How to Measure Incremental App Installs From Influencer Campaigns

By INFLUENCE · Published November 2, 2026 · 7 min read

To measure incremental installs from creator campaigns, compare what happened where creators ran with what happened where they didn’t, over the same dates. The cleanest version is a geo holdout: run creators in some US regions, hold others back, and measure the difference in total installs, not just tracked ones. If you can’t split by region, a before-and-after comparison against a solid baseline is the next best thing.

Our view: if creators are a real line in your budget, you should know your incrementality multiplier, and almost nobody does. Tracked links only count people who tapped, so they can’t answer the question your CFO actually cares about: would these users have shown up anyway? INFLUENCE prices on delivered views and controls exactly when every creator posts, which makes it unusually easy to test. Here’s how to run the test. For basic install tracking, start with our AppsFlyer guide.

What does “incremental” mean for app installs?

Incremental installs are the ones that happened because of the campaign and wouldn’t have happened without it. Attributed installs are the ones a tracking tool credited to the campaign. They aren’t the same number, and they can miss in both directions.

Attribution undercounts when people see a creator, skip the link and search the App Store later.
Attribution overcounts when people who were going to install anyway happen to tap a creator link first.

Incrementality testing has gone mainstream. An EMARKETER and TransUnion survey from July 2025 found 52% of US brand and agency marketers use incrementality testing, and 36.2% plan to increase spending on it over the next year. (EMARKETER)

A clean test needs a precise record of where and when each video went live. INFLUENCE keeps that record for every creator, plus view counts, which is the raw material any incrementality test needs.

Three ways to measure incremental installs

Pick the strongest method you can run with your budget and data. Each step up costs more effort and gives a more trustworthy answer.

Method	How it works	Strength	Watch out for
Before-and-after baseline	Compare installs during the campaign with a matched period before it	Easy; no special setup	Seasonality, launches and press can fake a lift
Geo holdout	Run creators in test regions; hold out matched control regions	Strong causal read	Creators’ audiences don’t respect state lines
Time-based on/off tests	Alternate creator bursts with dark periods, several times	Works when geo splits are impossible	Needs several cycles; carryover effects
Marketing mix modeling	Statistical model across all channels over months	Sees the whole picture	Needs lots of history; slow to update

For most apps spending meaningfully on creators, a geo holdout is the sweet spot.

Whichever method you choose, it only works if the creator side is controlled. INFLUENCE agrees every post’s timing with the creator before the test starts, so a surprise upload doesn’t wreck a control week.

How to run a geo holdout with creators

Split US markets into test and control groups, run creators whose audiences concentrate in the test markets, and compare total installs. It’s less tidy than geo-targeted ads, but it works if you plan for the spillover.

Pick matched market groups. Use states or metro areas with similar historical install trends. Tools like Meta’s open-source GeoLift can help choose them.
Choose creators with regional audiences. Local creators, regional food or sports accounts, city-specific lifestyle creators. Check their audience geography before booking.
Set a pre-period. Record at least four weeks of installs in every market before the campaign.
Run the campaign for two to four weeks in test markets only.
Measure total installs (organic plus paid plus creator) in test vs control, compared with the pre-period.
Account for spillover. Some views will leak into control markets, which makes your result conservative, not inflated.

Finding creators whose audiences actually cluster in your test markets takes real audience data. INFLUENCE checks audience geography when building the creator mix, so your test markets get the bulk of the views.

Launch a campaign from $1 CPM

Talk to us

How to calculate incremental installs and cost

Subtract the expected installs (what the control markets or the pre-period predict) from the actual installs in test markets. Then divide spend by that difference.

The basic math:

Expected installs = test-market installs in the pre-period × (control-market installs during the test ÷ control-market installs in the pre-period)
Incremental installs = actual test-market installs − expected installs
Incremental CPI = creator spend ÷ incremental installs
Incrementality multiplier = incremental installs ÷ attributed installs

That last number is gold. If your multiplier is 2.5, every tracked install from creators represents about two and a half real ones. You can use it to adjust future reports until you re-test.

Once you know your multiplier, apply it to retained users too. Incremental installs that churn on day two aren’t worth much; see our post on measuring creators beyond CPI.

Clean math depends on a clean cost line. INFLUENCE charges only for delivered views and never past your cap, so the spend side of the equation is exact.

Common incrementality mistakes

Most bad incrementality results come from noise, not bad creators.

Testing during a big event. Holiday season, a feature on the App Store or a viral press hit will swamp the signal.
Too short a test. One week rarely separates signal from noise.
Mismatched markets. If test and control markets trended differently before the test, the result is unreliable.
Only counting tracked installs. The whole point is measuring total installs.
Changing other channels mid-test. Pausing paid social in one region during the test ruins the comparison.
Never re-testing. Your multiplier drifts as creators, formats and seasons change. Re-test a few times a year.

Keeping a test clean means no surprise posts and no shifted dates. INFLUENCE locks posting windows with every creator, so your test calendar holds.

How to run a geo holdout with INFLUENCE

Creator geo tests usually fail on logistics, not math. Here’s how we keep them clean.

Set geography to your test markets in the campaign form, next to the per-1,000-view ceiling and total spend for the test.
Approve a geo-weighted roster. INFLUENCE composes the creator mix around audience geography, favoring local and regional creators whose viewers sit in your test markets.
Lock posting windows. Your INFLUENCE AI agent agrees every post date in advance, and no creator posts during the pre-period or outside the test window.
Keep controls quiet. Nothing in the campaign targets control markets, so spillover is the only leak, and it makes your result conservative.
Use exact spend. Delivered-view billing means incremental CPI uses the real cost, never an estimate.
Read the creator record. Every post, publish date and delivered-view count sits in one dashboard, ready to line up against installs by market.

Relevant campaigns can include a view guarantee, which helps a test reach the volume it needs to read cleanly.

Attributed installs tell you what you can see. Incremental installs tell you what you actually caused. Run a geo holdout or a careful before-and-after at least once, find your multiplier and use it. You’ll likely find creators are doing more than your links say. INFLUENCE makes that test practical: geo-weighted creators, locked dates and exact spend.

Links show you who tapped. Incrementality shows you who came because of the campaign.

Set up the campaign. Approve the creators and content. INFLUENCE handles the rest.

Launch a campaign from $1 CPM

Talk to us

FAQs

What are incremental app installs?

They’re installs that happened because of a campaign and wouldn’t have happened otherwise. They differ from attributed installs, which are simply the installs a tracking tool credited to the campaign.

How do you measure the incrementality of influencer marketing?

Compare total installs where creators ran against matched markets or periods where they didn’t. Geo holdouts are the strongest practical method; before-and-after baselines work when regional splits aren’t possible. INFLUENCE can weight a campaign toward creators whose audiences sit in your test markets and lock every post date.

What is a geo holdout test?

It’s an experiment where a campaign runs in some regions and is held back in similar control regions. The difference in outcomes, compared with each region’s history, estimates the campaign’s true effect.

How many marketers use incrementality testing?

An EMARKETER and TransUnion survey from July 2025 found 52% of US brand and agency marketers use incrementality testing, and 36.2% plan to increase spending on it in the following year.

Do influencer campaigns drive installs that links don’t track?

Usually, yes. A large share of viewers look the app up in the store later rather than tapping through. Geo tests often show creator campaigns driving more total installs than attribution reports, which is why an incrementality multiplier is useful.

Related guides

Sources

EMARKETER, FAQ on incrementality: How to prove your ads actually work in 2026 (2026)