The 5 Steps to Running an Incrementality Test for Your Performance Program

Most scaling apps run Google App Campaigns, Meta, TikTok, and performance/affiliate networks at the same time. Each channel has its own attribution model, and each one has a reason to claim credit. The result is that most UA teams are paying for the same users multiple times, and budget decisions get made on numbers that don’t reflect reality. That isn’t a measurement success story. It’s an attribution problem dressed up as performance.

Every channel can show you installs. The question that matters is whether that channel drove installs that wouldn’t have happened anyway. That distinction is incrementality: the lift in conversions (installs, registrations, deposits, and so on) that would not have occurred without a specific channel or campaign. 

Most mobile marketers understand incrementality as a concept. The harder part is building a test that holds up internally, actually informs budget decisions, and doesn’t require turning off your existing channel mix. Here’s how the teams that are solving it structure their approach.

Step 1: Define Your Quality Signal (Not Just Your North Star, Your Proof Point)

Before you test anything, decide what a valuable user looks like. Your quality signal proves that the users you acquire are worth paying for. It won’t prove they’re incremental, because that’s what a holdout test measures. It does need to be specific enough to hold up as your proof point.

Strong quality signals vary by vertical:

• Finance: first deposit, first trade, linked bank account, first money movement
• Shopping: first purchase, cart completion with payment
• Entertainment: subscription start, first content play
• Gaming: reaching a defined engagement threshold, such as level completion
• Health & Wellness: subscription activation, first logged activity

What matters most is specificity. “First deposit” is a north star. “First deposit within 7 days from a net-new user not attributed to any other channel in the prior 30-day window” is an incrementality-ready quality signal. The first tells you what you want. The second tells you exactly what counts.

Push your team to get that specific. Every step that follows depends on it.

Step 2: Align Your Attribution Source of Truth Before You Test Anything

MMPs are the standard source of truth in user acquisition. Many brands also run internal attribution models, though, and the two frequently disagree. That raises two questions. Which one decides whether a channel is working? And where they diverge, who wins?

Most teams have never answered either question explicitly, which is why incrementality conversations so often stall before they start. Resolve this before you launch a test, because it’s where programs break down.

Use this alignment checklist before any pilot goes live:

• Confirm whether the MMP or the internal model is the final arbiter
• Document where and how often the two have historically diverged
• Agree on a lookback window for attribution
• Decide how view-through conversions will be handled
• Get sign-off from every stakeholder who will see the results, including growth, finance, and channel partners

Sign-off from channel partners matters for another reason, and most measurement guides skip it. Proving incrementality sometimes means proving that an existing channel isn’t delivering what it claims, even one with a relationship, a contract, and an internal champion behind it. That’s a politically loaded finding. When every stakeholder has already signed off on the source of truth, the conversation moves to reallocating budget on evidence rather than gut feel or internal politics. 

This step can’t be skipped. Suppose you run a pilot, the MMP says it’s incremental, and your internal model disagrees. You haven’t proven anything. You’ve just created a new argument. Align first.

Step 3: Structure Your Pilot to Prove Incrementality, Not Just Performance

Most channel pilots are built to answer “does this channel perform?” An incrementality-focused pilot answers a harder question: “does this channel drive conversions we would not have gotten otherwise?” Getting there takes five deliberate design choices:

• Set shared KPIs before launch. Define success explicitly. That means incrementality targets, not just CPA targets.
• Define a control. Know what your baseline looks like without this channel active.
• Contain the test. Use a controlled budget, a defined time window, and a specific geo or audience segment.
• Plan for event lag. In finance, for example, down-funnel events can take 3–7+ days to materialize. Size your test window for how long your events actually take, not how long you’d like them to take.
• Compare against your existing channel mix. Evaluate the new channel’s contribution relative to what you’re already running, not in isolation.

A pilot structured this way produces evidence you can defend when budget is on the line.

Step 4: Benchmark Against Cross-Program Patterns

Single-program data has a built-in limitation. It tells you what happened in your program. It can’t tell you whether that result is good, normal, or a red flag compared to similar apps in your vertical.

Cross-program benchmarking fills that gap by answering the questions your own data can’t:

• Are your down-funnel conversion rates in line with similar apps, or are you an outlier, good or bad?
• Are the signals from a new channel consistent with what that channel typically delivers in your vertical?
• Is the lift you’re measuring real, or is it within the normal variance you’d expect from any new channel in its first 30-60 days?

Benchmarks also set a realistic ceiling. 100% incrementality isn’t a real target, because some conversions will always come from brand loyalty or organic behavior. In digital advertising, anything above 70% is considered exceptional.

Single-program data tells you what happened. Cross-program data tells you what’s likely to happen and whether your results are meaningful. Most teams can’t access that insight on their own. It requires either operating at significant scale across dozens of similar programs or working with a partner who has that cross-vertical visibility.

Step 5: Build a Feedback Loop That Compounds

Treat incrementality as a continuous measurement discipline. It gets more valuable the longer you run it. In practice, the feedback loop builds on itself in four ways:

• Ingest more down-funnel signal over time, and optimization sharpens.
• Expand from a single program-level grade to lift broken out by channel and source.
• Shift the question from “is this channel incremental?” to “what’s the optimal budget allocation across all channels based on proven incrementality?”
• Build internal reporting that shows incrementality trends over time, not just point-in-time snapshots, which feeds the next round of signal.

The teams making the most confident budget decisions build incrementality into their ongoing operating cadence.

How Perform^[cb] Proves Lift

Perform^[cb] is built around these principles, with measurement designed in from day one rather than added later.

For every campaign, Perform^[cb] builds a matched holdout: real users from the same vertical and partner mix, with zero ad exposure. That holdout is tagged as its own source in your MMP, so you can verify it at any time. Lift is calculated as (Campaign CVR − Holdout CVR) ÷ Campaign CVR, on a rotating, always-on cadence. Across verticals and channels, Perform^[cb] programs consistently measure 75-85% incrementality, in an industry where many channels struggle to clear 50%. In one recent holdout test, a leading fintech app measured 89% incrementality on iOS and 97% on Android.

The benchmarks behind that number come from scale. Perform^[cb] manages performance media for more than 35 of the top 100 finance apps in the U.S., across hundreds of active programs in multiple verticals. When we talk about what’s working across programs, we’re drawing on live signal from a significant share of the competitive set you’re actually benchmarking against.

That’s what Verified Lift means in practice. Real lift isn’t a number we report to you – it’s something you can open and verify yourself. 

Stop Reporting. Start Proving.

You don’t need another UA dashboard. You need clarity on what’s actually driving growth, and that clarity comes from treating incrementality as something to prove rather than assume.

Want to see where your program stands today? Reach out to our team of experts to see how Perform^[cb] proves incrementality for programs like yours.