in audience size and composition across the broadcast day.

21 3.27 Decoder Device (also called Converter) that decodes digital transmission signals and converts them for display on to the TV set DSO Digital Switchover, the process of replacing analogue with digital TV broadcast, for which prior switch-off is a prerequisite condition (see also 3.3 ASO) Disaggregated Viewing Data Same as Elementary or Respondent level viewing data - processed viewing data held at the level of individual respondents. The basic components of disaggregated viewing data are individual viewing statements consisting of complete time records across each broadcasting day off all viewing sessions by every family member and guest on all metered TV sets in the home Disproportional Sampling Sampling in which different sub-populations have different probabilities of selection, resulting in over-sampling / undersampling of some groups compared with others. Disproportional sampling by selected region(s) is quite common in TAMS research Effective Frequency The level of exposure frequency at which reach is deemed effectively delivered Effective Reach The number or percentage of a target audience that is exposed to a schedule of advertising spots at a set level frequency. This will typically specify a lower threshold value indicating the minimum level of exposure deemed as sufficient for effective advertising purposes (e.g. 4+ reach) and an upper threshold (e.g.12+ reach) above which additional exposures are considered as waste 3.33 Enforced (panel) Turnover Homes dropped from a panel on the initiative of the data supplier. The principal categories of enforced turnover are: turnover in order to preserve / improve panel balance, turnover to reduce panel age as a precaution against creeping panel bias and /or panel fatigue (with some TAMS systems setting a maximum length of service), or turnover due to faulty compliance with panel viewing instructions Engagement Broad terms / buzzword referring to user involvement in the media consumed Event Types Classifications of broadcast types found in the TV environment. They can be major, like commercial and noncommercial, or detailed like spot, sponsor, announcement or programme.

22 3.36 Exclusive Reach Individuals who, during the period of analysis, have watched only one channel for at least one minute Fragmentation (Audience) When broad television audiences break into smaller segments due to multiple viewing choices and niche programming that targets particular demographics Frequency The average number of times that members of a target audience who have been counted at least once as viewers to a schedule of advertising spots (or sequence of programmes), have counted as viewers Gold Standard Industry declared correct audience value, most often used in connection with viewer ratings. The object of laying down gold standards is to prevent disputes between buyers and sellers of commercial airtime over the correct audience figures that have arisen on account of them using different software yielding estimates of viewing Good Households or Respondents Every day the entire installed base of the SAARF panel is polled [to upload all meter data]. All of the polled households are then subjected to the validation & edit rules, to make absolutely certain that all households in the released data for that day, are good. IE: that they have been validated as providing good data that can be used to calculate the TV audiences for that day. It follows that all members of a good household, are also good. Its important to realize that a good household could be one with no viewing data for the previous day, as when that household is away on holiday, or simply when no-one viewed TV GRP Gross Rating Point, a unit of audience volume, which is based on the percentage of the target audience population that has viewed a transmission across a unit interval [based on the second by second audience]. e.g. a GRP of 10 implies an audience size that is equal to 10% of the audience being measured. Meanwhile the total GRP delivery of a schedule of advertising spots is equal to the sum of commercial GRP / ratings across all of the spots contained in the schedule. GRP totals or averages may be estimated for a wide range of different time periods, programmes or commercial selections. For the purposes of calculating GRP s and making comparisons, commercial GRPs for each advertising spot are typically adjusted to a standard 30 second spot interval [the so called 30 second equivalent]. (See also 3.82 Rating [TV]) 3.42 Home Language For the purposes of the SAARF TAMS panel, home language is a household variable, the classification resting on the claim of the responsible adult being interviewed: English / Other = English or some non South African language Afrikaans / Both = Afrikaans or Afrikaans and English equally Sotho languages = Sotho [North and South Sotho, Tswana, Venda, Tsonga] Nguni languages = Nguni [Zulu, Xhosa, Swazi, Ndebele or other]

23 When a guest registers viewership, the guest is deemed to be of the same language group as the household he / she is viewing in Household Purchaser Any respondent of either gender who claims to be solely or partly responsible for the day-to-day purchases of the household, is described as a household purchaser. There may be more than one person who could claim to be a household purchaser within any given household, although only one would be interviewed Housewife In previous SAARF AMPS reports, a housewife was defined as a female member of a household who claimed to be solely, mainly or partly responsible for the household purchases. Since SAARF AMPS 93 Female Housewives have been replaced by Household Purchasers. NOTE: Data users, however, are able to apply a gender filter to Household Purchasers HUT Homes Using TV, a term mainly used in the US that refers to the percentage of homes using (tuned in to) TV at a particular time (see also 3.78 PUT) Individual Viewing Statements Converted meter records (raw data) after data processing (i.e. after editing, validation and assignment of weights into summary statements of individual viewing over time. Each statement contains information concerning: start and end time of viewing session, identification of signal source and TV set being viewed, identity of viewer, coded demographic and other information about the individual s identity, and the individual s daily weight Processed individual viewing statements constitute the basic components of disaggregated viewing data Lead-in / Lead-out A programme preceding / following the time period of the programme being analysed Linear TV Umbrella term for a real time television services that transmit programme schedules. Almost all broadcast TV services count as linear TV, the main exception being Near Video-On-Demand (NVOD) transmissions of pay-per-view programmes over a large number of channel feeds. The alternative non-linear TV covers all on-demand programming, which is available to view at any time the user decides and not constrained by the real-time broadcast schedules. The linear versus non-linear TV distinction may also be applied to the nature of viewing, whether it is live (linear) or time shift (non-linear) Live Audience The audience of a commercial, daypart or programme at the time of its actual transmission.

24 3.50 Main Purchaser An adult claiming sole or main responsibility for the household shopping Makegood Extra advertising spots in compensation of a scheduling error / alteration by the TV channel or failure to deliver an audience guarantee target during the campaign period Media Dependent Media planning / buying specialist dependent on a creative agency with common ownership Media Independent Media planning / buying specialist on behalf of creative agencies under separate ownership Net Daily Reporting Sample (in-tab sample) The final sample yielding audience measurement data on a given day after polling and validation is equal to the in production sample, less households that were not successfully polled and households that were successfully polled but rejected during validation for technical or behavioral reasons. (See also 3.84 Reporting Homes) 3.55 OTS Opportunity To See, a commonly used term in Europe denoting frequency of media exposure. For television it is synonymous with frequency Out of Home Viewing All viewing that takes place outside the home (e.g. viewing at a friend s house or at a public venue, such as in pubs, clubs, hotels or work places). Out of home viewing may be particularly important for some broadcasters, e.g. niche satellite channels Overnight Viewing Data Viewing data delivered the next day [specifically the Viewing data and not the Event data]. The term is sometimes used in a narrower sense to refer to initial data output covering a restricted selection of key target audiences that is delivered to clients at the earliest opportunity the next day (i.e. during the morning). (See also table on page 37 for Release Timetable of ratings and events data) 3.58 Panel Balance Conformity of actual panel allocation to the ideal panel allocation for selected panel control variables Panel Fatigue Deterioration in quality of panel measurement that may occur as a function of the length of time that households / individuals have belonged to a panel (e.g. possible increased levels of measured nil or uncovered viewing as panel members cooperate less with their button-pressing instructions over time).

