It is a question digital and brand marketers have been trying to answer for a long time. What is the value of a display impression? Add to that the questions of does bigger format equal greater impact? Are ‘engagement metrics’ impacting brand recall? Is brand search worth the cost? And you have the attribution conundrum facing all digital advertisers.
Every digital advertiser would love to be able to answer this question. A lot of technology companies will tell you they have the answer. But is there even a definitive answer to find?
Of course, you could look at path to conversion analysis. Or maybe you could go one step further and look at incrementality. But is it really answering your question? What was the value of that particular impression, or that particular digital channel?
Everybody in the data and attribution industry seems to be searching for a silver bullet. And no doubt some advertisers would like one too. My belief when it comes to attribution is that there is no magic model. No single set up that will answer all of your measurement problems.
Some of the best data scientists I have worked with would happily sit in front of a client, after crunching data across hundreds of thousands of touch points and said, “I have no idea what caused that spike you see there.” And that scenario is not going away. Human behaviour can not be predicted down to such a granular level and online data does not give us the full view of what is influencing it. Yet some of us spend our days trying to understand every single peak and trough.
Not believing in a single model, doesn’t mean I don’t believe in seeking to attribute success accordingly. What stakeholders in a digital team need to do is come together and discuss the data, agreeing between them what it suggests as the optimal channel split and measurement approach for a campaign. They should then implement the agreed approach and assess at agreed intervals asking the question, “has the total generated from this new approach amounted to more than our old way of measuring?” If the answer is yes, persist with the new model and amend to improve accuracy. If the answer is no, address or go back to the old way.
Attribution is not a one size fits all system. And most of the time, the question of whether an individual impression had value is really not one which is worth pursuing. Accept that nobody has all of the answers, and your real goal is finding a model which works for you, and produces more pounds out, for every pound you put in.
This approach still requires the same AdTech. It still requires the right data solutions, modelling capability and smart thinkers driving it all. It just doesn’t rely on trying to find a single answer to a very complex question with lots of variables.