Capturing attention in the attention deficit economy
This article is part of a series of articles on capturing attention in the attention deficit economy. Read more.
The study of the attention economy is graduating to the wide-scale adoption of consumer attention as the key measure of advertising success. Likewise, there is acceptance that attention should become a cross-platform currency unto itself. This is exciting for those of us who have been advocates of appropriately captured attention in advertising research. By appropriately captured attention, we mean measures of an unconscious behavioural reaction, not an attitudinal construct, such as recall or self-reporting.
Not all attention measurement is equal. In our own work, we have been able to demonstrate the vast difference between attention measured using a machine and that reported by a human. We have measured attention in two ways across 28 advertising effectiveness studies over the past couple of years. We have used machine learning to determine whether research respondents are really paying attention to advertising, and have asked the respondents for their own perceptions of the attention they have paid to viewed material. Comparing the two measurement methods with the same individual respondents has highlighted dramatic differences between man and machine. Respondents struggle to accurately describe the attention they pay to advertising: their reports are not even close to what the machine tells us. Across our studies, 50% of respondents over-reported their attention by an average of 42%, while the remainder under-reported by an average of 36%. In addition, self-reported attention is not even closely aligned with change in later buying behaviour, while attention captured using a machine is. Clearly, human self-reporting and machine measurement of eyes-on-screen attention is not the same. Measurement by machine is more granular (i.e. output is five times a second), enjoys improved objectivity, and benefits from reduced interference from a myriad of other variables that can affect a human survey response.