Table of Contents
Retention-first user acquisition is a paid media strategy that judges every campaign, channel, audience, and creative by the share of installed users still active and paying at Day 7, Day 30, and beyond, rather than by install volume or first-day cost. In Admiral Media’s work across subscription, dating, fitness, and edtech apps, retention-first UA consistently exposes campaigns that look efficient on a cost-per-install basis but quietly destroy cohort economics, and reroutes spend toward audiences that compound revenue over time.
Most mobile apps acquire users who churn before the app earns back its acquisition cost. That is the core problem retention-first UA solves. Admiral Media uses Day 7 and Day 30 retention, paired with cohort revenue, as the primary spend signals because they correlate with lifetime value far more reliably than installs, signups, or first-session events. This article explains how the Admiral Media team operationalizes that approach, the framework used to wire retention back into bidding, and what the math has looked like across real Admiral Media accounts.
Why install-led UA breaks subscription and consumer apps
Install-led UA optimizes for the cheapest user, not the most valuable one. The result is a portfolio of campaigns that hit short-term CPI targets while bleeding cohort revenue. The Admiral Media team sees this pattern across categories: campaigns that win the install auction often lose the LTV auction.
The mechanism is straightforward. Bidding algorithms find users for whom the optimization event is cheapest, not users for whom downstream monetization is highest. When the optimization event is an install or a first-session trigger, the algorithm pools demand toward low-intent users who happen to be cheap to acquire. These users open the app once, fail to retain, and disappear. The cohort looks fine on Day 0 and collapses by Day 7. By Day 30, the campaign has paid for users who generate almost no revenue.
The Admiral Media team has measured this drag repeatedly: a campaign with a 30 percent lower CPI than its peers can produce a 50 to 70 percent lower D30 retention rate, and the cohort revenue gap is wider still because non-retained users almost never subscribe. The cheap install is, in most accounts the Admiral Media team manages, the expensive user.
What retention-first UA actually optimizes for
Retention-first UA optimizes the campaign against retained, monetizing users, not against the install. In practice, this means three operational shifts. First, the optimization event moves downstream from install to a retention or monetization proxy such as D3 retained session, trial start, or first subscription. Second, the bidding strategy moves from cost-cap or target CPI to target ROAS or value-based bidding wherever possible. Third, the reporting cadence shifts from daily CPI to cohort-level D7, D30, and D90 economics.
This is not a creative or audience tweak. It is a measurement and bidding reconfiguration that forces the algorithm to spend against the version of the user the business cares about. The Admiral Media team treats retention as a UA input, not a post-acquisition outcome.
The Admiral Media Retention-First UA Framework
Use this framework to convert a traditional install-led account into a retention-first account. Each step is required. Skipping one breaks the chain.
- Define the retention contract. Decide which retention milestone the business will treat as the acquisition success metric. For most subscription apps the Admiral Media team manages, this is D7 retention combined with trial start; for ad-monetized apps it is D7 retained sessions plus rewarded ad impressions; for dating and social apps it is D7 retention plus second-session signal. The contract is the single number every campaign is judged against.
- Instrument the proxy event. Pick a proxy event that fires inside the attribution window, correlates strongly with the retention contract, and has enough daily volume to clear the platform’s learning phase threshold. On iOS this often means engineering an AdAttributionKit or SKAdNetwork conversion value that encodes both early engagement and a paid signal. On Android it means firing a custom event for D3 retained session or trial start.
- Move bidding downstream. Reconfigure Google App Campaigns, Meta App Promotion, Apple Search Ads, and TikTok campaigns to optimize for the proxy event rather than installs. Where supported, switch from tCPI to tROAS or value-based bidding using modeled conversion values for the proxy event. Allow at least two weeks for the learning phase to complete before reading results.
- Rebuild the creative testing brief around retention. Brief creatives to attract the user who will retain, not the user who will tap install. This typically means longer hooks, more category-specific language, and onboarding-aligned messaging. The Admiral Media team uses creative tags to track which concepts produce the highest D7 retention, then concentrates spend on those tags.
- Read cohorts, not days. Replace the daily CPI dashboard with a weekly cohort dashboard that reports D0, D7, D30, and where possible D90 retained users and revenue. Make budget decisions on cohort economics. A campaign with a higher CPI and a better D30 cohort revenue wins.
- Reallocate on cohort signal. Every two weeks, redistribute spend toward audiences, geos, channels, and creative tags with the best D7-to-D30 retention curve. Cut campaigns where the curve flattens early, regardless of how cheap the install is. This is the step most accounts skip and it is where most of the lift comes from.
The framework works because it aligns the bidding algorithm, the creative brief, the reporting layer, and the reallocation cadence around the same downstream signal. Most accounts have one or two of these aligned and the rest pointing at installs. That misalignment is the leak.
How retention-first UA changes channel economics
Each major UA channel responds differently when the optimization event moves downstream. The Admiral Media team sees the same directional pattern across accounts: channels with strong machine learning and rich signal infrastructure gain efficiency, while channels with thin event signal lose efficiency or require manual workarounds.
