Incrementality Mix Modeling

The cost of not measuring right

Every media plan holds two numbers for each channel: the lift you assume it generates, and the lift a controlled experiment would actually measure. This tool shows what's hiding in the gap — and what reallocating on the truth adds to your top line.

Budget A = what you spend on ads today. Budget B = what you'd spend after learning each channel's true lift (usually larger: more budget flows to the channels that actually work, and those channels can absorb it profitably up to their reach ceiling). Blue cells are editable — change any value and the tool recalculates live. Hover the ? next to any row label for a quick definition.

Total ad budget (A)
$400M
Budget as % of revenue
10.0%
1
What you think your ads drive (Budget A, Assumed Lift)
2
What they actually drive (Budget A, True Lift)
3
What they could drive (Budget B, True Lift)
Reallocation upside (Budget B Actual − Budget A Actual)
Budget:
Total company revenue
Today Ads share of rev: assumed · actual
After reallocation Ads share of new rev:
top-line growth

Incremental revenue by channel

Budget A Assumed vs. Budget A Actual vs. Budget B Actual

Budget shift

Budget A vs. Budget B allocation

How to read this

The formula.

  • Reach = MIN(Budget × Reach Efficiency, Max Reach). Linear up to a cap.
  • Channel Incremental Revenue = HH TAM × Reach × Avg Spend per HH × Lift.
  • Every term has a physical meaning. No multipliers, no fudge factors.

The three-act story.

  • Act 1 — What you THINK your ads drive. Uses Assumed Lift. This is what attribution or MMM tells you today.
  • Act 2 — What they ACTUALLY drive (at the same budget). Uses True Lift. This is what controlled experiments would reveal.
  • Act 3 — What they COULD drive after reallocating. Uses True Lift at Budget B. Usually: total budget grows, total revenue grows more.

Napkin Math assumes linear reach (no diminishing returns) and independent channels (no audience overlap). For a more realistic view, switch to MMM-Lite.

Sum of channel reach (A)
Deduped combined reach (A)
Sum of channel reach (B)
Deduped combined reach (B)
1
What you think your ads drive (Budget A, Assumed Lift)
2
What they actually drive (Budget A, True Lift)
3
What they could drive (Budget B, True Lift)
Reallocation upside (Budget B Actual − Budget A Actual)
Budget:
Total company revenue
Today Ads share of rev: assumed · actual
After reallocation Ads share of new rev:
top-line growth

Incremental revenue by channel

Budget A Assumed vs. Budget A Actual vs. Budget B Actual

Reach saturation curves

S-curve: how reach responds to incremental spend

How this model works

Differences from Napkin Math.

  • S-curve reach. Each incremental dollar buys less reach as the channel approaches its Max Reach — how real media actually behaves.
  • Optional reach deduplication. When on, overlapping audiences across channels are netted out so a household reached by both TV and CTV isn't counted twice. Watch what happens to total revenue when you flip the toggle — that's the reach inflation hiding in most attribution models.
  • Equalized-marginal-ROAS optimizer. Adds dollars to whichever channel currently has the highest marginal return per $; stops when the best remaining channel returns less than $1 per $1.

The core point stands: the cost of not testing is almost certainly larger than the cost of testing. You can cap the cost of an experiment. You can't cap the cost of a bad allocation.