25 3.60 Panel Maintenance Panel staff functions of preserving panel balance over time and likewise accuracy of panel classifications and quality of panel response through the exercise of regular and periodic quality control procedures Patronage Patronage is the unduplicated proportion of a selected universe [such as adults, females etc.] which views a given TV channel during a specified time period for a minimum session of 16 consecutive seconds. [ Unduplicated : a person is counted only once, irrespective of the number of times they have viewed that channel]. Example: If the weekly patronage for TV channel Y for target market X is 43: then 43% of target market X viewed channel Y at some time during the week in question [each individual having at least 1 viewing session of longer than 15 seconds] Penetration (medium / channel) The percentage of people (or homes) within a defined universe that are physically able to be exposed to a medium / receive a particular TV channel PeopleMeter Generic name for the electronic measurement system which monitors the channel that a TV set is tuned to and the individuals present in the room while the TV set is switched on. Individual demographics are measured through a complimentary specialized remote control. (See also 3.25 Criterion of Viewing) 3.64 Personal Meters Portable metering devices that permit the measurement of an individual s overall exposure to Radio and TV (and possibly print media) both inside and outside the home. Designed to be worn or carried by selected individuals, Personal Meters can potentially capture viewing / listening in all types of out of home locations. The channel / station identification technique may be based on either audio comparison or recognition of a broadcaster code Person Viewing SAARF TAMS panel data is reliant on each member of every household logging into the system when viewing, and logging out when ceasing to view, this also includes visitors. In this regard see the comprehensive counseling procedures and data checks to root out bad button pushers [see counseling report annexure] Picture Matching Technique of signal identification in which the meter collects sample visual data from the image displayed on the TV screen, which it matches against an array of known signals from a central reference source in order to establish the identity of those measured signals Podcast Is a series of digital-media files which are distributed over the Internet using syndication feeds for playback on portable media players and computers. The term podcast, like broadcast, can refer either to the series of content itself or to the method by which it is syndicated; the latter is also called podcasting.

26 3.68 Polled Sample In production meter sample that has been successfully polled by the central processing base of the data supplier and is available for inclusion in the net daily reporting sample Polling Procedure for collecting data from meter panel homes, usually by means of a telephone call from the central processing base of the data supplier,which downloads in the early hours of the morning meter data from the previous broadcast day(s) via a modem connection with the central meter data storage unit in the home. Alternative cellular radio or one-way connection methods may be used in order to collect data from non-telephone households and are becoming increasingly common as an alternative to the standard fixed line modem connections. At the same time polling takes place in a few systems via telephone calls from the household to the central processing base of the data supplier rather than from the data supplier to the home Pollux Pollux is proprietary software of AGB Nielsen Media Research. The Pollux System is a compact software package to collect, validate weight and calculate television audience data. Audience events are collected by PeopleMeters positioned in participating families houses and connected to their television sets and telephones Position in Break Refers to the running order of a commercial break and where a specific advertisement fell within that e.g. 2/ Post (campaign) Evaluation Evaluation of a media schedule at the end of a campaign, with audience delivery data in the case of TV advertising PPM Portable PeopleMeter carried by the survey participant, this was originally developed to measure radio listener ship. There is an ongoing debate as how successfully it may also be used for purposes of measuring television viewing. These devices are not used in South Africa (see also Personal Meters) Primetime Evening day part associated with largest audiences, generally between and 23.00, though precise times may vary slightly by country Processed Viewing Statements Processed individual viewing statements after editing, validation and conversion of raw meter records into basic units, for calculating ratings and reach, as per the algorithms for assigning ratings Product Placement Paid for placement of an advertiser s product within a programme.

27 3.77 Profile (adhesion) The composition of a channel / programme audience, as defined by the proportionate contribution of different demographic categories against one or more variables, such as age and sex (e.g. a channel profile of 26% for Adults means that 26% of the total audience for that channel was aged 15-34) Programme Genre The classification of programme by type, e.g. sport, drama, chat show etc Programme Loyalty The measure of a constant programme audience across a series of episodes. Numerous different operational definitions may be used to define programme loyalty PUT People Using TV, a term mainly used in the USA to denote average percentage of people using TV across all channels within a set time period (see also 3.44 HUT) PVR Personal Video Recorder, a device also called Digital Video Recorder (DVR), which uses a hard drive to record and store digital video content. An important feature of the PVR is that it allows viewers to pause, fast-forward and rewind live programmes. Some PVR appliances also have the capability to suggest programmes for users by recognizing their viewing behavior Random Sample A sample in which every element in the population has an equal chance of being selected Rating [TV] In line with international convention, the old, but in every way identical, AMPS Rating [AR] has given way to TV rating. This is only a terminological and not a methodological change. For the purposes of this report the convention is to use TV rating or GRP. (See also 3.40 GRP) Definition of TV Rating A TV rating is the proportion of viewers, averaged across some time period [a commercial spot or break, a ¼ hour or a programme] and percentaged on the total number of individual panel members in the relevant group [e.g. all adults or females ]. Ratings are calculated by summing the products of numbers of viewers and relevant times spent viewing [within the period concerned] and dividing by the product of the total number of potential viewers and the duration of the period concerned. A TV rating is thus, in effect, a time-weighted average of audience size, indicated by an index which varies between 0 and 100 (the TV rating ). However, because visitors to panel households are included, a maximum rating greater than 100 is theoretically possible as visitors are not included in the base, but are included in the TV ratings. Visitors to metered households compensate in part for panel members who view in non-metered households. Informant numbers and universe sizes in the printed reports exclude visitors. The above explanation is summed up in the following formula, which is used to calculate a rating.

28 A practical example of this calculation is detailed below. 30 second spot with 4 viewers: Viewer Number Time Viewed (secs) Duration of Event (Secs) Σ Persistence Persistence is used both here and internationally, to suppress noise in the data without dispensing with good viewing data. Both the Mitron Eurometer and the AGBNMR TVM5 meter have been set up here to use persistence in an identical manner: when a person logs in for less than 15 seconds, that viewing is discarded, when a person changes TV channel for less than 15 seconds, that viewing is allocated to the TV channel viewed immediately prior to the change [this is known as the persistence setting, expressed in seconds], when a person changes a TV channel for a period longer than 15 seconds, the viewing is allocated to the new TV channel as from the time of the change. The change is only made after he / she has viewed for 15 seconds, but back-dated to where he / she was first found at the new TV channel Whilst the PeopleMeters continuously scan the status of TV sets and individuals in time intervals of less than a second, it ignores any viewing session of less than 15 seconds. The purpose of this filter is to eliminate momentary viewing that lacks persistence. The advantage of this international convention is to strip noise out of the data. At present, all ratings refer to real time, off-air viewing only. Viewing of VCR pre-recorded tapes or tuning through the VCR is not captured for inclusion in the published ratings. Those panel households having a Multichoice DualView decoder have their viewing on both decoder channels fully recorded on the panel Measurement of the PVR At this time there are no Multichoice PVR s on the panel, as Service Information [SI] protocols from the contractor have not been integrated into the firmware of the PVR, as has happened with the older model decoders. However, so called Banner recognition has been set up so as to measure live viewing on both channels on a PVR, whilst the Time Shifted Viewing cannot be allocated back to the channel where it occurred, the system does accumulate this TSV, and will report on total TSV. So the extent of TSV will be available, but not the source as at June Eight PVR households have been installed.

