Attribution
How to Read an Attribution Report Without Lying to Yourself
Attribution reports are useful, dangerous, and almost always misread. Here is how to interpret them honestly.
Every attribution model lies. First-touch overweights awareness. Last-touch overweights conversion. Linear flattens reality. Time-decay favors recent. There is no honest single number. The skill is reading multiple models in parallel and forming judgment.
The honest read goes like this. Start with last-touch attribution to identify which channel converted. Then look at first-touch to identify which channel sourced the buyer. Then look at multi-touch (linear or time-decay) to see what was in the middle. The truth is the convergence of all three, weighted by your knowledge of the deal.
The trap most marketers fall into is using one model (whichever makes their channel look best) and reporting it as fact. Boards see through this within two quarters. The credibility cost of cherry-picked attribution is far higher than the budget cost of being honest about what you do not know.
The right framing for the board is: 'Here is what each attribution model tells us. Here is what we believe based on the convergence and our judgment about which channels are leading vs lagging. Here is where we are increasing investment based on this read.' That framing earns trust. Single-number attribution reports do not.