Decoding the digital age
The Leaky Bucket
For over a decade, the app economy was defined by fast growth. Now retention is the key metric for many app companies.
App economics
The Leaky Bucket
Why retention, not acquisition, now defines the economics of the app economy.
Written by Peter Franks
For more than a decade, the app economy was defined by how quickly a product could grow. Cheap distribution, expanding smartphone penetration, forgiving capital, and increasingly sophisticated performance marketing let consumer apps scale at a pace that would have seemed implausible in earlier eras of software. If users churned, you simply bought more. If engagement dipped, you pushed notifications harder. If monetisation lagged, you introduced another tier and told investors you were still early in the journey.
That era is over. Not because people stopped downloading apps, but because the economics of keeping them changed. The modern app economy is constrained less by reach than by leakage. Users arrive in vast numbers, sample freely, and leave just as quickly. Acquisition still matters, but it no longer saves you. In many categories, the bucket leaks faster than it can be filled.
This is not a temporary correction. It is a structural shift in how consumer software behaves, how users value digital products, and how platforms extract rent from those who depend on them. Retention has become the defining economic problem of apps, subscriptions, and content businesses alike.
The leaky bucket is not a metaphor, it is the model
Everyone in mobile has heard the phrase. The leaky bucket is the mental model that reminds you why top line growth can be illusory: you can pour aggressively at the top, but if the bottom is full of holes the system never compounds. What is new is not the metaphor, but the scale of the leakage and the number of companies discovering it at the same time.
A decade ago, many categories could survive with mediocre retention because acquisition was cheap and relatively predictable. Teams treated churn as an annoyance rather than an existential variable. Today, churn is the variable that determines whether a company has a business or a treadmill.
Acquisition is not just expensive now, it is unreliable
User acquisition did not suddenly become costly. It became less controllable. Apple’s privacy changes, most notably App Tracking Transparency, did not kill performance marketing, but they changed its character. Attribution became probabilistic. Optimisation cycles slowed. Marginal returns became harder to forecast. In practice, this means many teams can still buy users, but struggle to know which spend is genuinely incremental.
Distribution and measurement
Why attribution became harder after privacy changes
User acquisition did not suddenly become costly. It became less controllable. The difference is measurement: what you can observe quickly and confidently versus what you must infer later.
Before: more observable
Faster, more granular
Ad clickcampaign, creative
Installdevice-level event
Attributionwho converted, from where
- Clear linkage between spend and outcomes.
- Short feedback loops for optimisation.
- More confidence in marginal returns.
After: less observable
Delayed, aggregated
Ad clickcampaign, creative
Privacy layer
Consent gating and reduced identifiers
Attributionmodelled and aggregated
- Coarser signals, fewer deterministic links.
- Slower learning because feedback arrives later.
- Incrementality fog: harder to know what was truly caused by spend.
The practical result: teams can still buy users, but struggle to know which spend is genuinely incremental,
and which would have happened anyway.
If you want the platform view of how this ecosystem works, Apple’s own App Store documentation is a useful starting point, not because it is neutral, but because it clarifies how distribution and monetisation are structured.
TikTok temporarily reopened the tap for many consumer businesses, offering scale, novelty, and algorithmic reach. But it introduced its own fragility. Creative fatigue sets in quickly. Formats change without warning. Audiences are rented, not owned. Distribution feels abundant right up until it disappears.
The short term response was predictable: pour faster. Increase spend. Expand channels. Chase virality. In the leaky bucket era, pouring faster often reveals the problem rather than solving it. If the product cannot retain, paid acquisition becomes a way of purchasing churn.
The LTV to CAC equation is the business
This is why the most important relationship in consumer apps remains brutally simple: customer acquisition cost versus lifetime value. If you spend £30 to acquire a user, but only earn £25 back over the life of that user, growth destroys value. If you spend £30 to earn £300, growth compounds.
In practice, teams talk about LTV to CAC ratios, payback periods, and contribution margins. The labels vary, but the logic does not. When you strip away narrative, an app business succeeds when it can convert attention into cash at a sustainable cost.
Retention sits at the heart of this equation because it is the primary driver of lifetime value. Higher retention increases total revenue per user, lowers the effective acquisition cost via word of mouth, and creates the time window to cross sell, upsell, or introduce new monetisation layers.
Unit economics
CAC versus LTV: the equation that decides if growth is worth it
Consumer apps succeed when the value of a user over time is higher than the cost of acquiring them. Everything else is implementation detail.
CAC (Customer acquisition cost)
Cost
The average amount you spend on marketing and sales to acquire one new user or subscriber.
LTV (Lifetime value)
Value
The total gross profit you expect to earn from that user over the entire time they stay with you.
The rule of thumb
Decision
If LTV is lower than CAC, growth destroys value. If LTV is much higher than CAC, growth compounds.
