Multichannel TV must trigger shake-up of the ratings system

When Lord Leverhulme uttered his famous dictum about not knowing which half of his advertising money was wasted, he had never heard of cable television and he had probably never envisaged his products being sold in Formosa, or Taiwan as it is nowadays known.

But he would still not have been amused – any more than current Unilever executives – by a tale told in the company’s staff magazine by Mike Kirkham, managing director of the research company Taylor Nelson AGB.

“In February and March 1996 a staggering number of 10,000 Unilever commercials were transmitted on cable TV stations in Taiwan. According to the official research data, all 10,000 spots achieved a rating of zero!”

The fabled nil rating, virtually unknown until the launch of Channel 4 some 15 years ago, is now no longer a disgrace but common-place – indeed, for some stations, a whole week’s output fails to trouble the scorer. “Does that really mean that no one saw the campaign?” asks Kirkham.

The answer, of course, is no – though if some of the cable stations in Taiwan are anything like some of those in the UK, a few of those zero ratings may indeed represent a mere handful of viewers.

But the problem is going to get worse. The launch of digital television next year will give most viewers a choice, initially, of 30 channels through terrestrial transmission and hundreds through satellite and cable. The number of programmes attracting zero ratings is likely to escalate alarmingly.

It is a conundrum that Taylor Nelson, as the main research contractor for BARB, has been grappling with for some time. It’s made worse by the fact that the existing peoplemeters, which recognise each TV channel by its frequency, won’t work with digital television because it pumps out several channels on a single frequency.

Eighteen months ago, Taylor Nelson revealed a new technique for keeping track of the hundreds of digital channels – PictureMatching – which uses computers to analyse the image on the screen in every home on the research panel.

But even when you can trace all the channels, the fundamental problem remains: how do you measure the number of people watching when the audience for most channels is so small?

“The core of the problem lies in sample sizes,” says Kirkham. “Thirty-nine of the 52 countries with peoplemeter systems have household panel sizes of less than 1,000 households. In Taiwan, the panel consists of just 600.

“These sample sizes may have been able to provide adequate measurement of two or three channels. But in the multichannel environment they cannot satisfy advertisers’ needs.”

This is because the smaller the rating, the larger the margin of error. The obvious answer is to increase the panel size and measure viewing in more homes, but with all sides of the industry trying to cut costs, there is a natural reluctance to do what is necessary. Taylor

Nelson believes it is possible to increase the BARB panel size – and thus the accuracy of the ratings – without increasing the number of peoplemeters.

“Our latest study shows that over 80 per cent of the variation in audiences is caused by the number of sets switched on,” says Kirkham. “The number of people watching each set accounts for very little variation. In addition, the ‘per-set’ viewing figures for major audience groups are highly predictable.”

He suggests it would be possible to reduce the number of homes monitored by people-meters and have a much larger panel monitored by set-meters, which measure what’s on the screen but not who is watching. Researchers would calculate the household ratings from the combined panels, and then apply a “per-set” factor derived just from the peoplemeter panel.

Taylor Nelson claims the accuracy of BARB could be improved by about 50 per cent – at no extra cost – by reducing the peoplemeter panel from 4,800 to 3,000 households, and adding a set-metered panel of 7,000.

One other benefit of the set-meter over the peoplemeter is it doesn’t require the household members to push any buttons, so they can be used for other research without risking “panel wear-out”.

This week, Taylor Nelson and TSMS will unveil the results of the first such research – a year-long study into advertising effectiveness, directly linking viewing data to consumer purchasing in the Meridian region.

For their TVSpan study, they put set-meters into 750 homes on Taylor Nelson’s consumer-purchase Superpanel, which monitors sales of over 60 of the most heavily advertised brands each week. The results are claimed to shed new light on the effectiveness of peaktime as compared with off-peak.

But valuable though such research may be to advertisers, the real potential of the Meridian panel could lie in the lessons it offers for the future of TV audience research. If the 750 households on the set-meter panel are combined with the 450 Meridian homes on the BARB peoplemeter panel, it produces a huge viewing panel – no fewer than 1,200 households.

This could be used by Taylor Nelson to test its theories about panel size and set-meters, and – more importantly – to convince the rest of the industry that they hold water. It is an expensive venture, and the company cites it as proof of its international ambitions, since – with multichannel TV expanding in most parts of the world – the problem of zero ratings is a global one.

Whether it can solve Unilever’s problem in Taiwan remains to be seen.