29 The plan is eventually to include live viewing in PVR households and to move to the UK VOSDAL methodology. This is Viewing on Same Day as Live, meaning that all viewership in the PVR households will be captured every day [including live or delayed viewing up to the beginning of daily polling starting at 02H00 daily] and this viewing will be uploaded for processing into the daily viewing statistics every day Reach (or Cover / Cume) The cumulative percentage or total (usually expressed in thousands) of a population that has been counted as viewers at least once during a specified interval. Examples of commonly used reach measures are TV channel daily / weekly / monthly reach, advertising campaign reach programme reach / programme series reach, daypart reach and so on. Commonly used synonyms are Cume and Cover. (See also 3.22 Cover) 3.85 Reporting Homes The number of panel homes that contribute to the daily ratings. Where failure to meet quality control standards or communications related problems occur, such installed homes are eliminated from the reporting panel (also known as Intab homes). (See also 3.53 Net Daily Reporting Sample) 3.86 Response Bias Bias in survey data due to measurement methods (e.g. recall bias in diary or day after recall TAMS research, or panel fatigue in PeopleMeter measurement) Return Path A data link that goes from a digital television system subscriber back to the system head-end. For a cable system, this may be the same cable. For a satellite or IPTV system, it may be a telephone landline or a GPRS link Return Path Data (or Click Stream Data) Any information sent via a Return Path, including system information such as subscription and pay per view request, also user interactivity, and potentially research information on audience size by channel RIM Weighting Random Iterative Method Weighting that puts selected non-interlocking and grouped interlocking variables in isolation through an iterative sequence of weighting adjustments. The sequence adjusts for each RIM in turn and then repeats itself as many times as is required in order to obtain a convergence, in which the sum of the weighted RIMs matches the target population estimates, or is as close as it is possible to achieve. (See also Weight Factor) 3.90 SAARF LSM s (Living Standards Measure) The set of standard questions applied to the SAARF AMPS sample to determine SAARF LSM status is similarly applied to the SAARF TAMS panel. All the variables in the SAARF LSM index are household variables, therefore all members of a specific household fall into the same SAARF LSM group. The SAARF TAMS panel SAARF LSM status is based on SAARF AMPS and is adjusted to always be comparable Sales House Separate organization or department within a TV channel responsible for selling commercial airtime.

30 3.92 Sample One or more elements (individuals or households) selected from a universe to represent that universe Sample Error Statistical measure of the possible deviation of a sample estimate from the true population value, assuming the sample to be representative of the population from which it has been drawn. The sample error is normally expressed as a margin of difference either side of the reported value within specified confidence limits (i.e. there is an X% probability that the true population value lies within Y units either side of the sample estimate ). Sample error is wholly distinct and not to be confused with sample bias, for which no parametric statistical assumptions can be made Sample Size The number of households or individuals selected for a research sample Sample Frame Source of addresses / household telephone numbers from which a pre-selected probability sample of identified individuals or households is drawn for interview. Usually the sampling frame is external (e.g. official population register of households / individuals, telephone list, list of postal addresses, electoral lists). Where such pre-existing sources do not exist, data suppliers may create their own sampling frames by means of conducting a micro-census or by carrying out a prior enumeration study of households in areas selected for the establishment survey fieldwork. Alternatively they may work without a sampling frame, but in that case it would not be completely random SI Service Information, proprietary information about the applications software in a set-top box, which can be used to facilitate the measurement of non real-time broadcast viewing (e.g. PVR use) and on-demand viewing, but relies on the co-operation of the service provider Slave Meter Meter installed on a TV set or other equipment that is subservient to another master meter in the household, which controls some of its functions (e.g. time synchrony) and collects data from it for temporary storage and delivery to the central processing base of the data supplier during polling Solo-Viewing Defines the condition where the members of a reference target are the focus of an analysis only when they are watching television alone Standard Error Standard deviation of the sample error distribution of a sample estimates standard Errors denotes the upper and lower bound margins of sample error that correspond with 95% confidence limits SAP Station Average Price, estimated cost of unit audience delivery on a TV channel based on advertising Cost Per Rating point (CPRs) or advertising Cost Per Thousand (CPTs or CPMs). SAP is normally calculated with reference to specified time periods (usually calendar months) and selected target audience(s).

31 3.101 STB Set Top Box, a device that receives, processes and converts incoming TV signals, for display on TV sets. The Set Top Box may be designed to receive signals in analogue or digital form and from cable, satellite and terrestrial sources Stratification A means of improving the quality of a probability sample by selecting sample elements according to population variables with known distribution profiles in order to determine a proportional or disproportional allocation of the survey sample. Examples of commonly used stratification variables are region, type of settlement and household size Target Market (or Target Audience) Core TV audience which an advertiser is aiming to reach, typically specified in terms of sex, age, socio-economic grade and housewife / main shopper categories. In many countries, airtime prices are negotiated with respect to specified target audience TSV Timeshift Viewing, later viewing of video recordings of live television transmissions (i.e. Playback of recorded live transmissions). (See also 3.25 Criterion of Viewing) Trick Mode Pausing, rewinding and fast forwarding live or stored video content TVEvents System TVEvents is the broadcast monitoring system of AGB Nielsen Media Research that offers a complete, highly reliable and easy to use suite of tools for the creation, collection and maintenance of high quality TVEvents databases Uncovered Viewing / Set Use Meter statements indicating that the TV set is switched on, but without any persons registered as present Unidentified Channel Viewing The viewing of an unknown frequency as ascertained by DFM [Direct Frequency Measurement] or tuner meters. The frequency is usually presumed to belong to a channel, although it may indicate a separate output source (e.g. a video games console) Visitors (also known as Guest Viewing) A visitor is anyone from the age of 4 upwards, who is not normally resident in the panel household and whose name is therefore not registered on the system. Every time a visitor logs in a new record is created. If the same visitor views in the same household on three different occasions, three different visitors will be recorded. Visitors are not and cannot be taken into account when reach is calculated. Data on each visitor s gender and age is recorded on each occasion, but they are automatically assigned the income and home language of the hosting household VOD Video On-Demand, a programme service where the content is not broadcast, but stored in a library which users can access on-demand. Typical VOD content offerings include recently aired television programmes (as in catch-up TV),

32 popular series, selected categories of thematic programming (e.g. music, children s programmes), and movies. There are three models of VOD: content-free VOD, pay-per-title, where the user pays an individual fee per programme or event, and subscription VOD, where the user pays a flat fee for access VOSDAL Viewing On Same Day As Live, when South Africa moves to measuring viewership in satellite households, the plan is to follow the UK and use the VOSDAL methodology i.e. to measure all viewership on each day [live and time shifted] up to the time that polling begins at 02H Wearout A level of frequency or a point in time when an advertising message loses its ability to effectively communicate Weight Factor Multiplication factor for converting the size of a sample to the population estimate for the survey universe. (See also 3.88 RIM Weighting) Within Programme Break Commercial break within a programme Zapping Flicking through different TV channels, often to avoid a commercial break Zipping Fast forwarding through recorded commercials when watching a home-recorded videotape.