Value destructive
LTV < CAC
CAC
£30
LTV
£25
Net value per user
-£5
Value compounding
LTV > CAC
CAC
£30
LTV
£300
Net value per user
+£270
Why retention matters: retention is the main driver of LTV. It determines how long a user has to generate revenue,
and it creates the time window to upsell, cross-sell, or introduce new monetisation layers.
In practice teams add ratios and payback periods. But the core logic stays the same: LTV must exceed CAC.
Subscription, ads, marketplaces: three app economies inside one
One reason the app economy feels confused is that it is not one economy. It is several, overlapping on the same devices and distribution rails. At a high level, most consumer apps monetise via one of three models: subscription, advertising, or marketplace take rate.
Subscriptions exchange a recurring fee for ongoing value. Ads exchange attention for revenue, usually at the cost of user experience. Marketplaces monetise transactions and can, in theory, scale efficiently once liquidity is reached.
Each model has a different relationship with retention. Subscription businesses live or die by renewal. Ad businesses can sometimes tolerate churn because they monetise quickly, but they still depend on repeat sessions to build inventory. Marketplaces depend on retention from both sides, buyers and sellers, and often have strong switching costs once network effects take hold.
Many of the most prominent consumer businesses blend models. A growing number of subscriptions are freemium, with free users subsidised by ads or by a funnel into paid tiers. This is the closest analogue to free to play mobile games, where the majority of users pay nothing but provide liquidity, data, and social context for those who do.
Why subscription content apps feel especially fragile right now
Subscription content businesses, language learning, wellness, mindfulness, health, fitness, are often vulnerable for two reasons. First, they sit in the discretionary zone of household budgets. Second, they compete against an expanding pool of cheap alternatives.
The post-Covid economy has been less forgiving to discretionary spend. Users have learned to cancel, pause, and rotate subscriptions tactically. A monthly subscription is no longer a commitment, it is an evaluation that repeats every month.
At the same time, content supply has exploded. In media, every niche has multiple newsletters, podcasts, and YouTube channels. In wellness, there are endless guided meditations, breathwork routines, and fitness plans. AI makes the problem worse by lowering the cost of producing competent content. When supply becomes infinite, pricing power collapses.
This is why retention has shifted from being a byproduct of content quality to being a deliberate system. Content alone is not sticky enough. The stickiness comes from identity, progress, community, and the subtle pressure of systems that users do not want to break.
Duolingo: a language app built like a live service game
Duolingo is the clearest illustration of what retention-first product design looks like in a subscription and freemium context. It is often described as a gamified education app. A more precise description is that it is a live service consumer system that happens to teach languages.
The social layer is central. Leaderboards and leagues turn learning into weekly competition. Friends quests create cooperative goals that encourage mutual accountability. Friend streaks formalise a shared commitment. These are not superficial features. They are retention infrastructure that increases the cost of quitting, emotionally and socially, without feeling like a lock in.
Duolingo itself explains the mechanics of these features in posts such as how Leagues and leaderboards work, Friends Quests, and Friend Streaks. The point is not the details, but the pattern.
This pattern is familiar to anyone who has built or studied mobile games. Competitive ladders, cooperative quests, daily streaks, and limited time rewards map cleanly to live game operations. A live ops team is, in effect, a retention engineering team. It watches behaviour, runs experiments, ships new content, and uses event cadence to pull users back in.
Duolingo borrows from this playbook with intent. The app behaves like a live product, constantly iterating, adding new challenges, and adjusting feedback loops. It uses streak pressure in the same way games use daily rewards. It uses leagues in the same way games use ranked seasons. It uses social reinforcement in the same way games use guilds and team events, but packaged for a mainstream audience.
Retention design
How Duolingo turns social mechanics into retention infrastructure
Duolingo’s social features are not cosmetic. They translate live game mechanics into a mainstream subscription product, increasing the emotional and social cost of quitting.
Leagues and leaderboards
Weekly rankings that turn learning into ongoing competition.
Friends quests
Shared goals that create cooperative accountability.
Friend streaks
Formalised commitment to show up together.
Social pressure
Progress becomes visible to others, not just the user.
Loss aversion
Quitting means losing status, streaks, or letting others down.
Habit formation
Regular participation becomes the default behaviour.
Higher retention over time
Users stay longer, return more frequently, and give monetisation systems time to work. The product behaves like a live service, not a static content app.
This structure mirrors live game operations: ranked seasons, cooperative events, and streak-based rewards.
Duolingo adopts the same playbook, packaged for a mainstream audience.
The result is not simply higher engagement. It is a different relationship between user and product. Progress is tracked, visible, and socially legible. Quitting means losing status, breaking a streak, or letting down a partner. This changes the emotional cost of churn, which is often more powerful than any rational calculation of price.