Google App Campaigns and Meta App Promotion are the channels that benefit most. Both have mature value-based bidding stacks that consume modeled conversion values and optimize toward predicted revenue. When the proxy event is wired correctly, tROAS bidding on these channels typically requires a minimum of 30 to 50 weekly conversion events to exit the learning phase. Apple Search Ads benefits in a different way: retention-first reporting exposes which keywords produce retained subscribers rather than installs, and the Admiral Media team shifts bids toward the retained-subscriber keyword set.
Channel comparison: install-led versus retention-first economics
| Channel | Best optimization event under retention-first | Bidding strategy | Min weekly conversions to exit learning | Typical retention-first impact |
|---|---|---|---|---|
| Google App Campaigns (UAC) | In-app purchase or modeled value | Target ROAS | 30 to 50 conversion events | Stronger compounding once tROAS stabilizes |
| Meta App Promotion | Subscription or value optimization | Lowest cost with value rules, or minimum ROAS | 50 events per ad set per week | Higher quality cohorts, lower volatility |
| Apple Search Ads | Retained subscriber by keyword | Manual CPT with retention-weighted bids | Per-keyword post-install signal | Higher LTV at similar CPI |
| TikTok For Business | App event optimization on monetization event | Cost cap with value floor | 50 events per ad group per week | Variable, creative-dependent retention lift |
| iOS via AdAttributionKit / SKAN | Conversion value encoding retention plus revenue | Value-based via mapped CV | Privacy-thresholded postbacks | Recovered iOS efficiency when CV is rebuilt |
The Admiral Media team uses this matrix to set channel-level expectations before changing bidding. The numbers in the volume column are platform-documented minimums, not Admiral Media estimates, and they govern how long the team waits before reading results. Underestimating the learning phase is the second most common cause of retention-first programs failing in their first month.
What the Admiral Media account data shows
Three Admiral Media case studies illustrate the retention-first pattern from different angles: a fitness subscription app scaling from a small base, a brain-training subscription app with mature campaigns, and a dating subscription app with a broken iOS measurement stack.
Fastic: retention-led scaling from near zero to one million users
Admiral Media managed Fastic’s app growth across Facebook, Google, Apple Search, Snapchat, Pinterest, and TikTok, with creative testing and audience exploration prioritized against post-install monetization rather than install volume. The Admiral Media team scaled spend by concentrating budget on channels and creatives that produced retained, paying users, not the cheapest installs.
- +639% installs: campaign scale comparing May 2020 with December 2019, driven by channels validated against post-install monetization.
- +1655% purchases: paid users grew far faster than installs, the defining signature of a retention-first cohort: the marginal install was disproportionately a paying user.
- +439% revenue: revenue compounded ahead of CPI, the outcome the framework is designed to produce.
- -50% cost per purchase: efficiency at the monetization layer improved as the optimization event moved downstream from installs to purchases.
- +952% monthly active users: MAU growth outpaced install growth, indicating retained cohorts rather than churn-and-replace acquisition.
Full results: Fastic case study.
NeuroNation: retention-weighted creative testing for a mature account
Admiral Media ran intensive creative testing and channel exploration for NeuroNation with results judged by net cohort revenue and ROAS rather than install volume. The Admiral Media team built a categorized testing system that ranked winners using the pRank methodology, which weighted client-defined downstream KPIs above install-level efficiency.
- +117% ROAS: revenue per ad dollar more than doubled as the testing system surfaced creatives that produced retained, paying users.
- +42% net cohort revenue: cohort-level revenue grew faster than install volume, the key signal of a retention-first reallocation working.
- +66% installs: install scale grew alongside, but well behind, revenue and cohort gains.
- +32% purchases: paid user growth compounded.
- -39% CPI: install efficiency improved even though CPI was not the optimization target, a common second-order effect when the downstream signal is wired correctly.
Full results: NeuroNation case study.
FET: rebuilding iOS measurement to expose true monetization
Admiral Media took over FET’s paid UA when iOS was failing to convert, signup-to-subscription conversion was low, and subscription CPA was high. The Admiral Media team rebuilt the SKAN conversion value mapping from scratch to align campaigns with true monetization drivers, ran 60 ad iterations across 9 unique concepts, and identified the right optimization event combination for subscription performance.
- -66% cost per subscription: subscription CPA fell by two thirds once campaigns were optimized against the correct downstream event.
- +162% subscriptions: paid subscription volume more than doubled.
- +181% conversion rate: signup-to-subscription conversion lifted sharply, confirming that the new optimization event pulled in higher-intent users.
- iOS turned profitable: SKAN remapping rebuilt iOS into a profitable channel and rebalanced spend across platforms.
Full results: FET case study.