33 4. Hardware Description and Operation 4.1 SAARF TAMS Panel Uses 2 Different PeopleMeters At the time of awarding the new TAMS contract, the Finnish Eurometer was the PeopleMeter in use in South Africa. For reasons of economy it was decided to retain the Eurometer for non-satellite households on the panel, as the Eurometer was suited to the less demanding measurement challenges in these households. The more advanced TVM5 PeopleMeter was introduced for satellite households, which is a more demanding measurement challenge. 4.2 Eurometer The hardware placed in each non-satellite SAARF TAMS panel household in order to measure TV viewing is the Eurometer from Mitron, a Finnish electronics company. The Eurometer Audience Measurement Equipment is widely used throughout Europe, South America and South East Asia. It is designed to be a flexible device that can handle the demands of different TV systems in the various countries where it is used. 4.3 An Intrusive Installation In South Africa, SAARF opted for intrusive channel measurement identification, as most broadcasters do not carry Channel Network Identification (CNI) within their broadcast signal. [ Intrusive refers to the presence of a device connected to the TV signal receiving circuitry] Channel measurement is achieved by having an AVPALPRE module installed inside the meter. This module comprises a tuner which is known as the pre-tuner, a tuner inserted in the signal flow in front of the TV s tuner and a Radio Frequency (RF) modulator. It has Antenna Input and RF Output plus one Audio Video (AV) output. 4.4 Eurometer Function The Eurometer takes over the tuning activity that would normally be performed by the TV. The household selects the desired channel using the Eurometer s Remote Control handset or the Local Keyboard [an option in households without a remote control unit], and the Eurometer does the tuning. The TV in turn is tuned to the Eurometer s RF modulator and the desired channel is transmitted to the TV set by the RF modulator. The AV option from the Eurometer may also be used for this purpose. All other off-air signals to the TV set are attenuated. The Eurometer measures the frequency of the terrestrial channel being viewed and expresses it as a channel number. The processing software, Pollux, identifies the broadcaster [TV channel] by using this channel number. The Eurometer is also equipped with a Network Identification Processor (NIP) board for capturing CNI codes. Broadcasters that carry CNI codes are identified based on their code rather than on their channel frequency. Currently, only MNET and CSN are broadcasting a CNI signal. The Eurometer s Remote Control handset is also used for logging individuals IN when viewing TV and OUT when no longer viewing. Each individual in the household is allocated a button on the Remote Control handset for this purpose, and his / her personal data is linked to that button number. The Eurometer normally sits on top of the TV set and displays the identity of the viewers as well as the program number of the channel being viewed. Messages such as birthday greetings can be displayed. This facility is also used to prompt viewers to log in. A feature promoting compliance is that if nobody is logged in, the household cannot make channel changes.

34 Each time an individual logs IN or OUT and / or a channel is changed, a real time record of the event is stored in the memory of the Eurometer. This memory has battery backup to prevent erasure in the event of a power failure. A Eurometer is connected to every TV within the household, and all other equipment linked to that set [for example VCR s or MNET decoders]. The convention is to connect up to three TV sets per household in non-satellite homes, and four TV sets in Satellite homes. When the TV is off, the Eurometer will display date and time. A MIAMOD Telkom-approved modem is installed inside the Eurometer, which is used to download the viewing data records every 24 hours to the collection computers at head office. Every night starting at 02H00, the collection computers dial out to each home where, by that time, the Eurometer modem will be in answer mode. When entering the call time window the modem disconnects the household s telephone. The Eurometer can detect if the household picks up the phone during this period and will then interrupt the time window or actual data flow / communication, so the household can make a call. In non-telephone households the viewing data records are down loaded via FastNet radio links. This is a packet radio system specifically designed for data transmission. A Radio Pad connected to the Eurometer is installed in each nontelephone household. During a set time window, the collection computers call up each installed Radio Pad and set up a communication link to download the viewing data records to the collection computers. 4.5 TVM5 System Description The TVM5 is a PeopleMeter designed and manufactured by AGB Lab, mainly to accommodate the measurements of digital broadcasting. The TVM5 System consists of a TVM5 base unit, display unit, remote control, transmission unit, and corresponding connections. The most important features of the TVM5 System are: non-intrusiveness, reliability, and ease of installation Display unit and Remote control The Display unit and the TVM5 remote control are the interface between the panel members and the meter. The Display unit has to be placed in the most convenient point that allows best visibility. Preferably on top of the TV set, this point also being most suitable for the RF module, RF communication link from the Meter to the Transmission Unit. The RF module is housed in the same plastic casing with the Display Unit Base Unit The Base unit is the core of the metering equipment that detects and stores all information concerning the viewing habits of the panel members. The TVM5 System requires that a PeopleMeter (Base Unit and Display Unit) be connected to all the TV sets and TV equipment (VCR, Digital Decoder, DVD player, etc.) in the home. The viewing data or statements generated, is stored in the memory of each of the PeopleMeters installed. This viewing data is progressively transmitted via the Display RF communication link to the Transmission Unit as mentioned above. The basic information gathered by the TVM5 is: On/Off time of the TV set the selected source of viewing i.e. TV set, VCR, Digital Decoder, or DVD player the selected channel Diagnostic Statements as motion detection, panel member interference with the installation, etc (The above category of statements are automatically detected and stored without intervention of the panel member).

35 identification of panel member viewer, arrival and leaving time of panel member viewer, identification of sex and age of visiting guest viewer, arrival and leaving time of guest viewer, and departure of the family for holiday (The above category of statements require input action from the panel member by using the remote control which is provided with each meter) Transmission Unit (TU) The Transmission Unit has two main functions: to connect to the TVM5 Meter(s) via RF communication, and to connect to Host Server by Land Line / GSM network / Internet Connection The TU polls the recorded viewing data from one or more TVM5 Meters through the RF communication link and also stores the polled viewing data. The TU has the ability to be called by / or to call the Host Server at certain predetermined time periods (time-slots) during the night, in order to download the viewing data statements. The connection to the Host Server is made by either Land Line / GSM network or Internet Connection (GPRS). 4.6 TVM5 Principles of Measurement Principles of measurement Due to the variety of platforms and the need to measure all possible present and future services, various methods are required. Currently the TVM5 uses 7 methods to identify channel and various states of the measured devices. The expansion slot and the modular nature of the TVM5 allow adding new methods in the future if so required. Some measurement methods can be added just by upgrading the software. The identification of a channel is performed in two steps: source identification (TV set tuner, Digital Decoder, VCR, DVD, etc), and broadcast identification (channel number, channel name or Service Information (SI) that is related to a channel) Some methods used for channel identification are given below. Some are used for both (Source and Broadcast ID) while some are used just to identify a certain state of a device Phase measurement Phase measurement method measures the phase between the sync pulses induced on the Sync sensor and the sync pulses extracted from the video signal coming from the internal tuner of the TVM5. The method exploits the fact that broadcasted video signals from various TV channels are not synchronized, except the one being viewed Audio comparison This method uses the audio signal picked up by the Audio sensor and compares it with the audio coming from the internal tuner of the TVM5 to find a match Electrostatic Video Average (EVA) comparison Every TV set screen based on the Cathode Ray Tube generates an electromagnetic field. The EVA sensor captures the electrostatic component of the electromagnetic field that is in correlation with the video signal of the displayed image. This signal picked up by the EVA sensor is compared with the video signal of the TVM5 internal tuner to find a match. Instead of using sensors, all the above signals may be derived from the Audio / Video (AV) output of the TV set.

