Marketing, Advertising & AdTech// topic
Attribution, Incrementality & Marketing Mix
In short
Deciding what actually caused the sale once the data is trustworthy: attribution models and their assumptions, geo holdouts, incrementality tests, and mix modelling at budgets too small for a vendor MMM.
4 pages
definitions
- Adstock: the assumption that last week's spend is still selling this weekAdstock carries a share of this week's media into next week's model input. The half-life is chosen, rarely identified from the data, and it moves the answer.definition7 min
- The lookback window: the assumption inside every conversion figureA lookback window has two dimensions set independently, so two platforms reporting different totals for the same sales are not disagreeing. They are counting different event classes.definition6 min
diagnostics
- The model hands a channel a negative coefficient nobody believesAn implausible coefficient is usually a statement about your spend history, not about the channel. Five causes, the checks that separate them, and where each branch ends.diagnostic9 min
- The test finished and the confidence interval still straddles zeroAn interval containing zero can mean the channel does nothing, or that the test was never able to see an effect this size. One calculation separates them before anyone moves budget.diagnostic9 min
Other topics in Marketing, Advertising & AdTech
See allWorking on something in this space?
Tell us where you are in a sentence or two. We'll tell you honestly whether we're the right team, and what a sensible first slice of the work looks like.
Start the conversation