THOUGHT FOR THE DAY: WHERE IS THE SCIENCE?
Despite the huge rise in the number of data scientists (see the chart), the exponential growth of data, and the sterling efforts of Byron Sharp, Les Binet, Peter Field, and Mark Ritson, the advertising world is sadly lacking major new theories that explain how brands grow in the digital age.
Yes, brands grow by getting more buyers and, yes, the five Ps of marketing are as relevant as ever and, yes, behavioural economics is terrific. But, as programmatic advertising grows ever larger and machine learned algorithms increasingly dictate who is exposed to what machine learned ad — where is the new science that explains what works and what doesn’t. How do the algorithms know?
There is plenty of scope for the AI to get it wrong. Here is an example:
Let’s say that the goal for the programmatic advertising algorithm (and there must always be a human defined goal for machines to target) is to reduce the cost per sale. A machine learned algorithm, using clickstream data and not limited in any other way, will ‘learn’ that some people are more likely to buy than others so will increasingly target those people with the highest probability to buy. Increasingly exposing higher probability buyers to the ad will result in a higher sales conversion per person exposed and successfully deliver a lower cost per sale. Unfortunately, while the goal will be met, this will not be good for the sales of the brand because of two things:
- The people who have the highest probability of buying a brand are the heaviest existing buyers of that brand. They are more likely to buy the brand next time around, without being exposed to the ad, than someone who does not buy the brand. So targeting highest probability buyers is actually an unnecessary waste of money — they are existing customers with high loyalty.
- Failing to target lower probability buyers, people who do not buy the brand but who do buy other brands in the category, cuts the brand off from its acquisition target, reducing its reach and potential for penetration growth.
Programmatically targeting higher probability buyers will reduce the reach of a campaign, it will waste exposure on people who are most likely to buy the brand, it will (slowly) unsell the brand — but it will reduce the cost per sale. As sales go down!
So, where is debate and the new science that shows how to programmatically maximise sales? — predictably, over the long term. Where is the new theory that explains how to grow brands in the digital age?
Here’s my theory: the best way to grow a brand through programmatic advertising is to target category buyers who do not already buy the brand. What do you think? Worth testing?
— — — — — — — — — — — — — — — — — — — — — — —