Identify the question you need to answer
Attribution can help understand recorded paths and support operational reporting. Incrementality becomes important when deciding whether spending creates enough additional value, especially where customers already have strong intent or several channels overlap.
Neither approach is free of limitations. Attribution depends on observable events and rules; causal comparisons depend on design, execution and the relevance of the comparison.
Compare the evidence
| Evidence | What it can support |
|---|---|
| Attributed conversion report | Outcomes credited under the selected rule. |
| Customer journey records | Observable interactions and progression. |
| Randomized holdout where suitable | Added impact under a valid experimental design. |
| Before-and-after observation | A descriptive change with possible confounding factors. |
Design a better decision
- State whether the question is about credit, journey diagnosis or added impact.
- Identify the outcome, population and observation period.
- Check whether a suitable comparison or holdout is feasible.
- Account for overlap, spillover and conversion lag.
- Report the conclusion within the design's actual limits.
A causal test needs more than a dashboard labeled “incremental.” Review how groups were formed and whether other conditions changed differently between them.
Worked example: retargeting credit
Illustrative example: a person visits after a recommendation, later clicks a retargeting ad and purchases. Attribution may credit the ad according to the platform's rules. To estimate added impact, the business needs evidence about comparable people who did not receive the intervention under an appropriate design.
The ad may still help. The distinction prevents reporting an observed path as proof of causation.
Avoid false precision
Different systems may credit the same purchase to different channels. Adding their attributed revenue can exceed actual revenue. Use operational totals as the business record and explain why platform reports are not independent pieces of one additive total.
Is last-click reporting useless?
No. It can answer a limited operational question when its definition is clear. It should not be presented as complete causal truth.
Does incrementality require a large company?
Some useful controlled comparisons are possible at smaller scale, but feasibility depends on volume, timing and design. When evidence is limited, state the uncertainty.
Put this into practice
Take one channel report and rewrite its claim from “caused” to “was credited with” where appropriate. Identify which spending decision needs stronger added-impact evidence.
Primary-source reading for platform details: Google Analytics: attribution.
Related foundation: CAC vs ROAS: which metric should guide your marketing budget?. How these guides are prepared.
Ayoub Mouhachtt
Growth & performance marketing. Explore the portfolio and working background.
Related portfolio work: Etsy & Shopify. The worked examples in this guide are illustrative and are separate from the portfolio’s project evidence.