Attribution is correlation
Correlation is how strongly variables follow each other (e.g., sales of bags of ice & suntan lotion sales move in the same direction). With Facebook advertising, attribution is bound by click & view conversion windows, as well as a last-touch attribution model, and answers this digital advertising question:
Are clicks or views of a campaign’s ads associated with conversions?
Incrementality is causation
Causation is a change in one variable causing a change in another variable (e.g., the more you drive your car, the more gas you buy). With Facebook advertising, incrementality is independent of conversion windows & attribution models, and answers this digital advertising question:
Did clicks or views of a campaign’s ads cause conversions that would have not happened otherwise?
Which should I focus on?
Attribution is a means to an end, not an end unto itself. Ultimately we want to see digital advertising cause intended top & bottom line business impact. In a world of increasing digital tracking disintegration, measuring both attribution & incrementality is becoming more difficult. Facebook has a means of measuring incrementality for Facebook ads. But, it is not perfect, and it only considers Facebook ads. If that’s not your only advertising, you do not get a complete picture of incrementality.
Problems with incrementality / causation
- You can’t co-mingle in analysis the full impact of ads served from different ad platforms.
- Studies can be costly in terms of resources (they require work to set up) as well as time.
- It’s hard to draw conclusions over campaigns with different ad messaging and/or business & news cycle environments.
The bottom line
If you have zero attribution (correlation) you’ll have zero incrementality (causation) as well. Even where you only can see attribution, there may very well be underlying incrementality you just don’t see yet.
So do both. Attribution is easier, and should be your baseline performance evaluation tool. As resources allow, confirm attribution periodically with incrementality studies. Know such studies won’t be perfect, and they may only truly consider a single ad platform.
You do not need to be perfect. You just need to be better than your competition.
“It is better to be roughly right than precisely wrong.”
John Maynard Keynes