Why it matters
Platform dashboards attribute conversions that might have happened organically. Before/after comparisons confound seasonality, budget shifts, and creative fatigue. A conversion lift study randomizes ad exposure (treatment sees ads vs control withheld) so marketing can answer a causal question: How many extra conversions did this spend cause?
Lift studies matter when finance asks for proof beyond platform ROAS, especially after turning on value-based bidding or user-level pLTV. Churney case studies (for example, Søstrene Grene on Meta) cite conversion lift readouts for incremental ROAS vs business as usual (BAU) automatic bidding.
Lift studies have limits: platform-defined conversion windows, eligible campaign types, minimum spend, and metrics that may not match your margin or cohort LTV definition. They measure platform-visible outcomes, not full finance-grade economics unless you align definitions upfront.
conversion-lift-study
Conversion lift studies fit the experiment layer of pLTV programs:
- Baseline: Document BAU conversion setup (event type, value field, bidding strategy).
- Treatment: Enable user-level pLTV from first-party data in your data warehouse; Churney sends values directly to ad networks (Meta CAPI, Google Ads Conversion API, app paths).
- Platform lift (optional): Where eligible, run a platform conversion lift study to measure incremental conversions from ad exposure vs a withheld control.
- Custom holdout (often needed): Withhold the pLTV value signal on a control slice when the platform lift design cannot isolate your signal change.
- Readout: Compare incrementality and incremental ROAS at agreed cohort maturity, not only platform-attributed ROAS.
Lift studies measure ad exposure incrementality. They do not replace calibration checks on predicted values or data warehouse-based cohort readout for returns and renewals.
Category variants
| Model | How conversion lift studies show up |
|---|---|
| Ecommerce / DTC | Platform conversion lift on prospecting or retargeting after enabling pLTV value events via CAPI; pair ad-exposure lift with custom signal holdouts when testing value changes. |
| Subscription app | Lift on trial-start campaigns; supplement with data warehouse readout on trial-to-paid and renewal because platform windows may be short. |
| SaaS / PLG | Lift on lead or signup campaigns; align on whether platform "conversion" matches qualified pipeline or paid conversion. |
Common mistakes
- Treating lift as full LTV proof. Lift studies use platform conversion definitions; they may miss refunds, churn, or offline revenue in your data warehouse.
- Underpowered studies. Insufficient spend or conversion volume yields inconclusive results and false negatives.
- Running lift without BAU documentation. No clear control definition makes results hard to reproduce or scale.
- Confusing lift with creative A/B tests. Creative tests swap assets; lift studies measure incremental impact of ad exposure, not bidding or value-signal changes alone.
- Stopping at platform learning phase. Ending before signal volume stabilizes or cohort maturity misstates long-term value impact.
- Skipping custom signal holdout. Platform lift may not isolate withholding pLTV events; pair with holdout test design when needed.
Advertiser lens
| Role | What they ask | What good looks like |
|---|---|---|
| Head of Performance / UA | Can Meta/Google run lift on our campaigns? | Eligibility check, minimum budget, treatment definition, and timeline agreed pre-launch. |
| VP Growth / CMO | What proves pLTV is worth scaling? | Incremental ROAS from lift plus data warehouse cohort quality at maturity. |
| Marketing Analytics / Data Science | Does lift match our incrementality definition? | Mapped conversion events, windows, and supplemental first-party readout plan. |
| Data Engineering | Do we need separate signal routing for holdouts? | Clear test vs control campaign map when platform lift cannot withhold value events. |
| Finance / Procurement | What triggers renewal or scale? | Pre-registered success metrics on incremental outcomes, not total attributed ROAS. |
FAQ
What is a conversion lift study?
A platform-run experiment that withholds ads from a random control group and compares conversions to an exposed group to estimate incremental conversions caused by ad delivery.
Who runs conversion lift studies?
Meta offers Conversion Lift (and related lift products). Google offers conversion lift and user-based conversion lift experiments for eligible Google Ads accounts and campaign types.
How is a conversion lift study different from a holdout test?
Platform lift studies randomize ad exposure. Custom holdout tests often withhold your pLTV value signal or bidding change on campaigns you control. Many pLTV programs use both.
Can a conversion lift study prove pLTV works?
Platform lift measures incremental conversions from ad exposure, not from a bidding-strategy or pLTV-signal change alone. Proving a value-signal change usually requires a custom signal holdout plus calibration, signal volume checks, and data warehouse cohort LTV at maturity.
What metrics does a lift study report?
Typically incremental conversions, incremental revenue (platform-defined), incremental ROAS, and confidence intervals. Exact fields vary by platform and study type.
How long should a conversion lift study run?
Follow platform minimums and your optimization window. For delayed value businesses, extend analysis with first-party cohort readout beyond the platform study window.
When is a lift study not enough?
When conversions omit returns, subscription renewals, or margin; when you need user-level signal withholding the platform cannot enforce; or when cross-channel effects matter. Add holdout tests and data warehouse readout.
Not the same as
| Term | Difference |
|---|---|
| Holdout test (custom) | You withhold pLTV or value events on control campaigns; not always platform-randomized. |
| Google Ads Experiments | Campaign split tests for bidding or creative; related but distinct tooling from Google conversion lift products. |
| A/B test (creative) | Tests assets or copy; lift studies measure incremental impact of ad exposure. |
| Geo experiment | Geography as the experimental unit; lift studies typically randomize users within platform audiences. |
| Platform ROAS | Attributed ratio, not incrementality; lift estimates causal lift vs control. |
| Media mix modeling (MMM) | Aggregate statistical model across channels; lift is a randomized platform experiment. |
