Measurement Design
Part of Marketing mix modelling for media decisions
Distinguishing a marketing mix model from attribution reporting
See what MMM and attribution reporting each measure, why their channel figures can differ, and which result suits a media decision.
Attribution reporting allocates credit to eligible actions under a report's rules. A marketing mix model (MMM) estimates aggregate media effects under its data and causal assumptions. Use attribution to examine credited response within its coverage; use an MMM estimate when assessing a broader channel investment change.
Check what each result represents
An attribution report depends on the selected action, available interaction paths, lookback window and credit rule. Changing those settings can change the channel breakdown without changing the underlying business record. Credit does not show whether an action would have happened without the ad.
A platform report may also have missing or modelled data. Do not assume every credited action is a separately verified business order.
An MMM uses an outcome series, media activity and relevant other influences across time, often with geographic detail where suitable data exist. It can assess activity that cannot be traced through an individual customer path.
Its effect estimates still depend on variation, controls, model choices and the quality of the outcome data.
| Question | Attribution reporting | MMM |
|---|---|---|
| Main unit | Eligible actions and credited interactions | Aggregate outcome and activity by period, sometimes by area |
| Useful decision | Investigate response within the report's coverage | Assess channel effects and possible spend changes |
| Key limit | Path coverage and credit settings | Data variation and causal assumptions |
| Causal claim | Credit alone does not establish an additional outcome | An effect estimate needs defensible assumptions |
Investigate a disagreement
Suppose a hypothetical Australian retailer sees strong search credit in an analytics report while its MMM estimates a smaller search effect. The figures answer different questions.
Search may receive credit for appearing late in a recorded path; the model estimates how the outcome changes with search activity after accounting for measured influences. Either result could also be weakened by missing data or unsuitable assumptions.
First align the product, market, outcome and dates. Record the attribution settings.
For the MMM, inspect how search demand, promotions and other channels were handled, and whether search activity varied enough to estimate its effect. An experiment can provide another check when its tested change matches the decision, but its result applies to the conditions tested.
Keep the decision view clear
Show the agreed business outcome once. Place credited actions beside it as diagnostics, with their rules attached. Show MMM contribution and marginal-return estimates separately with uncertainty and assumptions.
Do not add supplier-credited actions as though they were unique additional customers. If the reports still disagree, decide whether to repair measurement, improve model inputs or test a bounded media change.