36 4.6.5 Service Information (SI) This method of channel identification requires modification of the Digital Receiver / Decoder or Set Top Box (STB) software and sometimes hardware, so that the SI information will be available on an RS232 output immediately after channel change and from then on periodically. This obviously necessitates co-operation from the STB manufacturer and Broadcaster. The TVM5 detects the source and identifies the Channel generated by the STB present on the TV screen, by capturing the SI from the STB. The SI used to identify the Channel is known as a Triplet and is unique for every available Channel on the STB. The Triplet is a hexadecimal number made up of Network ID, Transport Steam ID and Service ID. Multichoice Decoders that have been modified to provide SI for the TVM5 Meter are: DSD 990 DSD 820 Dual View Image capturing or Banner Recognition Image capturing can be divided into two layers: the first layer is hardware based and is used to capture the whole image even the hidden lines, and the second layer which is software based, processes the captured image The fact that the second layer is software based, gives the flexibility to detect various signs and patterns in the image, which identify a channel. Character recognition is the most used image processing to identify the channel. Beside the character recognition, there are also other possibilities. Any permanent unique pattern present in the video image can be a potential for channel identification. Since no SI software is available for the latest Multichoice DSD 1110 Decoder the Banner Recognition method is being used for measurement purposes I 2 C Bus Monitoring (Sniffer Box) The I 2 C Bus Sniffer allows for non-intrusive monitoring of an I 2 C bus. The TVM5 Meter is equipped to interface via a Sniffer Box to monitor an I 2 C Bus on any device using that technology. Since the digital tuner used on some MNET terrestrial Decoders uses I 2 C bus technology for tuning purposes it becomes a convenient tool to measure encrypted MNET terrestrial viewership. Anytime a channel change takes place on the MNET Decoder the applicable tuning information is sent on the I 2 C bus. The Sniffer Box converts the information into a format accepted by the TVM5 Meter, which in turn relates the information to the off air Channel that the Decoder is tuned to. The Decoder On / Off status can also be derived from the I 2 C data. Decoders that can be I 2 C bus monitored are: Compact 9000 Delta 9000 Plus Comcrypt 4000I

37 5. Software 5.1 Pollux Processing Software: Overview PolluxCS System is a compact software package to collect, to weight and to calculate TV audience data. Audience events are collected by specific equipment positioned in participating families houses and connected to their TV sets and telephones. It has a Client / Server structure. The server side part of PolluxCS runs on the SCO Unix operative system; the client part of PolluxCS runs on Windows operating systems. Client / Server communication is achieved both via ODBC (Open Data Base Connectivity) and RPC (Remote Procedure Calls). All the databases, configuration and viewing files reside under Unix Server, the Windows Client is the user s interface to access the data. Pollux produces audience data every night from 11:30 pm to 6:45 am on the basis of parameters set by users with the Pollux System menu as follows. The production system contains the following elements: polling (handling of multiple lines) and line test, audience data validation and weighting, data quality control printouts, audience data printout tables, panel (household) management, and Pollux files maintenance. 5.2 Pollux Processing Software: Architecture The Pollux menu consists of screen based components for management and control. There are 9 main menus: 1. Recruitment: a set of programs that allows users to carry out control and administration of the recruitment phase. 2. Panel Management: a set of programs that allows users to carry out panel control and administration, engineering and equipment administration duties. 3. Polling: performs a set of polling analysis to check and test the quality of line dialogues and results of the overnight polling. 4. Production: runs production and analyses audience data quality. 5. Output: this option generates export files from Pollux production environment for data analysis and planning tools. 6. File Maintenance: a set of data entry menu programs that maintains the control and reference file used throughout the Pollux system. 7. Config: this menu provides a set of utilities to help the system administrator to manage the Pollux system. 8. Help: this menu provides a help online on most of Pollux features (user s manual, specific functions, etc). 9. Customize Menu: it allows to dynamically configure the main menu of the PolluxCS Client application.

38 5.3 Pollux Processing Software: Production Process The daily production can be monitored and checked by users controlling the log files, one for each phase. The daily production is split into three main phases: pre-poll, night poll, and post-poll phases Pre-poll The first phase needs to prepare the archives directory and parameters for the polling: the first archive directory, contains the list of families to be called with the telephone number, and the second file parameters, contains the list of parameters to be passed to the meter, home by home, during the parameters exchange dialogue in the poll phase Poll The second phase needs to call home by home, exchanges parameters and collects statements Post-poll The post-poll phase starts after the communication phase, it splits in more phases: format, production, and quality control Format The format phase converts the statements collected from the meter in a format readable by the pre-validation process Production The production phase runs: validation, creating channels viewing files, and weighting Validation The validation step first creates a binary file containing the image of HOUSEHOLD Master file; for each panel it displays the list of installed TV sets and then creates validated statements file for all elementary channels, checks inconsistencies in the data, creates a validation detail statistics file and eventually creates the Validation Summary and the Accumulated Viewing files Creating Channels Viewing Files The Creating Channels Viewing Files procedure creates the Networks, Groups, Supergroups and Totals viewing channels, grouping elementary channels created by validation according to the relations defined in the channels file Weighting The RIM weighting procedure tries to find a unique weight for each individual or home to ensure that the total of the weights of all individuals belonging to a category matches the universe for that category. The procedure is split into two phases: the first phase creates the individual (and home) weights file which lists each individual (or home), the initial weight (set to 1) and the list of RIM weighting categories that the individual belongs to; and the second phase of RIM weighting calculates the weight for each individual or home, balancing the universe

39 The RIM weighting process operates with a set of parameters, as specified by the contractor Quality Control The quality control phase runs programs to: build quality control poll files (telephone lines, panel calls), list daily and periodic poll statistics, dump the unidentified channels statements, list frequency states for each elementary channel, list high viewers, list nil viewers, list lazy viewers, list discarded by validation process 5.4 Pollux Processing Software: Production Process Flow Chart Pre-Poll Pre-Validation Pre-Poll 23:30 Main Poll Poll 02:00 Format Validation Preparation of raw data Production 07:00 Pre-Weighting RIM Weighting Remove temporary files Compress files File Maintenance 07:30 Daily reports QC reports Reports 08:00

40 6. Daily Processing Cycle 6.1 Data Collection All meters are programmed to communicate with the contractor collection system at least once every 24 hours [starting at 02H00]. To reduce the probability of a functioning meter not being polled, the polling system will continue to cycle through the panel until all households are polled or the window closes. There are four polling systems to cater for all infrastructures, i.e. landline, FastNet, GSM and GPRS. Once successful polling has occurred that household is not called again. 6.2 Daily Polling Record A report detailing successfully polled meters is produced every weekday morning. This is analysed on a rolling 3 day basis [except for Mondays and the first working day after every public holiday, when a period of more than 3 days could be covered]. Meters which have not reported for 3 days in succession are listed for Call Back and a Job Card is issued for a technician visit. 6.3 Validation Report Log Every weekday morning all data collected from every meter called, is processed through the Pollux Validation Module. A log file is produced that lists all rejected households, and households with warning messages covering No Meter Statement, No Data, Statement Overflow, Uncovered Viewing etc. Corrective action is taken immediately, either in the form of counseling the household concerned, or by issuing a Job Card for a technician visit. 6.4 TVEvents and Independent Broadcast Log The data captured per broadcast event is set out below: Date of broadcast; The start time of the event [the broadcast day is defined as the 24 hours starting at 02H00]. Therefore the latest time that can be recorded in any broadcast day is in fact 25H59M59S. The start time is recorded to nearest second; Duration of broadcast event [recorded to the nearest second]; Title of broadcast event [taken directly from the channel log], but limited to a maximum of 25 characters; In the case of advertising breaks and spots, additional information is captured from separate spot schedules to enable identification of tariffs, advertised brand description, etc; Each program is coded by program genre taken from a list of codes. Channel broadcast logs are sent to the contractor daily by all broadcasters, logs are normally received by midmorning following the day of broadcast, except in the case of weekends where the logs for Friday, Saturday and Sunday are received on Monday morning. The logs are used for verification of spot flighting codes. Weekly reports cannot be processed until all channel logs, for all days in the week have been received and coded. Close cooperation with broadcasters is therefore an integral part of the SAARF TAMS.

