Why it matters
Attributed ROAS can rise while total revenue stays flat. Users who would have converted organically still get credit when they click an ad on the way to purchase. Creative refreshes, seasonality, and audience overlap all inflate platform metrics without proving the spend or signal change added net value.
That blind spot becomes expensive when teams roll out predicted lifetime value (pLTV) or shift to value-based bidding. Platform-reported value ROAS may improve because the algorithm reweights who it buys, but finance still needs to know whether incremental customers are higher quality or whether you paid twice for the same demand.
Incrementality is the discipline that separates correlation from causation. Performance marketing owns the test design; analytics owns the readout methodology; finance owns what counts as success. Without incrementality framing, signal pilots devolve into arguments about attribution settings instead of answers about lift.
Incrementality
pLTV changes what the platform optimizes on. Incrementality answers whether that change worked:
- First-party data in your data warehouse supplies baseline conversion and revenue patterns.
- Churney models and calibrates user-level pLTV, then sends values directly to ad networks via server-side paths.
- A structured pilot compares the new signal against business as usual (BAU) conversion or a holdout test that withholds the signal from a segment.
- After cohort maturity, teams measure incremental conversions, incremental revenue, and incremental ROAS, not platform totals alone.
- Causal modeling can extend readout when randomized holdouts are impractical, with clear assumptions documented upfront.
Churney does not replace your incrementality design. It supplies the signal under test; your experiment framework supplies the counterfactual.
Category variants
| Model | How incrementality shows up |
|---|---|
| Ecommerce / DTC | Holdout on pLTV value events vs purchase-only BAU; readout on net revenue after returns when maturity allows. |
| Subscription app | Test modeled subscription value vs trial-start proxy; watch trial-to-paid and early churn, not just install volume. |
| SaaS / PLG | Compare activation-value signals vs lead-gen BAU; longer sales cycles require extended maturity windows before judging lift. |
Common mistakes
- Equating platform ROAS with incremental lift. Attribution is not incrementality; always define a counterfactual (BAU, holdout, or geo design).
- Ending tests before cohort maturity. Early windows favor short-proxy signals; true incrementality on LTV may only appear after repeat or renewal patterns stabilize.
- Changing multiple variables at once. New creative, audience, bid strategy, and pLTV signal together make lift impossible to attribute.
- Holdouts too small to detect signal. Underpowered tests produce inconclusive readouts that teams interpret as failure or success arbitrarily.
- Ignoring organic baseline shifts. Seasonality and promos move BAU conversion at the same time as your pilot; pre-register comparison windows.
- No pre-registered success criteria. Teams move goalposts when platform metrics and finance metrics disagree.
Advertiser lens
| Role | What they ask | What good looks like |
|---|---|---|
| Head of Performance / UA | Did pLTV actually beat BAU? | Holdout or BAU design, stable campaigns during test, and agreed primary metric. |
| VP Growth / CMO | Are we scaling real lift or reallocating credit? | Executive summary tied to incremental ROAS and customer quality, not attribution alone. |
| Marketing Analytics / Data Science | Is the experiment valid? | Power analysis, leakage checks, maturity window, and documented causal assumptions. |
| Data Engineering | Can we implement holdout routing? | Clean event flags, no accidental signal bleed into control, and audit trail. |
| Finance / Procurement | What proof triggers continued spend? | Pre-signed incrementality thresholds and billing terms aligned to legal review. |
FAQ
What is incrementality in marketing?
Incrementality is the measure of additional outcomes caused by a marketing action compared with a counterfactual where that action did not occur. It focuses on causal lift, not attributed credit alone.
How is incrementality different from attribution?
Attribution assigns credit to touchpoints along a path. Incrementality asks whether the touchpoint or change caused extra conversions or revenue that would not have happened otherwise.
Why does incrementality matter for pLTV pilots?
pLTV changes the value signal sent to ad platforms. Without incrementality testing against BAU or a holdout, teams cannot tell whether higher platform ROAS reflects true lift or re-labeled existing demand.
What is the simplest incrementality test for value signals?
Withhold the new pLTV signal from a randomized segment (holdout) while keeping BAU conversion active for the control. Compare incremental conversions, revenue, and ROAS after an agreed maturity window.
Can geo experiments measure incrementality?
Yes. Geo holdouts or matched-market tests are common when user-level randomization is hard. Design and analysis require careful baseline matching and documented assumptions.
How long should an incrementality test run?
Long enough for cohort maturity on your primary economic metric. Separate the date the signal went live from the date you judge lift; short windows favor proxy metrics over true value.
Do ad platforms measure incrementality for you?
Some platforms offer conversion lift studies or experiment frameworks. They can help but may not cover your full value definition (returns, subscription LTV, margin). Complement with first-party readout where needed.
Not the same as
| Term | Difference |
|---|---|
| Platform ROAS | Ratio using platform-attributed value; does not prove causal lift vs organic demand. |
| Multi-touch attribution (MTA) | Path-based credit assignment; not a randomized or quasi-experimental counterfactual. |
| Blended ROAS | Cross-channel efficiency metric; still not incrementality without a control design. |
