In app growth, failure is rarely sudden. It announces itself early, often quietly, through patterns that experienced operators learn to recognize long before a chart collapses or a budget runs dry. The difficulty is not that these signals are obscure, but that they are easy to rationalise away while momentum still exists.

That diagnostic perspective shaped a recent App Talk recorded at Business of Apps Berlin 2025, where Samet Durgun, better known as the Growth Therapist, approached app growth less as a problem of tactics and more as a system that either supports durable outcomes or undermines them. Rather than prescribing optimisations, he focused on the recurring conditions that separate leaders from laggards across categories, business models, and stages of maturity.

What follows is not a checklist of tricks, but a map of structural fault lines that tend to appear well before performance collapses.

Cash flow as a structural constraint, not a finance problem

One of the first warning signs appears long before campaigns collapse or retention curves flatten. It shows up in timing.

App businesses, particularly subscription-led ones, live with a structural delay between revenue generation and payout. That lag, often stretching to six weeks via app stores, turns scale into a risk multiplier if it is not explicitly modelled.

“You are already spending money that you have not really received yet,” Durgun noted, pointing out that growth decisions made without a reliable LTV baseline can burn future runway. The danger is not ignorance, but optimism. Teams assume the next payout will arrive in time, until it doesn’t.

Cash flow discipline, in this framing, is less about austerity than visibility. If growth requires bridging gaps, those gaps need names, numbers, and contingency plans, not faith.

Retention as the only metric that compounds

Retention is often discussed as a post-install metric, something to be optimised once acquisition is stable. That sequencing misses the point.

From the first app open, retention is already being shaped by onboarding friction, paywall clarity, perceived value, and the coherence of the initial experience. “It’s not just day seven or day thirty,” Durgun observed. “It’s everything from the first second to whether someone still cares months later.”

“It’s not just day seven or day thirty. It’s everything from the first second to whether someone still cares months later.”

Rather than chasing absolute benchmarks, he argued for working within healthy ranges, informed by credible industry medians and top-quartile data, rather than averages that flatten meaningful variation. Retention, treated this way, becomes less a single KPI and more a continuous diagnostic signal.

Diagnosing broken app growth

Source: Business of Apps via YouTube

Instrumentation as strategy, not plumbing

Event setup is usually framed as a technical prerequisite rather than a strategic decision. In practice, it determines what a growth system is capable of learning.

Many apps optimise around signals that are convenient rather than meaningful. A trial start, for example, may look like progress while masking immediate cancellation behaviour. More predictive signals often sit a step later in the journey, requiring deliberate definition and consistent transmission across analytics, attribution, and media platforms.

Instrumentation failures tend to surface indirectly. Teams struggle to explain performance differences between channels, cannot reconcile app store data with MMP reports, or adjust campaigns without understanding which behaviours actually correlate with revenue. These problems rarely originate in media buying. They begin upstream, in decisions about what the system is allowed to measure.

Attribution as an organising system

Attribution problems rarely announce themselves. They accumulate through small inconsistencies: delayed signals, missing parameters, revenue visible in one system but not another.

What stood out was how often these gaps persist even at scale. “I still see companies doing millions,” Durgun said, “but not really knowing which channel is driving what.”

The issue is rarely tooling. Most stacks already contain the necessary components. The failure lies in orchestration, deciding when and how data moves between product, MMPs, and acquisition platforms. Without that map, optimisation becomes guesswork disguised as analysis.

Creative scale and the illusion of cheap AI

The growing availability of AI-generated creative has shifted the conversation from scarcity to abundance. Volume is no longer difficult to achieve. What has become harder to see is the true cost of producing a winning asset.

Creative production carries opportunity cost, whether the labor is human or machine-assisted. Prompting, iteration, testing, and interpretation all consume time and attention. Measuring success by output count obscures the more relevant question: how much did it cost, in total effort, to arrive at a creative that actually performs?

The more useful metric is not cost per asset, but cost per winner. Without that lens, teams risk optimising for throughput while eroding emotional resonance, mistaking speed for effectiveness.

Spend without learning is not scale

Statistically significant data requires investment, but feeding algorithms without clear hypotheses quickly becomes wasteful. Spend that is not designed to generate learning is indistinguishable from spend that simply fails.

The warning sign here is activity that increases complexity without increasing clarity. More campaigns, more creatives, more channels, paired with diminishing insight into what actually works. Growth becomes louder but less informed.

Product-market fit as an ongoing test

The final signal is also the most familiar, and often the most misinterpreted.

Product-market fit is frequently treated as a milestone crossed early, then assumed permanent. In reality, it degrades unless actively maintained. Markets shift, expectations rise, and user tolerance narrows.

Apps built from genuine personal friction tend to recalibrate faster, not because they are purer, but because their creators feel the problem directly. “If you would not use your own app,” Durgun said, “it’s very hard to know when it stops helping.”

“If you would not use your own app. It’s very hard to know when it stops helping.”

Focus as a growth strategy

The conversation ended not with predictions, but with restraint.

Rather than chasing new tactics, channels, or formats, the recommendation for the year ahead was focus. Focus on what already works. Focus on the strongest product. Focus on signals that matter.

Referencing Jeff Bezos, Durgun framed it simply: what matters most is what will not change. Relevance, clarity, and value remain stubbornly resistant to trend cycles.

In growth, as in therapy, progress often begins not by adding something new, but by understanding what is undermining the system already in place.