41 7. Daily Weighting and Panel Balancing 7.1 Daily RIM Weighting: Household Pre-weight Each day is treated as new survey by Pollux, and is re-weighted to population on a daily basis. The reason is that each day inevitably sees households coming back to report after a period of non-reporting, or reliably reporting households dropping off the panel because of meter problems, lightning strikes, telephone suspensions, or non-compliance issues etc. Each new day could therefore potentially see skews within the panel, because the sample is deviating from the ideal. The daily weighting process removes these skews from each of the weighting variables by weighting each RIM to the correct population size. The RIM method firstly weights the good reporting households to Household population; this is known as the Preweight as it this process produces a weighted household population base, with skews removed, which becomes the input to the second stage, the individual weighting procedure. Then Pollux performs 50 iterations of the data, to arrive at its final best fit for that day. The current RIM weight structure for households is: RIM 1 LSM 1-4 LSM 5-6 LSM 7-10 RIM 2 Eng / Other by Household size [1-3, 4-5, 6+] by non Satellite Main sample + Over sampled by Satellite Afrik / Both by Household size [1-3, 4-5, 6+] by non Satellite Main sample + Over sampled by Satellite Nguni by Household size [1-3, 4-5, 6+] by non Satellite Sotho by Household size [1-3, 4-5, 6+] by non Satellite Nguni Main Sample by Satellite Sotho Main Sample by Satellite RIM 3 Provinces All 9 by Total RIM 4 Community Size All 4 RIM 5 Platform No Pay, Satellite, MNET 7.2 Daily RIM Weighting: Individuals Current RIM weight structure for individuals is: RIM 1 Language [2 groups] + age RIM 2 Language [4 groups] + gender RIM 3 Non Pay TV + Pay TV Analogue + Pay TV Digital + Pay TV Both RIM 4 LSM + Homebound + Out & About RIM 5 All 9 provinces [LSM design: 1-5, 6, 7-8, 9, 10] RIM 6 Community Size All 4 [inclusion of Rural panel]

42 7.3 RIM Weights and the Disproportionate Sample To compensate for Oversample of the DSTV panel and the under-sample of the rural panel [as covered under Brief History of TAMS ], it is necessary to re-weight satellite and rural households, back to their proper representation. Consequently, the RIM weight structure specifically includes separate weights for satellite and non-satellite households, as well as community size. The weighting procedure for Black households who have satellite is slightly different because of sample size issues. However, with the growth in this market as a result of the Compact bouquet, this will be reviewed in due course. 7.4 Universe Updates The establishment survey, SAARF AMPS, is released twice yearly. As a consequence, SAARF TAMS universes for weighting variables are also updated twice a year, as soon as possible after the release of the latest AMPS. LSM s are also updated when variables change in SAARF AMPS. 7.5 Role of Phone Counselors and the Monthly Update Correct household composition records for every panel household, is critical to producing good, reliable data. This means that for every installed panel household, Pollux must have a complete and accurate record of every member of the household and all demographic information pertaining to all members. TAMS panels in first world countries do not have much contact with their households, as household composition does not change often or significantly. South Africa is different, and experience has taught us that composition changes often and dramatically. For this reason we employ phone counselors to contact each household regularly, and that we have done an annual face to face mini recruitment questionnaire with every household. It has become apparent that an annual household composition and demographic check is not sufficient, particularly in times of economic change such as South Africa is experiencing now. AGBNMR has recommended that each panel household is updated twice each year, approximately 80% via telephonic interview and 20% via a face to face interview (difficult or unreachable households).

43 8. Daily Reporting Cycle 8.1 Overview Since July of 2006, the SAARF TAMS viewing data has been issued on an Overnight basis. The Event data [comprising programmes and advertising activity] is issued the day after the Overnight data. In overview, the following takes place: the data are polled starting at 02H00 every morning, 7 days a week, polling finishes at 06H00 Post polling takes all the polled raw data, and runs the validation checks on all households, at this stage households are rejected e.g. new households on directory but not yet allowed into production, null viewers, constant viewing for 24 hours and unknown channel where a TV frequency is unrecognised by the meter. Polling reports are accessed to establish how many good reporting households there are [from Eurometer and TVM5 polling]: The performance statistics of each of the polling modems is checked The Pollux validation summary of rejected households is checked Then Pollux produces the industry output files next [TX3 and D_TX3] A converter then produces the ratings file Every weekday morning TVEvents supervisor completes closing the day At 10H00 daily the TVEvents data for a particular day is exported as RF, TEL and AIS files Batch files are run to convert the data for Arianna clients as well as the industry data The necessary files are then put up on the FTP server for clients to access. On the following days these files are released [to the bureau for processing] On these days: Monday Tuesday Wednesday Thursday Friday This data is released 10H00 Full Event for Thursday Full Event for Sunday Full Event for Monday Full Event for Tuesday Full Event for Wednesday Saturday & Sunday This data is released 10H00 Ratings for Friday Ratings for Monday Ratings for Tuesday Ratings for Wednesday Ratings for Thursday This data is released 11H00 Full Event for Friday Ratings for Saturday This data is released 12H00 Ratings for Sunday This data is released 13H00 Full Event for Saturday

44 9. The Edit Rules Current edit rules have been agreed by the SAARF TAMS Council together with the contractor. The critical decision was that the data must be manipulated or massaged as little as possible in the application of the edit rules. The edit rules being applied are therefore those that allow the raw data to come through with as little change as possible, with editing out or flagging only suspect viewing behaviour. 9.1 Current Edit Rules Uncovered Viewing : The permissible sum of all the gaps in viewing time [this is uncovered viewing where the TV set is switched on but no one is logged in] has been set to reject any household where total logged in time was less than 10% of total time switched on. The upper limit to the fatal gap and permissible logged-in percentage may be adjusted from time to time. (The South African situation is different from many countries in that TV sets are often left on for security reasons when panel members leave the house. There is also reason to believe that the TV is used to calm pets or babies.) All gaps [leading, embedded and trailing] have been disabled. All VCR tuning and any DVD playback is captured and recorded under VCR or DVD, but none of this viewership is included in the daily viewing statistics. Invalid Channel (unknown frequency): If a household views an unknown frequency, and viewership exceeds the set threshold, the household is rejected from the daily in-tab. Constant Viewing: If a household views TV for the full 24 hours, by one or more members, the household is rejected from the daily in-tab.

45 10. Reports to Industry 10.1 Weekly Report The standard industry weekly report, setting out TV viewing statistics for the week in question covering: Universe Headlines Top 10 Patronage Programmes Sample Quarter Hours N.B. this report has not been included due to large size, but can be viewed by going to and selecting TAMS(tv), then weekly updates Monthly Cume Channel Reach Report This report lists sample sizes and cume reach (from daily to annual) for the various TV channels including satellite. The prime reason for the report is to publish basic coverage statistics for channels too small to report, and also to track channels reaching the 60 respondent threshold for three consecutive weeks. This can be viewed by going to and selecting TAMS(tv), then Cume Channel Reach. (See also Annexure 4 for an example)

