App Install Growth and Retention Messaging for a Nairobi Fintech Lender

Context

The client operates a mobile lending app in Nairobi offering short term, small ticket loans disbursed directly to M-Pesa, targeting the mass market segment of gig workers, small traders, and salaried employees who need quick access to cash between paydays. The category is crowded. At the time we started, dozens of similar apps competed for the same Android users on the Play Store, many funded aggressively and willing to burn cash on installs regardless of whether those users ever borrowed or repaid.

The client had raised a modest seed round and needed to show investors a path to sustainable unit economics rather than just download numbers. Their monthly acquisition budget was around KES 3.2 million, almost entirely spent on Facebook and Google App campaigns, and their finance team had started asking pointed questions about why install volume kept climbing while active borrowers barely moved.

The Problem

The core issue was that the acquisition strategy optimized for the wrong outcome. Campaigns were structured around cost per install, which rewarded creative that promised the largest, fastest loans with the fewest conditions, exactly the kind of promise that attracted users who installed, browsed, got declined or approved for a token amount, and then churned immediately. Cost per install looked healthy on a dashboard while the business underneath was hollow.

Retention was worse than acquisition. Once a user completed their first loan cycle, there was no structured nudge to bring them back for a second one, no differentiated messaging for good repayers versus users who had struggled, and no attempt to use M-Pesa repayment behavior as a signal for anything beyond the credit decision itself. Safaricom's dominant position in mobile money meant nearly every user's financial life ran through M-Pesa, yet the client's marketing never referenced that behavior to build trust or relevance.

Finally, there was a reputational risk sitting quietly in the background. Digital lenders in Kenya had drawn regulatory and public scrutiny over aggressive collection practices and unclear pricing, and the client's existing ad creative leaned on urgency language that skated close to the kind of messaging regulators had flagged industry wide.

What We Did

1. Reworked acquisition targets from install to activation. We rebuilt the Google App and Meta campaigns around cost per activated borrower, defined as a user who completed KYC and disbursed a first loan, rather than raw installs, which meant restructuring bidding strategy and giving both platforms enough activation signal through the app's event tracking to optimize properly.

2. Rewrote creative around transparency rather than urgency. We replaced countdown timers and maximum loan size headlines with creative that led with clear terms, real repayment examples, and straightforward eligibility criteria, betting that a segment of the market was tired of aggressive lending ads and would respond better to a calmer, more credible tone.

3. Built a first loan onboarding sequence. New borrowers received a structured sequence of in app messages and SMS over their first loan cycle explaining exactly how repayment via M-Pesa worked, what happened if they repaid early, and what their next available loan limit would be on time repayment, addressing the biggest source of first cycle confusion we found in support ticket data.

4. Segmented retention messaging by repayment behavior. We split the user base into on time repayers, late but resolved repayers, and defaulted users, and built distinct SMS and push messaging tracks for each, including a loyalty style limit increase offer for consistent on time repayers that had not existed before.

5. Introduced win back campaigns tied to loan cycle timing. Rather than generic reminders, we timed win back messages to arrive a few days before a user's typical payday cycle based on their historical borrowing pattern, since a message offering a loan the day after payday performed far worse than one that arrived just before a likely cash crunch.

6. Set up a weekly unit economics dashboard. We connected acquisition spend, install, activation, first loan disbursement, and repayment data into one weekly view so the founders could see blended cost per activated borrower and repeat borrowing rate rather than relying on separate exports from the ad platforms and the loan management system.

Results

Cost per activated borrower, which is the metric that mattered far more than raw install cost, fell as targeting improved, but the headline efficiency gain the client cared most about was a 47% reduction in cost per install once we cut the lowest quality acquisition sources that were driving volume without activation. Activation rate, meaning the share of installs that completed KYC and took a first loan, rose to 61% from a starting point in the high thirties.

The retention segmentation work had the largest downstream impact on the business. Repeat borrowers, meaning users who took a second loan within ninety days of their first, grew 3.1x over the engagement period, driven mostly by the payday timed win back messages and the limit increase incentive for reliable repayers. Support tickets related to repayment confusion dropped noticeably after the onboarding sequence launched, which the client's operations team flagged as an unplanned but welcome side effect.

What We Would Do Differently

We spent the first six weeks optimizing acquisition before touching retention, and in hindsight the sequencing should have been reversed. Fixing the onboarding confusion and building repayment segmentation earlier would have made every subsequently acquired user more valuable from day one, rather than pouring improved traffic into a leaky retention funnel for over a month.

We would also push earlier for API level integration with the credit decision engine so that acquisition targeting could exclude segments the credit team already knew were unlikely to be approved, since a portion of our activation spend in the first two months went toward installs that were technically activated by our narrower definition but rejected at underwriting, which frustrated users and wasted budget on both sides.