For a brief time in 2003 I ran away and joined a media circus. We had finally published the first results from a London Business School study on advertising viewing.
The study was based on eight different households that were videotaped for two weeks. We studied TV viewers' behaviour during every commercial break that occurred in the second week of the study in an attempt to understand what happens when programmes end and advertising begins.
With the analysis complete I wrote an article for the Financial Times. I explained that while many households did watch the ads, other activities such as social interaction, reading, flicking and advertising interaction often took precedence. I also described how the peoplemeter system was flawed because measuring the number of people in the room had a variable correlation with the number watching the ads. Indeed, in many instances, increasing numbers of people in the room led to significantly less people watching the ads.
I thought that this was the end of the story. But then all hell broke lose. After the space shuttle and Saddam it became the week's most downloaded article on the FT's web page. The story was then picked up by The Guardian, The Independent, and The Daily Mail. It was featured on Five News and, most importantly, on Richard and Judy. As the story went further afield, I even found myself in the rather bizarre position of doing a major Australian breakfast show live from my bed, drinking Horlicks.
This kind of media exposure created quite a response from the ad industry.
It seemed to revolve around two key methodological criticisms. The first was based on the sample size. Clearly eight households was inferior to BARB's 5000 or more. Well this depends. Eight households can never be representative, but it is better to have eight measurements of the phenomenon you are interested in (advertising viewing) than 5000 measurements of something you are not (people in a room when the ads are on). It seems that much of the ad industry has become obsessed with quantities rather than the qualitative nature of what is being measured. Good research always combines the qualitative 'what' with the quantitative 'how much'.