This also creates a lesson for the broader app economy. As the cost of acquisition rises, the highest leverage feature is often not a new paywall or a new marketing channel, but a new mechanism that makes users feel embedded in a system they do not want to abandon.
Peloton: the Covid darling that met the churn wall
If Duolingo represents the best of retention design, Peloton is a reminder of what happens when a high growth narrative runs into churn. During the pandemic, Peloton captured a perfect storm: closed gyms, homebound consumers, and a premium identity brand.
The challenge was always what happened next. As the world reopened, usage patterns normalised. A connected fitness subscription is valuable when it becomes a habit. It is fragile when it is treated as a phase.
Peloton’s own investor reporting shows how closely the business is tied to subscriber dynamics, including churn and net adds, as seen in its quarterly results updates. The specifics vary quarter to quarter, but the lesson is stable: when churn rises and adds slow, the economics of the model compress fast.
Peloton’s story matters because it reflects a wider truth about subscription content and lifestyle businesses. Many of them grew explosively when external conditions created forced habits. When those conditions faded, the retention system was tested. Some passed. Many did not.
This is why churn is so dangerous. It hits revenue twice. First, you lose recurring revenue from churned users. Second, you have to replace those users with new acquisition spend. The bucket leaks and you pay for the water.
Why the app economy is becoming a data science problem
As leakage becomes the constraint, retention becomes the focus. And retention at scale is, increasingly, a data science problem. Not in the simplistic sense of dashboards, but in the deeper sense of understanding cohorts, segments, and behavioural patterns.
Modern consumer apps run like ongoing experiments. Teams segment users by intent, engagement depth, price sensitivity, and likelihood to churn. They model lifetime value across cohorts. They test onboarding flows, pricing prompts, content cadence, notifications, and social features. They ship continuously, not in annual releases.
This is the shift from product as an object to product as a system. A decade ago, an app launched and then added features. Today, an app is a live product that constantly re shapes itself around user behaviour. The boundary between product, growth, and monetisation has blurred into one loop: observe, test, ship, measure, repeat.
The best teams treat retention like a discipline. They know which segments churn in week one, which segments churn at month three, and which segments become long term users. They know which experiences create habit and which create fatigue. They invest in features that look small on a roadmap but large on a retention curve.
The Great Game: monetisation is behavioural design
Retention is inseparable from monetisation because the same psychological levers often drive both. Mobile games understood this first. Their economics forced them to become laboratories of behavioural design.
If you want a deeper exploration of how monetisation systems shape behaviour, see No Latency’s own piece The Great Game of Monetization. The core idea applies beyond games: monetisation works best when it is integrated into the experience rather than bolted on at the end.
In the leaky bucket era, this integration becomes more important because users are more sceptical and more price sensitive. Aggressive monetisation can lift short term revenue but increase long term churn. A retention led system searches for mechanisms that feel fair, rewarding, and voluntary, even when they are strategically designed.
Which categories look durable, and which look exposed
The app economy is not collapsing, but it is bifurcating. Some categories have structural advantages that reinforce the bucket. Others rely on habits that are easy to break and content that is easy to replace.
Mobile games remain the most durable because live ops, progression systems, and social competition are native to the category. Dating remains strong where network effects create switching costs, though retention is cyclical. Finance retains users where regulation, trust, and friction make switching painful.
The exposed categories are those that sell content or motivation without a strong system. Wellness, mindfulness, generic fitness, and broad education products are vulnerable to subscription fatigue and to an expanding set of low cost alternatives. AI accelerates this by making content cheaper to produce and easier to personalise outside of a paid subscription.
The emerging edge is hybrid. Apps that combine utility with entertainment, or content with community, tend to leak less. So do products that bundle services rather than sell isolated features. The unifying theme is that retention is built through systems users participate in, not content they consume and move on from.
What the leaky bucket era demands
It is tempting to treat this moment as a crisis of growth. It is more accurately a crisis of compounding. The companies that win will be those that design for staying, not just for downloading.
That implies practical changes. Acquisition teams must work alongside product, not around it. Pricing must be treated as part of experience design. Retention mechanisms must be explicit, not accidental. And live product operations, content cadence, events, experimentation, become core capabilities even outside games.
The lesson of Duolingo is not that every app should become a game. It is that the best consumer apps borrow the operational discipline of games: ship continuously, create progress, build social context, and respect cadence.
The lesson of Peloton is that narrative cannot compensate for churn. When a product is tied to habit, retention is the business. If the habit breaks, everything else follows.
The leaky bucket metaphor endures because it captures a truth that founders and investors prefer to avoid. You cannot build a durable business by pouring forever. At some point, you must fix the leaks.
The app economy has entered that phase. Growth is harder. Capital is more selective. Users are more sceptical. The survivors will not be those that spend the fastest, but those that design systems users choose to remain inside.
Retention is not a metric. It is the model.