Cohort revenue lift across three retention-first engagements
How to instrument the proxy event correctly
The proxy event is the single most important decision in a retention-first program. Pick the wrong event and the algorithm optimizes against noise. The Admiral Media team uses three rules to select it. The event must correlate with the retention contract at the cohort level. It must produce enough weekly volume to exit the platform’s learning phase. And it must fire inside the attribution window so the platform can attribute it back to the impression.
For iOS, the Admiral Media team usually rebuilds the AdAttributionKit or SKAdNetwork conversion value to encode both an early retention signal and a paid signal in a single value. Apple’s own developer documentation on AdAttributionKit and on SKAdNetwork defines the postback windows and value-encoding constraints, which the team treats as hard limits when designing the value schema. For Android, the proxy is usually a custom event on D3 retained session or trial start, fired through the platform’s app event SDK.
Reading cohorts without fooling yourself
Cohort reporting is where retention-first programs are most often misread. The Admiral Media team applies three checks to every cohort dashboard before acting on it. The first is sample size: a cohort smaller than the platform’s privacy threshold or smaller than roughly 500 users is too noisy to drive bidding decisions. The second is age: a cohort needs to be at least seven days old before D7 retention is meaningful, and at least 30 days old before D30 is. Acting on a four-day-old cohort is a common, expensive mistake. The third is mix: a cohort drawn from a single channel and creative tag is decision-grade; a cohort blended across channels hides the signal the program is trying to surface.
External benchmarks help but should never substitute for the account’s own cohort data. Industry retention curves from sources such as AppsFlyer performance reports or Adjust mobile benchmark reports are useful for ballparking what good looks like by vertical, but the Admiral Media team always overweights first-party cohort data when making spend decisions.
What retention-first does not solve
Retention-first UA does not fix a weak product. If the app does not retain organically, no bidding configuration can manufacture cohort revenue. The Admiral Media team treats organic D1 to D7 retention as a precondition for the program: below a category-specific floor, the recommendation is to fix product and onboarding first and rerun the program once organic retention is healthier. Retention-first UA amplifies a retaining product; it cannot rescue a churning one.
Retention-first UA also does not eliminate measurement uncertainty on iOS. SKAdNetwork and AdAttributionKit postbacks remain delayed, aggregated, and privacy-thresholded. The Admiral Media team manages this with conversion value engineering and modeled value bidding, but a perfectly clean iOS cohort is not on offer. The framework is built to be robust to that noise, not to remove it.
Frequently Asked Questions
What is retention-first user acquisition?
Retention-first user acquisition is a paid media strategy that selects campaigns, audiences, channels, and creatives based on Day 7 and Day 30 retention and cohort revenue, rather than on install volume or cost per install. In Admiral Media’s accounts, the team moves the bidding optimization event downstream from install to a retention or monetization proxy, and reallocates spend on cohort signal every two weeks.
Why does optimizing for installs hurt subscription apps?
Install-led bidding finds users for whom the install event is cheap, not users for whom downstream monetization is high. In subscription categories, those two groups overlap only weakly. The cheap install is often a low-intent user who opens the app once and never returns. Subscription apps lose more cohort revenue to this mismatch than to any other UA factor, in the Admiral Media team’s experience.
Which optimization event should I move to first?
Pick the earliest event that correlates with the retention contract and produces enough weekly volume to exit the learning phase. For most subscription apps the Admiral Media team manages, the first move is from install to trial start or D3 retained session. Trial start usually has the strongest correlation with subscription, and D3 retained session usually has the highest volume. Choose based on which constraint is binding in your account.
How long should I wait before reading retention-first results?
Wait at least two weeks for the learning phase to complete on each major channel, then read cohorts that are at least 30 days old before judging program impact. Acting on D7 retention alone or on a fresh cohort is the most common cause of premature program cancellation. The Admiral Media team uses a 90-day evaluation window for full program ROI.
Does retention-first UA work for ad-monetized apps?
Yes, with a different retention contract. For ad-monetized apps the Admiral Media team usually uses D7 retained sessions plus rewarded ad impressions as the contract, because impressions are the closest available revenue proxy. The mechanics of the framework are identical: choose a proxy event that correlates with the contract, move bidding downstream, and reallocate on cohort signal.
How does iOS measurement constrain retention-first UA?
iOS postbacks under SKAdNetwork and AdAttributionKit are delayed, aggregated, and subject to privacy thresholds. The Admiral Media team handles this by engineering conversion values that encode both retention and revenue signals in a single value, and by using modeled value bidding where the platform supports it. The constraint is real and permanent, but it does not block a retention-first program. The Admiral Media team has rebuilt iOS measurement stacks from scratch on multiple accounts and recovered profitable iOS spend.
What is the fastest way to start a retention-first program?
Start with one channel and one cohort. Pick the channel with the most spend and the cleanest signal, instrument a single proxy event, and switch one campaign to value-based or downstream-event bidding. Build the weekly cohort dashboard before changing anything else. Once the first channel is stable and the dashboard is trusted, extend the same setup across other channels. Programs that try to convert every channel simultaneously almost always fail.
