← Every exhibit
The calibration log
A model becomes more credible when its assumptions and outside anchors are visible in the same row.
| Model element | Assumption | External anchor | If the anchor moves |
|---|---|---|---|
| Baseline demand | The non-media demand path has this shape | Control series, category data or known interruption | Re-estimate the base before reallocating |
| Carryover | Exposure persists for this long | Delayed response in an experiment or prior with a stated basis | Widen the scenario range |
| Saturation | Additional spend produces less response after this point | Spend variation with an independent shock | Do not use the point estimate as a ceiling |
| Channel interaction | Two channels reinforce or substitute | Designed test or a documented mechanism | Keep the interaction as a scenario, not a fact |
| Incremental lift | The allocation reflects causal contribution | Holdout, ghost-ad or lift result | Mark the model coefficient as uncalibrated |
Swipe or scroll horizontally if the table is wider than your screen.
Reference & Evidence
Source: Author's worksheet for documenting what enters a marketing mix model and what anchors it outside the model.
Related exhibits
-
What each measurement instrument can carry
From the essay Marketing mix modeling is a calibration problem
-
Accumulation versus activity, indexed over twelve quarters
From the essay Why growth compounds, and activity doesn't.
-
The same work, in two forms
From the essay Why growth compounds, and activity doesn't.