A New Approach To TV Audience Measurement

Abstract

This paper presents an innovative and potentially revolutionary alternative to the peoplemeter technology for measurement of television audiences.Peoplemeters, which were first introduced to commercial use in 1984, are now installed in more than 40 countries. Japan is alone among industrialised countries in continuing to rely on viewer diaries for television audience measurement.Peoplemeters are ciearly a more satisfactory approach than diaries; the data are available (through overnight telephone polling) much sooner, and there is some evidence that the data are more accurate than diaries (Friedman 1989; Rubens 1991). Certainly the method allows for analysis over periods much shorter than the quarter-hours of the diary method. There are however some problems with people meters. The first is that they can only be used with panels. The cost of recruiting and training households has to be amortized over as iong a period as possible, but his means that the problems of panel research - low final response rates, and possible unrepresentative members - camlot be avoided. The second is that the measurement depends on members of panel households accurately complying with the requirement that they log their presence by operating the buttons of the remote control device. Peoplemeter companies devote much effort and ingenuity to maximising compliance, and there is some research evidence that compliance is high. Unfortunately the evidence is not entirely convincing. A recent study in New Zealand (Danaher & Beed 1993) investigated compliance by telephoning panel members and comparing their self-reports of viewing with the behaviour recorded by the peoplemeter. In 92% of cases the panel member reported behaviour consistent with the peoplemeter record. Furthermore, the 8% non-compliers were made up of roughly equal numbers of over-reporting (peoplemeter showing viewing when panel member was not actually present) and underreporting. The audience measured by the peoplemeters was thus almost exactly the same size as the actual audience, as measured by coincidental self-report. From the point of view of advertisers however the interesting question