App Retargeting · Facebook & Google · Re-engagement
Retargeting is one of the most misused budget lines in app marketing. The measurement is hard, the attribution is confusing, and the ROI is almost impossible to defend to a sceptical stakeholder. This is what we've learned from running retargeting across mobile games, consumer apps, and subscription products — what works, what doesn't, and how to allocate budget when you can't prove it conclusively.
Retargeting budget as share of total — conservative starting point for most products
Budget share where retargeting can stabilise revenue without a conclusive ROI proof
Minimum active-user exclusion window — below this, your "retargeting revenue" is just existing user activity
One-time lifecycle events (registration, first purchase) are often the highest-ROI retargeting targets
Retargeting fit by product type — indicative ROI potential
What retargeting actually is — and isn't
This is fundamentally different from acquisition campaigns. The audience is already installed. The challenge is not getting a download — it's getting a specific in-app action from someone who stopped doing it. Without deep linking (routing the user directly into a specific screen on app open), retargeting campaigns deliver users to a home screen and lose most of the conversion potential immediately.
Attribution. The core question — did this user return because of the ad, or were they going to return anyway? — is genuinely difficult to answer. Standard attribution windows overcount retargeting effectiveness because they include users who would have re-engaged without any ad.
This is why retargeting budgets are hard to defend. The data shows conversions. The question is how many of those conversions were caused by the ad versus correlated with it. Without a holdout test or carefully tuned exclusion windows, the number is unreliable.
Running retargeting on Facebook and Google
FACEBOOK — CONVERSION CAMPAIGNS
Use Conversion campaign type with a specific in-app event as the optimisation target. Create custom audiences from your installed user base using in-app events passed back via the SDK or Conversion API. Layer in exclusions — for example, exclude users who completed the target event in the last 30 days, so you're only reaching users who haven't yet taken that action.
Audience combinations: users who installed but never registered, users who registered but never made a first purchase, users who made one purchase 60+ days ago but not since. Each segment gets a different creative and a different optimisation event.
GOOGLE — APP CAMPAIGNS FOR ENGAGEMENT
Google's retargeting product for apps is App Campaigns for Engagement (ACE). Requires Firebase integration — the app must pass events through Firebase for audience segmentation to work. Targets existing installed users with re-engagement goals.
The setup mirrors Facebook: define the audience segment (installed, not active in 30 days), define the target event (return to app and complete a purchase), set the creative. Deep linking is essential on both platforms — without it, the ad opens the app to the home screen and the conversion chain breaks.
Deep linking is not optional. Retargeting an lapsed user to a specific promotion — a limited bonus, a level unlock, a discounted pack — and then routing them to a home screen wastes most of the ad's work. Deep linking routes the user directly into the specific screen that matches the ad. Without it, drop-off between ad click and target action is very high.
Attribution setup — the part most teams get wrong
Setting these correctly — especially the active user exclusion window — is the difference between data that looks impressive and data that actually reflects retargeting's contribution. The most common mistake: not excluding active users, which causes the dashboard to attribute organic re-engagement activity to the retargeting campaign.
Even with correct settings, the data is still ambiguous. AppsFlyer shows you attributed conversions — but it cannot tell you whether those conversions would have happened without the ad. The only reliable way to answer that question is a holdout test: run retargeting on half the eligible audience and compare conversion rates. Without a holdout, the retargeting ROI number is an estimate at best.
Which products should run retargeting
GOOD FIT — RETARGETING WORKS
Core users who pay repeatedly. Retarget with limited-time bonus packs, event promotions, or seasonal offers. Each return visit has monetisation potential. High-value user segments make the economics work even at relatively high retargeting CPAs.
Users who got stuck on a level and churned. Retarget with a relevant offer tied to that level — a hint pack, a power-up, a discounted continue. The user has a specific unresolved problem; the ad addresses it directly. Conversion rates are meaningfully higher than generic re-engagement.
Retarget lapsed users with social triggers — "someone sent you a message", "your friend posted". The incentive is personal and specific. Even if the retargeting attribution is imprecise, DAU stabilisation is observable and defensible as a business outcome.
Users who installed but never completed registration, or registered but never made a first purchase. These events happen once per user lifetime. Retargeting users who are inactive for 30+ days but haven't completed the event produces retargeting CPAs well below normal acquisition cost for the same event.
BAD FIT — DON'T BOTHER
Products where the value proposition is a single transaction — a scam, a predatory loan, anything where the user's experience of the first interaction was negative. These users left with a reason. No creative recovers from a fundamentally bad product experience.
Users who were charged high fees or had a bad borrowing experience are not coming back for more. Retargeting spend in this category is visible in the comment sections — the return on negative sentiment far outweighs any re-engagement revenue.
