New BARB Contract – Update Presentation

Today’s seminar on the new BARB contract updated users on the findings of the parallel run and the expected state of play on 5 August.

Bob Hulks, who chaired the meeting, began by describing the session as “a midwife’s pre-birth statement on the state of the baby”.

Hulks set out the objectives of the parallel run, pointing out that at the time of the last changeover in 1984there was only one day overlap, and a lot of confusion and misunderstanding. The parallel run is intended to check that all stages of the operational systems work, and to test the new data formats and systems.

In addition, the old and new data must be compared, to enable calibration, and coverage and frequency has to be monitored.

The task force overseeing the parallel run (comprising the IPA, the ITV companies, BBC, the contractors and BARB) has been studying the establish- ment survey, installation rates, the weighting process and the pollingsystem.

Alan Yates of RSMB outlined the key changes to the establishment survey, including the new design,the continuous survey, and the increase to 43,200 interviews per annum.

The new panel will implement new recruitment methods, there will be an improved panel maintenance system and panel controls, new sample sizes, a disproportionate structure, and the latest peoplemeter system.

The data will include VCR timeshift, guest demographics (sex and age will be specified), and rim weighting is beingintroduced.

Yates went on to summarise the progress that has been made so far. The Establishment Survey has been running for almost a year (57,000 interviews), and a 78% response rate has been recorded. The new report schedule involves reports every three months (electronic and hard copy), with the first one expected to be available this month.

The current rate of acceptance to become a panel member is 40%, and this is expected to level out at 30%. So far, there have been low drop-out rates and this is expected to level out at 2% per month.

Yates concluded by stating that “we cannot claim that the panel will be in perfect balance by 5 August, but imperfections will not have a signific- ant effect on the results because of the new weighting system.”

AGB’s Brian Roberts began by once again describing how the new contract is split between AGB and RSMB, and summar- ising the nature of the changes.

The new panel will have almost 300 cable and satellite homes, and the Astra panel itself ceases to exist on 4 August. The new system will have a 4,435 panel, at the moment the panel is up to 4,055, and by August this is likely to have reached 4,432. Around 98% polling is being achieved every night during this parallel run period, and after two and a half months, 8 weeks of Database One and 7 weeks of Database 2 are avail- able, despite initial problems with the rim weighting.

Roberts rounded off his presentation with the assurance that “the panel is largely installed today, and will be virtually in place by 5 August.”

Tony Twyman of BARB looked at the differences encountered in the move to the new contract. There are differences in universes, weighting, consolidation, the addition of non-terrestrial channels, changes in the measurement of guest viewing, and the task force is currently trying to assess the scale of these differences.

Twyman was pleased, “and slightly surprised” to be able to assure those present that there have been no major changes in the recorded viewing levels: Weekly Hours Viewed, All Individuals

With regard to non-terrestrial viewing,Twyman admitted that so far this is under-reported because of the use of old targets; the issue will be examined in the near future.

There may also be a possible short-fall in the amount of guest viewing recorded, down to around 3.5%-4%. This may be because of unallocated button pressing (0.5%), loss of regular guests (1%), the deterrent effect of entering demographics. A ‘regular guest’ procedure is being gradually introduced: NEW PANEL INDEXED ON OLD PANEL All Individuals, 3 weeks ending 2 June

Twyman acknowledged that the panels will not be in perfect balance at the start of the contract, but a reasonable balance is expected after three months. At the moment there are too few one- person households, too many larger households, and too few young house- holds and old adults. However, all important dimensions which are not in balance are being addressed by rim weighting.

Twyman concluded by affirming that there are no major systematic method- ological differences or biases. There are more differences by area and sub- groups because of sampling variability. Also, guest viewing is lower than anticipated.

Young And Rubicam’s Jane Perry had the difficult task of trying to explain the new reach and frequency methodology.

At the moment, buyers and sellers use ratings as the trading currency, butthere is a general trend towards coverage, dictated in part by the growing complexity of the TV market.

The current coverage and frequency methodology is based on continuous reporters, but this leads to some bias, particularly over a period of time, and currently there is no weighting to re- adjust this.

There are also inconsistencies with “currency” ratings, and the current methodology won’t work on a disproport- ionate sample.

Instead of using continuous reporters, the new technique is based on reporters who are present on a given day in the campaign (many favour the first or the middle day), and their weights are taken for that day. Their viewing across the whole campaign is then examined.

By using this method, the maximum panel is encompassed, panel members are weighted, nett reach/coverage is marginally under-stated(by unique view- ing on missing days), spot ratings are under-stated (by panel members on missing days and guests) and gross TVRs are understated.

The reach and GRPs under-statement is addressed by the use of negativebinomial distribution. This is applied to sample frequency so that those registered on lower frequencies are cut slightly, whilst the number recorded as seeing a higher number of spots is bumped up slightly to compensate.

As a result, there is far less bias in the sample used, and an element of guest viewing is included in nett cover. Gross TVRs now match the Gold Standard.

Hulks rounded-off the presentations with the assurance that a currency convertor will be released before 30 July, based on impacts delivered and encompassing 9 audience categories. This will compare old panel viewing with viewing on the new panel, new panel consolidated viewing, new panel live viewing indexed on old live view- ing, and new consolidated indexed on old live.

Tuesday 6 August – the first (live only) { hour overnights will be available Tuesday 13 August – the first full reporting on spots for Monday 5 August will be available Tuesday 20 August – the first full reporting for w/e 12 August, on spots { hours and programmes will be avail- able. Three printed reports will also be issued – the BBC Grey Book, the Network Report, and the Astra Report (now based on the main panel)

Press releases will be available from 20 August, providing consolidated and aggredated figures.

During question time, Twyman acknowledged that the new panel is recording around 1% more live viewing (including the guest viewing short- fall), and around 1.5% for time-shift. This is spread fairly evenly amongst the individual channels.

RSMB’s Steve Wilcox confirmed that it is possible to analyse coverage and frequency for individual satellitechannels on a network basis.

There followed a somewhat heated debate on whether there should be an industry- agreed day during campaigns which would always be the basis for coverage and frequency.

Perry and Wilcox argued for flexibility over this matter, because for some campaigns, to use the first or middle or any arbitrary day would be unsatis- factory. Others foresee confusion amongst buyers and sellers if there is no widely-accepted rule.

The general concensus amongst the speakers seemed to be one of ‘it willbe all right on the night, or at least it will be fairly OK within about three months’.