Attribution is one of the most important topics in marketing. How does one give credit for a sale to the medium that caused it? Did the TV ad generate the sales, or the online advert, or even a sales clerk in a physical store? Last touch attribution — the final touchpoint gets full credit — is a common assumption but it is self evidently a bad one. Yet, I must confess I have a hard time coming up with much better answers to questions about attribution. I usually try and give a relatively gnomic answer and hope no one presses too much.
Part of the problem is that cause is a challenging concept. Often ‘all of the above’ is the correct answer but even then this doesn’t really provide a satisfactory answer. Are all equally responsible?
One challenge that Peter Danaher and Harald van Heerde tackle is what attribution modeling is useful for. Often marketers want to know what caused a sale in order to do more of it. This is a understandable intuition but runs into problems when attribution crudely drives budget setting. The authors therefore ask “Should attribution be used for allocation?” (Danaher and van Heerde, 2018, page 668).
To start they define what they mean by attribution. In essence attribution is the percentage of sales generated by a specific touchpoint. They show how to create a model of attribution that has appealing qualities, such as “..a more effective medium gets more credit” and “attribution increased in the number of exposures to a medium” (Danaher and van Heerde, 2018, page 672).
After showing their model they note that it isn’t useful for budget allocation. This is a valid but somewhat disappointing conclusion. Attribution describes what happened before but budgets should be set to maximize effectiveness going forward. This makes sense when you think about it. If only TV advertising were used in the past attribution would be 100% for TV advertising but that doesn’t mean that TV advertising alone is the best plan going forward.
They do give a practical approach, noting that nothing gets away from running tests to see which is the most effective medium.
Read: Peter J. Danaher and Harald J. van Heerde. “Delusion in attribution: Caveats in using attribution for multimedia budget allocation.” Journal of Marketing Research 55, no. 5 (2018): 667-685.