46 11. Costs Reference for TV Broadcasters and 3 rd Party Processors 11.1 Overview The situation with new satellite broadcasters [or broadcasters requiring additional measurement services], and the emergence of additional 3 rd Party TV analysis software companies has necessitated that SAARF looks at and amends the costs charged for either additional services to broadcasters or the provision of TAMS data to the 3 rd party suppliers LCA Levy Collection Since inception of the SAARF TAMS panel, the funding has worked as follows: The industry in consultation with the LCA [or its predecessors] together with SAARF determines the quantum of the levy that the LCA will collect. The levy is applied to above the line revenue [excluding sponsorships] that each TV channel generates; presently this is 1.0% of revenue. However, the methodology used to capture TV viewership has changed significantly over time, particularly with the introduction of TVEvents to create an independent broadcast log of events [programmes and advertising activity]. The TVEvents system requires hardware and data entry staff for every additional TV channel requiring Event data to be captured [if this does not happen, only ¼ hour viewing data is available to industry users]. There is a cost associated with this. This section therefore sets out these existing and any additional costs associated with measuring TV channels in South Africa, and the provision of the viewing and event data files to the industry TV Broadcasters Requiring Additional Services The collection and capture of TVEvents data for the creation of independent broadcast logs is fully funded via the SAARF TAMS contract, for existing TV channels. Therefore staff were employed and trained in using the TVEvents software for the creation of the independent broadcast log for the TV channels at the time TVEvents was installed. Any TV channels requiring their channels to also have an independent log created [i.e. those channels not included in the list below], will need SAARF approval and will also be invoiced by SAARF for the additional costs involved. SAARF has taken the principle decision that the total TV market will be measured, including all commercial but also the non-commercial channels, thereby reporting on TOTAL TV consumption in South Africa. This highlights the funding problem, that if a non commercial channel requires Event data to be captured, but the 1% levy yields no revenue, how is this funded? In principle the channel will pay an annual fee for the programme information.

47 11.4 Current TVEvents Channel List which the system is Identifying - June 2009 ID Channel Name 160 Cape Town TV 14 CSN 77 DSTV Action TV 2 DSTV African Magic 150 DSTV African Magic Plus 135 DSTV Al Jazeera 96 DSTV Animal Planet 140 DSTV Animax 153 DSTV BBC Entertainment 155 DSTV BBC Knowledge 154 DSTV BBC Lifestyle 41 DSTV BBC World 44 DSTV Bloomberg 11 DSTV Boomerang 138 DSTV Botswana TV 54 DSTV Cartoon 156 DSTV Cbeebies 63 DSTV CCTV4 106 DSTV CCTV9 58 DSTV Channel O 45 DSTV CNBC Africa 42 DSTV CNN News 134 DSTV Crime and Investigation 48 DSTV Discovery 131 DSTV Disney Channel 84 DSTV Dutch TV (BVN) 59 DSTV DW 151 DSTV e News 17 DSTV E! Entertainment 66 DSTV Ert 34 DSTV ESPN 149 DSTV ESPN Classic 9 DSTV E TV 133 DSTV Euro News 51 DSTV Fashion TV 91 DSTV Game Channel 12 DSTV Guide 27 DSTV Hallmark 98 DSTV History Channel 137 DSTV Home Channel 162 DSTV Islam Channel 158 DSTV Kowee 25 DSTV Kyk Net 94 DSTV Learning Channel 145 DSTV Magic World 3 DSTV MK89 5 DSTV MNET 141 DSTV MNET Stars 26 DSTV Movie Magic 1 29 DSTV Movie Magic 2 7 DSTV MTV Base 56 DSTV MTV Europe

48 148 DSTV Nat Geo Wild 50 DSTV National Geographic 152 DSTV Nickelodeon 143 DSTV One Gospel 60 DSTV Parliament 35 DSTV Racing Channel 65 DSTV RAI 68 DSTV Rhema 61 DSTV RTPI 130 DSTV Russian TV 159 DSTV SABC News International 4 DSTV SABC 1 1 DSTV SABC 2 8 DSTV SABC DSTV Saffron Indian Channel 21 DSTV Series 43 DSTV Sky News 139 DSTV Sony Entertainment 144 DSTV Soweto TV 30 DSTV SS Blitz 81 DSTV SS Max 142 DSTV Style Network 47 DSTV Summit 32 DSTV Super Sport 1 33 DSTV Super Sport 2 36 DSTV Super Sport DSTV Super Sport 4 37 DSTV Super Sport 5 38 DSTV Super Sport 6 80 DSTV Super Sport 7 86 DSTV TBN Christian Channel 28 DSTV TCM 73 DSTV Trace Channel 49 DSTV Travel Channel 99 DSTV TV5 French Channel 55 DSTV VH1 Music 161 DSTV Vuzu 10 DSTV Weather Channel 79 DSTV Zone Reality 9 E TV 5 MNET 4 SABC 1 1 SABC 2 8 SABC Soweto TV rd Party TV Analysis Software Companies (incl. Media Auditing Companies) 3 rd party TV analysis software companies are defined as companies who receive industry media consumption data from SAARF [i.e. AMPS, RAMS and TAMS ] and restructure this data and integrate it into their media analysis software, which is then made available to industry users [i.e. broadcasters, print and radio media owners, advertisers and advertising agencies]. 3 rd party TV analysis software companies would be Telmar, AGB Nielsen Media Research, IMS, Techedge etc.

49 When assessing the costs to be paid by these companies, SAARF took into account that they perform a valuable function disseminating SAARF s data, and distributing it widely to end users throughout South Africa, and in fact internationally, to the industry s advantage. Consequently, these companies all pay an annual fee to SAARF. This annual fee covers the daily data dissemination from the contractor FTP server, and other administration costs. However, with the increase in international usage of SAARF TAMS data, SAARF will now require a confidential listing of all clients that these 3 rd party TV analysis software companies make the SAARF data available to. The reason is that SAARF has a duty to the industry here to make absolutely certain that all levies have been paid and collected. It is also noted that clients who are users of SAARF data [i.e. broadcasters, advertising agencies and advertisers] do not pay any fees to SAARF, as this has been covered via the 1% LCA levy. Media Auditing companies that receive their own copy of the SAARF data are also invoiced annually by SAARF. The basis for this is that the Media Auditors do not perform the function of disseminating the SAARF data to end users, but rather integrate the SAARF data into their own in-house monitoring software which is not available to any end user, but rather derive revenue via expert analysis of the SAARF data for the furtherance of their own businesses.

50 12. Gold Standard Methodology Algorithms 12.1 Overview Following standard practice internationally where Joint Industry Committees establish Gold standard methodologies, in order to ensure that all 3 rd party TV analysis software providers supply software to their clients that will all give the same results for identical analyses, SAARF is setting down its Gold Standard methodologies here. The intention is that all 3 rd party providers will embed these methodologies in their software, thereby ensuring that analysis results are to Gold standard specifications General Principles in Regard to the Reach Algorithm Weighting Average Weights, the arithmetic average of each individual s weights are used for the date range in question. The advantages are that no richness of data is lost in the process, and that the universes remain constant & consistent. Average Weights are used in all calculations where reach is calculated. Questions Posed & Answers Provided: Question: Campaign / Spot by Spot Alignments Here we consider situations where the legitimate mathematical calculations do not appear to the lay user as correct or legitimate. An example is where the 1 st event in an Event by event report the cumulative reach may be defined in such a way that the weight used to calculate it, is the average of the whole campaign, whereas for the single event, it is the weight on the day of the event. Therefore cume reach & reach for the 1 st event may legitimately not be equal. This result may not make logical sense to the lay user, & these adjustments are designed simply to avoid these situations. Therefore, in general the following are made to be true The cume reach of the last event always equals the campaign reach. In the first event (spot), cume reach will equal the differential reach. For the final event, the frequency distribution will equal the frequency distribution of the campaign. For each event e, the frequency distribution of e will equal the cume reach Reach adjustments Negative Binomial Distribution is the standard in South Africa, and will remain so Mandated Viewing Threshold In South Africa there is no mandated minimum reach threshold ie: once an individual appears in the TX3 file for that day with associated viewing, that individual is counted into reach. In other words we have no minimum viewing threshold, such as 1 minute, that the UK or Australia might apply [here we should note that Australia has minute by minute viewing data not second by second as South Africa has]. In short, there is no viewing threshold to qualify for reach, in South Africa Required South African Configuration Post Evaluation Viewing Behavior Weight mode Average weights Average weights Single spot adjustment On / yes On / yes