When the core product relies on deception (fake engagement, scripted interactions), users who experienced it once have no reason to return. The product itself is the problem — retargeting cannot overcome that.
Products where a user's need was satisfied by the first purchase and is unlikely to recur in a meaningful timeframe. The retargeting audience exists, but the conversion opportunity doesn't.
The test: Does a returning user generate real value each time they return? If yes — retargeting has a logical economic case. If the product's revenue model depends on a user not understanding the full cost until it's too late, retargeting just brings more people to that realisation faster.
Audience segmentation — the right people, the right offer
Exclude users active in the last 30 days. Target: complete registration. Creative: what they're missing by not having an account. This cohort has the highest gap between install cost (already paid) and conversion value (not yet captured). Retargeting CPA for registration is often 40–60% below acquisition CPA for the same event.
Exclude users active in the last 15 days. Target: first purchase with a limited offer. Creative: first-time buyer incentive tied to a specific product or pack. Users who registered showed intent — the gap between registration and first purchase is often a price or value perception barrier, not a disinterest signal.
These users converted once and stopped. Target: a time-limited re-engagement offer relevant to what they bought originally. For mobile gaming: an in-game reward tied to previous progression. For casual games: a pack for the level they were on. Personalisation at this segment level produces significantly better conversion rates than generic win-back creatives.
Your most valuable users who stopped engaging. Highest retargeting budget priority. Custom creative that reflects their user level or history where possible. Offer should match their previous spending tier — offering a $5 pack to someone who historically bought $100 packs is a signal you don't know who they are.
Users who stopped at a specific level or point in a progression system. Target: a relevant offer that addresses the specific barrier. Completed levels 1–5, dropped at 6 — target with a level 6 progression pack. This requires event data granular enough to segment by progress state, but produces retargeting ROI that's defensible even without a clean holdout test.
Budget allocation — the honest answer
The uncomfortable truth: retargeting budget is often set by feel, calibrated by observing revenue stability rather than proven ROI. This is not a failure of analysis — it's the reality of a measurement problem that hasn't been fully solved. Here's how to approach it honestly.
STARTING POINT
For most products: start at 5% of total budget or below. Run for 30 days. Observe revenue stability, DAU impact, and retargeting-attributed CPA. Adjust from there based on what you see.
WHEN IT'S WORKING
When retargeting demonstrably stabilises revenue — reducing revenue drop when acquisition slows, or maintaining DAU during periods of lower new user volume — 20–30% can be justified. The metric here is revenue stability, not attributed ROAS.
When the CPA makes the case for you
The clearest case for retargeting budget increase: a one-time lifecycle event (registration, first purchase) that costs $X to acquire from new users — and costs $X/2 to recover from lapsed users who installed but never completed it. When the retargeting CPA is materially below acquisition CPA for the same event, scale that retargeting line until the CPA converges or the addressable audience is exhausted.
ACQUISITION CPA — REGISTRATION
$4.20
Cost to acquire a new registered user from acquisition campaign.
RETARGETING CPA — REGISTRATION
$1.90
Cost to recover a registration from a user who installed but never registered. Same event — half the cost. Scale this line.
The honest summary
01
Attribution will always be ambiguous. No platform-side measurement fully answers "would they have returned without the ad?" Accept this, and measure retargeting by revenue stability rather than attributed ROAS.
02
The active-user exclusion window is the most important parameter. Set it wrong and you're measuring organic re-engagement and calling it retargeting performance. Minimum 15 days, ideally longer for low-frequency products.
03
One-time lifecycle events are the clearest retargeting use case. Registration, first purchase — events a user can only ever complete once. Users who installed but never converted are an addressable audience with a known gap. Retargeting CPA for these events is almost always below acquisition CPA.
04
Deep linking is non-negotiable. An ad that promises a specific offer and delivers users to a home screen loses the majority of conversions at the moment of arrival. Deep link directly to the relevant screen — the offer, the level, the purchase flow.
05
Segment the audience. Match the offer to the user state. Generic lapsed-user campaigns with generic creatives produce weak results. Users who got stuck on level 6 need a level 6 offer. Users who registered but never paid need a first-purchase incentive. The more specific the match, the higher the conversion rate.
06
Some products should not run retargeting at all. If users left because the product experience was bad, retargeting brings them back to a bad experience. Fix the product first. And if the product's revenue model depends on users not fully understanding what they signed up for — retargeting that audience is not a sustainable growth strategy.
We audit attribution setup, audience segmentation, and budget allocation — and tell you honestly whether retargeting makes sense for your product at the current stage. Sometimes the answer is no, and that's worth knowing before spending the budget.
Talk to us on WhatsApp →