What do you believe?
Analytics and attribution in advertising are similar to religion. You have to believe in one approach based on the result of how it makes you feel, and you
stick with it for a long time. Maybe somewhere along the way you make a switch, but not without much consternation and debate. For an attribution model to work, you have to have faith.
Last week I moderated a panel on this topic where this analogy resonated with me. Many advertisers use many different models for attribution, from media-mix models to last click, from
weighted averages based on previous experience to massively complex algorithms.
The only thing consistent for these models is, they all have flaws because they are trying to measure
human behavior, which is ultimately unpredictable.
To that end, you have to pick an approach and stick with it because what you want is to measure trends over a period of time. If you
can count on the consistent inconsistency in your model, then you can be effective in your analytics!
The goal is to understand the model’s flaws, account for them, and ensure they are
consistently wrong in the same place. If you do this, you can look at the trend over time.
What doesn’t make any sense is testing out new models of attribution on a regular
basis.
The dirty little secret is that if you work the numbers enough, you can get them to say whatever you want them to say. Numbers are malleable in one-off instances.
Trends are not malleable.
The trend of improvement over time and a trend that correlates to increased sales results in an effective campaign. You can accept fluidity as long as it
is over a long period of time.
My advice is to identify a model and stick with it for at least a year, and preferably for two to three. At one year, you might start to see some trends
and can determine if those trends are in line with your sales numbers. Over a two-to-three-year period, you can absolutely identify connections.
During that panel last week,
one of the panelists made a great comment: Being data-driven is not the same as being data-informed. Being data-driven means you rely very heavily on the numbers to make your decisions.
Being data-informed means you read the numbers, but you still have the choice to ignore them and go with your gut.
If you have the right analytics model in place, you are very likely to
know that the numbers are what they tell you, but you will never be afraid to make the final decision based on what your instincts are telling you.
If the model is accurate (enough), then your
decision will either pay out or prove wrong quickly, and the trend line of your results will tell you this. It’s about belief. Do you believe in your instincts — and are there
data points to support your ideas?