51 Coverage Adjustment On /yes On /yes Effective Reach Adjustment NBD NBD Viewing Threshold Off Off FAQ s Inevitably questions will be posed, & answered. It seems most appropriate to keep these questions & answers in one specific place, rather than to spread them in different areas where it will be more difficult to find answers. This section will therefore grow through time. Question 1: Answer 1: Reach: Qualifying Respondents & Demographic Changes What happens when a person s demographic changes during the period of the of the reach analysis, & thereby falls out of the target market specification? Where Average Weights are used, on the days when the person falls outside of the defined target, that person s weight should be set to 0 [zero] Question 2: Answer 2: Question 3: Answer 3: Question 4: Answer 4: Question 5: Exact Age: Non-standard Age Specification In the case of an age target market being specified as non-standard, ie: 20-30, should exact age be used to select respondents, rather than an age group such as 15-24? Age groups cannot precisely determine qualifying respondents, therefore where specific nonstandard age groups are specified, the exact age of all respondents must determine eligibility. In the question above only respondents whose exact age is between years of age should qualify. Reach: Minimum Days to Qualify Is there a minimum number of good days that each respondent must achieve, in order to be included in the reach analysis? There is no set minimum number of days that each respondent must achieve [ie: be in the database as a good reporter] in order to qualify for inclusion in a reach analysis. If a respondent is only in the data on 1 day out of a 30 day reach analysis, this respondent is included. In the Average weights calculations, this respondents contribution to the reach calculation is proportionately diminished, & is therefore in its proper representation. Floating Point: Number of Decimals to carry In all intermediary & final internal calculations, to what precision must the software work? The methodology is Double floating Point, where each number is stored effectively to 17 digits, irrespective of where the decimal point occurs. Time Shifted Viewing: Calculation Description How is the quantum of Time Shifted Viewing calculated? Answer 5: Calculation Methodology: As used in all countries to December 2009, & all countries bar UK, 2010 Assuming an advertiser buys a spot on 12 April, essentially the advertiser needs to know the total audience that this spot delivers, including its Time Shifted Viewing component. Therefore the calculation of the total final audience is based on Transmission Date. This makes most sense. In addition, the 2 requirements below, should be met by the chosen methodology a.) The methodology chosen should avoid the re-issuing of previously released data.

52 b.) The live & VOSDAL audience [& respondent sample] of a spot, should not change through time. The process: On April 12, polling takes place, after all live & VOSDAL viewing has been captured in the meter. During data production early morning 13 April, weights are assigned to all qualifying respondents. During the next 7 days all additional time shifted viewing by respondents is captured but only of respondents good & in the viewing database for 12 April. Time shifted viewing by other respondents, not good on 12 April, is discarded. This time shifted viewing is added to the Live & VOSDAL viewing already issued for 12 April This methodology avoids a recalculation & re-issue of the original 12 April Live & VOSDAL audience. Also, 12 April Live & VOSDAL data is not re-released The final addition of all time shifted viewing to Live & VOSDAL audience on 12 April, gives the advertiser what he is looking for. All guest viewership is used, based on the Guest s weight, on day of Playback [guests are not identified therefore using any guest s weight for 12 April is not possible.]

53 12.3 TV Rating (also known as AMR Average Minute Rating or AR AMPS Rating) Average number of individuals who have seen a specific event or daypart. Formulas: Absolute (000s): Percentage (%): Where: i = individuals of the target weight i = daily weight of individual i viewing i = viewed seconds of individual i on program/daypart duration = duration of program/daypart Univ = daily Universe Calculation Rules: For calculations of TVR for a group of events or dayparts the duration of each event / daypart is considered in the calculation of the Average TVR (weighted by duration). The TVR% on a group of events is divided by the average daily universe (averaged by the number of daily events). The Average by duration is calculated as: 12.4 ATV (Average Time Viewed) Average number of minutes seen by any individual Formulas: Absolute (000s): Percentage (%): Where: Calculation Rules: Dur = duration of program/daypart Univ = daily Universe On a group of events or day parts it is calculated as average weighted by duration.

54 12.5 ATS (Average Time Spent) Average number of minutes seen by each viewer Formulas: Absolute (000s): Percentage (%): Where: Calculation Rules: Dur = duration of program/daypart On a group of events or day parts it is calculated as average weighted by duration 12.6 TTVR (Total Television Rating) Average number of individuals watching TV Formulas: Absolute (000s): Percentage (%): Where: Calculation Rules: i = individuals of the target weight i = daily weight of individual i viewing i = seconds of individual i with TV on duration = duration of program/daypart Univ = daily Universe On a group of events or day parts it is calculated as average weighted by duration 12.7 SHR (Share) Proportion of individuals viewing a program or day part on the total number of individuals watching TV at that time Formulas: Percentage (%): Calculation Rules: On a group of events or day parts it is calculated as the proportion of average TVR and average TTVR

55 12.8 RCH (Reach) Number of individuals having seen greater than 15 seconds of an event or daypart Formulas: Absolute (000s): Percentage (%): Where: Calculation Rules: i = individuals of the target weight i = weight of individual I having seen the specified viewing threshold Univ = daily Universe No threshold is defined The reach and universe are calculated with daily weights and universe. Average reach: the average reach is calculated as average by duration. In Reach and Frequency average weights are used Effective Reach (n+, n-m) [Synonyms: effective reach, reach n+] Number of different individuals having seen at least n or a range n-m tv-items of the schedule Formulas: Absolute (000s): Percentage (%): Where: Calculation Rules: i = individuals of the target n+ =at least n ; n-m=frequency corridor weight i = weight of individual having seen exactly n items Univ = target Universe No threshold is defined The reach and universe are calculated with daily weights and universe. Average reach: the average reach is calculated as average by duration In Reach and Frequency average weights are used.

56 12.10 GRP (Gross Rating Point) Total number of contacts Formulas: Absolute (000s): Percentage (%): Where: Calculation Rules: Country Specifics: i = individuals of the target weight i = daily weight of spot viewers Univ = daily Universe GRP is calculated with daily individual weights. The GRP of a campaign or group of spots is the sum of each spot s GRP (in absolute or percentage). -South Africa: GRP per Spot Average spot rating Formulas: Absolute (000s): Percentage (%): OTS (Opportunity to see) Average number of spots seen by the viewers Formulas: Absolute (000s): Calculation Rules: OTS is the proportion between campaign s absolute GRP and Reach Cost/GRP Cost per 000 Cost per rating Formulas: Absolute (000s): Percentage (%): Calculation Rules: For a campaign or group of spots the Cost per rating is the total investment by total rating

57 CPP (Equivalent Cost per Point) Equivalent cost per GRP Usage (Why): The Equivalent cost per rating point is the estimator of Cost efficiency of campaigns Formulas: Absolute (000s): Percentage (%): Calculation Rules: For a campaign or group of spots the Eq. cost per point is calculated as the total eq. investment by the total GRP