CPI is a metric. Retention is the business.

CPI (cost per install) measures how much you paid for each download. Cost-per-activated-install measures how much you paid for each user who reached the first value moment in the app. These two numbers can differ by 3–5x depending on onboarding quality and install source. Optimising for CPI without measuring post-install activation rate leads to acquiring cheap installs from users who never engage — which produces a low CPI headline number and disastrous LTV economics. We set up activation event tracking from the start and use it as the primary campaign optimisation signal rather than install volume.

iOS signal recovery operates on three layers. First, SKAdNetwork configuration — Apple's privacy-preserving attribution framework — must be correctly set up through your MMP to receive any deterministic signal from iOS campaigns. Second, Meta's App Events API (the mobile equivalent of CAPI) sends in-app events server-side, partially recovering signal that ATT opt-outs would otherwise block. Third, probabilistic modelling through your MMP fills gaps where deterministic attribution is unavailable. The combination typically recovers 50–65% of the signal that ATT removed. Apps running iOS campaigns without all three layers in place are making budget allocation decisions on noise.

D7 retention varies significantly by app category. Gaming apps typically have D7 retention of 15–25%. Utility and productivity apps can reach 30–45% with well-designed onboarding. Social and communication apps are often 25–35%. The most important benchmark is your own D1-to-D7 retention ratio — if more than 60% of D1 actives are gone by D7, the re-engagement and habit formation architecture needs attention before UA spend is increased. Industry median D7 retention across all categories is around 20%, which is the floor for economically sustainable acquisition. Below 15%, scaling paid UA is likely to compound losses rather than drive growth.

We run a four-stage pipeline: concept identification, hook testing, full creative production, and scale. In the first stage, we identify 8–12 distinct creative concepts based on user research, competitor analysis, and product benefits. In the second stage, we test only the first 5 seconds — different hooks against the same offer — using low-budget tests to identify which concepts earn attention. Winning hooks move to full creative production. Only creatives that pass CPI and post-install quality thresholds move to scale. This approach means most creative spend goes on concepts that have already proven their hook, not on full productions that may fail immediately. Creative refresh cadence is set by frequency cap data from the MMP, not by calendar.

All three are capable MMPs for most consumer app use cases. AppsFlyer is the most widely adopted and has the broadest ad network integrations, making it a safe default for apps running across Meta, TikTok, Google, and multiple smaller networks. Adjust has strong fraud prevention tooling and is favoured by gaming and high-spend UA teams. Branch is particularly strong for apps with significant web-to-app journeys or deep linking requirements. For GCC-focused apps, all three MMPs handle the local ad landscape. The correct answer depends on your existing tech stack, current integrations, and whether your primary UA channels are already integrated with one provider. We work with all three and will recommend based on your specific situation.

Onboarding drop-off is almost always caused by one of three patterns: too many steps before the first value moment, account creation required before the user sees any value, or a permissions request (push notifications, camera, location) appearing before the user has a reason to grant it. The intervention for each is specific: map the minimum steps required to reach first value, defer account creation to post-activation, and time permission requests to moments immediately after the user has experienced value rather than immediately on launch. We audit the onboarding flow with session recording data and funnel step analysis, identify the highest-drop-off step, and test a specific intervention before moving to the next. D1 retention improvement of 20–40% is achievable in most apps within 6–8 weeks of a systematic onboarding audit.

CPI in UAE is materially higher than global averages because of the high-income market dynamics and relatively smaller addressable audience. On Meta, iOS CPI for consumer apps in UAE typically ranges from AED 8–25 depending on category, creative quality, and targeting precision. Android CPI is 30–50% lower. Google UAC and Apple Search Ads tend to produce lower CPI but often at lower volume. These ranges are highly variable — a well-tested creative hook can produce CPI at the lower end of the range, while broad creative without hook testing typically produces CPI at the upper end or above. The more useful metric is cost-per-activated-install, which accounts for how many of those installs actually engage with the app.

The sequence matters: MMP attribution should be configured and validated before any paid UA spend, because every install and in-app event from day one needs clean attribution. After MMP is in place, the next prerequisite is a D1 retention rate above 25% — if D1 retention is below this, paid UA spend is adding users to a leaky bucket. The most common mistake is starting paid UA before the onboarding flow has been audited and optimised. We typically recommend 2–4 weeks of organic install analysis with MMP data in place before launching paid campaigns — the organic cohort data establishes retention baselines and identifies onboarding friction before you add acquisition cost to the equation.