Module 0.3 introduced what these KPIs mean; here you compute them. The formulas are the program's own (Pre-Read slide 8), worked on one synthetic channel: spend 2.0M, support 400M impressions, incremental revenue 5.0M; the brand's total ad spend is 9M against 300M revenue.

Footnote on Effectiveness units: the Pre-Read's label says "per 10 Impressions" while its multiplier (×10) sits against a CPM defined per 1000 - a verify marker for this convention already lives in module 0.3, so treat Effectiveness here as a ranking metric (compare channels against each other and against CPM) rather than leaning on its absolute unit. The interpretation pair to internalise: CPM is what an impression costs, Effectiveness is what an impression earns - higher-CPM channels are more targeted and should rank higher on Effectiveness too (the coherence check from module 1.7).

Linear TV has no native impressions - it is bought and measured in GRPs (gross rating points), defined against a client-specified target audience. The program's conversion sheet gives the identities:

Worked: a campaign reaching 60% of the target audience 3 times each delivers 60 × 3 = 180 GRPs. Against a synthetic target audience of 25M adults, that is 180 × 25M / 100 = 45M impressions. GRPs are normalised to a 30-second ad length, and the same schedule produces different GRPs for different target demographies - the audience definition comes from the client, and the conversion methodology from the media agencies.

Why bother converting? Because a GRP-based effectiveness and an impression-based one differ by orders of magnitude - comparing them raw is a named pitfall. Standardising every channel to PMI (paid media impressions) is what makes CPM and cross-channel effectiveness comparisons legitimate.

These two views answer different client questions and are the most commonly confused pair in read-outs. Both tables below are fresh synthetic examples - build them yourself before reading the answers.

Contribution %: "what share of my volume does each driver produce?"

Due-to %: "what explains my growth vs last period?"

Each channel's due-to is its change in driven volume, divided by the prior period's total:

Read the two tables together and the teaching points fall out. The driver due-tos sum to the total growth (3.0 - 1.0 + 3.5 + 4.5 = 10.0%) - that additivity is the whole point of the view, and in practice base drivers and macro fill the same table. OOH is the smallest contributor (20%) yet the biggest growth driver (4.5 of the 10 points). Digital Video contributes a healthy 22.5% while dragging growth down (-1.0%) - a big contributor and a negative due-to at once, because it drove less than last period. When the client asks "what is driving my growth?", the answer is always the due-to table; when they ask "how dependent am I on TV?", it is the contribution table.

Where these numbers come from: the decomposition file

Both views are computed from the model's contribution decomposition export - one row per period, with columns Predicted, Actual, Residual, Intercept, and one contrib_<driver> column per variable (the same contrib_ names you met in the bounds template, module 1.6). Sum a contrib column across periods and divide by total predicted volume: Contribution %. Difference two periods' sums against the prior total: Due-to %. When channel results are broken down to platform and campaign levels, a weighted error-allocation step keeps children summing to their parent (Track 3.2 works the method).