Increasingly, in meta analyses conducted by Kantar and other firms we are beginning to see that digital is just another media channel; one that has its own unique properties and challenges. In our CrossMedia studies we find that once share of spend is equivalized, digital performs very similarly to other channels in terms of brand building. But given the much-touted advantages of digital – better targeted, more timely and adaptable – how could that be?
Listening to the Ted Radio Hour recently, I was struck by a statement made by Brian Little, a research professor in psychology at the University of Cambridge; he was reflecting on the fact that while personality traits are in part inherited it is possible for people to go against those traits and that they change over time, and said,
“We’re wonderfully complex creatures. And I think that part of the delight of our complexity is that we’re not as predictable as we might be.”
In many ways my own experience suggests that this is true. I have reflected elsewhere on the fact even though I may set out to buy a specific brand, writing it down on a shopping list, if that brand is not available I will end up buying another one. Kyle Findlay recently sent me a paper by Constantin Michael, Jan Hofmeyr and himself that compares stated purchase intention, e.g. buy most often, against actual purchasing behavior from panel and loyalty records finds little relationship over three months, but a far stronger relationship over 12 months. Given the chaotic influences on any individual packaged goods purchase this is hardly surprising.
But what if we look at it from the other side, using only behavioral data, could we accurately infer what a person’s real brand preference was? Yes, we probably could, given enough data and time. If you have all my purchasing data, across different stores, you could probably figure out that I prefer Pete’s Coffee, ground French Roast, even though I will buy other brands as the need arises. That might work for packaged goods which are typically instinctive and habitual purchases, but what about cars, financial services or airlines? Now things get a lot more complex and other factors start to influence choice, working against my attitudinal preferences.
Based on the work we at Kantar have done over the years I know that we can predict behavior fairly accurately based on an individual’s stated intentions, but the outcome is very dependent on circumstance. Across hundreds of people, for instance when predicting market share, the predictions turn out to be very accurate and stable, particularly when you take market size factors like brand size into effect, but at the respondent level not so much (something confirmed by the paper referenced above).
Now apply that learning to digital advertising. Do you really think that ad targeting is going to be accurate any time soon given the quality of data used?