More on Attribution

Few days ago, I had a conversation with an acquaintance of mine about marketing attribution. He kept quizzing me about my stance that attribution is a useless and unnecessary excercise. Of course, in the current marketing analytics climate, this is blasphemy and, coming from one of the attribution meccas -- Digitas, I should have known that I would get significant push-back.

The crux of the argument for attribution was, that it is the only way to measure how much an individual (digital) channel drives conversions and to optimize it. My argument was that,
1. You can't measure it
2. You don't need to optimize to it.

Of course, as usual, I did not make my argument as eloquent as I could or should have, and spent the next few days going over it in my head. Here's what I think.

I still believe there is no right way to measure attribution, especially in digital space. Most common approaches, first click, last click, even factorial attribution will miss the true attribution. We cannot, and should not look at attribution on macro-levels, attribution is purely individual, and we cannot use averages to simplify it. As Jim Lenskold rightly spoke in his book "Marketing ROI", customer value is not a constant, we should not be using a constant, or average customer value in calculating ROI. Similarly, why are we falling into the same pit by calculating average/constant attribution factor for a given digital channel? By assuming that attribution is constant for a given combination of a digital channel and client, we are saying that a given channel cannot drive more conversions.

As I've argued before, one measure cannot be both performance and optimization measure. For instance, if you are trying to optimize 'quality' visits to your website, 'quality visits' is not a performance measure, it is a mean to an end. Similarly with attribution, if you want to optimize how effective is each channel in contributing to the conversion, you should not be using the same measure to talk about the performance of the channel, and, more importantly, assign budgetary values to each channel based on that performance.

One may ask, and rightfully so, 'if you are not using attribution models to find out how well each channel is doing, how do you know it's doing it's job?'

My answer is -- determine what job that channel has and manage accordingly. It is naive to think that the job of each channel is to drive conversions. At the same time, however, it is naive to think that channels have a job of 'awareness' or consideration. Here's a simple objective for an online media channel "Drive 1,000,000 qualified visitors to xyz.com within 3 months using $200k budget." Determine what 'qualified' visitor is. Same with paid search and social. Same with website. Only website now, has a bigger objective, in taking these visitors and converting them into paid customers. Yes, we are building a story, but each chapter has it's own job to do.

We should stop thinking how much each channel is contributing to conversion, and start thinking and concentrating on making each channel as effective as possible in what it is designed to do. Engineers and assembly workers designing and building doors for the car are not thinking on how well the engine is going to be running, they are concentrating on designing and building the best door possible that serves it's function well. So should we.