Carrying Capacity: Why Your App's Growth Has a Ceiling

There is a growth story that repeats so consistently it should be taught as a law. An app launches, grows quickly, and the team gets confident. Six months later growth slows. Twelve months later the monthly active user count is essentially flat, even though acquisition spend has never been higher and the marketing team has never worked harder. Everyone starts looking for someone to blame.

Nobody is to blame. The app hit its carrying capacity — the ceiling at which the number of users arriving exactly equals the number leaving. It is a mathematical property of the system, and it was entirely predictable from data the team already had on month three.

The formula

If you acquire 10,000 new monthly active users each month and 5 percent of your existing active users go dormant each month, your carrying capacity is 10,000 divided by 0.05, which equals 200,000 monthly active users. No matter how long you run that system unchanged, it converges on 200,000. Not 300,000. Not eventually a million. Two hundred thousand, and then flat forever.

The term comes from population ecology, where it describes the maximum population an environment can sustain. The mechanics translate almost perfectly: a birth rate, a death rate, and an equilibrium where they balance.

The intuition: a bucket with a hole

Picture a bucket with a hole in the bottom. You pour water in at a constant rate. The hole drains at a rate proportional to how much water is currently in the bucket — more water means more pressure means faster draining.

At first the bucket fills fast, because there is little water and therefore little drainage. As the level rises, drainage accelerates. Eventually the water level reaches a point where the drain rate exactly matches the pour rate, and the level stops changing. That level is the carrying capacity. It does not depend on how long you pour. It depends only on the pour rate and the size of the hole.

This is why growth feels amazing in the first year and then mysteriously stalls. Nothing broke. You simply approached the equilibrium that your acquisition and churn rates always implied.

Deriving it, briefly

Let A be new active users added per period, c be the fraction of active users who churn per period, and U(t) be active users at time t. Then:

U(t+1) = U(t) + A - c * U(t)

At equilibrium the population stops changing, so U(t+1) equals U(t). Substituting and solving:

0 = A - c * U   =>   U = A / c

That is the whole derivation. The model is deliberately simple, and its simplicity is the reason it is useful: it isolates the two variables that dominate long-run outcomes and ignores everything that only affects the path there.

Worked examples

Look carefully at rows two and three. They produce identical ceilings. One of them requires doubling your acquisition budget every month indefinitely. The other requires a retention improvement that, once shipped, costs nothing to maintain. This asymmetry is the single most important practical implication of the model.

And the asymmetry compounds. Going from 10 percent churn to 3.3 percent triples your ceiling. Tripling acquisition to reach the same ceiling means tripling a recurring cost forever, in a market where acquisition costs tend to rise rather than fall as you scale.

Why churn is the highest-leverage lever

Churn sits in the denominator. Anything in a denominator has nonlinear influence: a 10 percent improvement in acquisition raises the ceiling by 10 percent, but a 10 percent improvement in churn raises it by about 11 percent, and the gap widens dramatically as churn gets lower. Going from 5 percent to 4 percent churn — a one point change — raises the ceiling by 25 percent.

There is a second, less obvious reason churn dominates. Acquisition costs almost always increase as you scale, because you exhaust the cheapest, highest-intent audiences first and move into progressively broader ones. So the practical curve is worse than linear: doubling installs typically costs more than double. Retention improvements have the opposite dynamic — they are usually one-time engineering and design investments that keep paying.

How to compute yours from real data

The formula is trivial. Getting honest inputs is where the work is, and where most calculations go wrong.

Defining new users per period

Count new users who became genuinely active, not raw installs. An install that never opens the app twice was never part of the population — including it inflates the numerator and produces a ceiling you will never reach. Use your activation definition here if you have one, which is exactly where an identified aha moment becomes practically useful.

Defining churn rate

Churn is the fraction of users active in the previous period who are not active in the current one. Two mistakes to avoid. First, do not measure churn over a window shorter than your product's natural usage frequency — a weekly product will show terrifying daily churn that means nothing. Second, use a stable cohort window; blending users who joined last week with users who joined two years ago produces a churn rate dominated by whichever group happens to be larger this month.

In practice, compute monthly churn as: users active last month who were not active this month, divided by users active last month. Track it over at least six months, because a single month can be distorted by a marketing spike, a seasonal effect, or an outage.

A worked calculation

1. 1Last month: 84,000 monthly active users.
2. 2This month: 87,200 monthly active users, of which 9,400 are newly activated.
3. 3Users from last month who did not return: 6,200.
4. 4Monthly churn rate: 6,200 / 84,000 = 7.4 percent.
5. 5Carrying capacity: 9,400 / 0.074 = about 127,000 MAU.
6. 6Interpretation: you are at 87,200, roughly 69 percent of the way to your ceiling. Growth will visibly decelerate within the next few quarters.

That last line is the point of the exercise. It converts a vague feeling that things are slowing into a specific, dated expectation — and it does so early enough to act.

The three levers, in order of leverage

1. Reduce churn

Highest leverage, hardest work, most durable. Churn reduction is fundamentally a product problem rather than a marketing one. The most productive starting points, in rough order of typical impact:

• Improve activation. Users who never reach the aha moment churn at multiples of the base rate. Early-lifecycle churn is usually the largest single block of total churn, and it is the most fixable.
• Find the churn cliff. Plot retention by day and find where the curve drops most steeply. That drop has a cause — a paywall, a permission request, a bug, an empty state — and it is usually identifiable from session recordings or funnel data.
• Build habit triggers. External triggers such as well-timed notifications and internal triggers such as recurring needs both extend lifespan, but only after the product has proven its value.
• Interview churned users. Twenty conversations with people who left will teach you more than a hundred with people who stayed. Retained users are survivorship bias made flesh.

2. Increase acquisition

Linear, expensive, and reversible the moment you stop. That does not make it wrong — it makes it a lever you should pull deliberately rather than reflexively. Acquisition is the right lever when your retention is already strong for your category, when you have a genuine time-limited window, or when you are still exploring channels rather than scaling one.

One important nuance: acquisition quality changes your churn rate. A cheap channel that delivers poorly matched users increases c while increasing A, and the net effect on the ceiling can be zero or negative. Always compute carrying capacity per channel, because a channel with half the volume and a third of the churn can be worth more than the one everybody is celebrating.

3. Resurrect dormant users

The most underused lever. Every app accumulates a large population of users who installed, used it for a while, and drifted away. They already know what the product is, they already have an account, and they have demonstrated at least initial intent. Reactivating them is usually far cheaper per active user than acquiring a stranger.

In the model, resurrection acts as an addition to the numerator. If you reactivate 1,500 dormant users per month on top of 9,400 new ones, your effective A becomes 10,900 and your ceiling rises by about 16 percent — from a channel that costs a fraction of paid acquisition.

Carrying Capacity = (New Users + Resurrected Users) / Churn Rate

Adding virality

If your product has word-of-mouth or referral mechanics, part of your acquisition scales with your user base rather than with your budget. Let k be the number of new users each existing active user generates per period. Then:

U(t+1) = U(t) + A + k * U(t) - c * U(t)
Equilibrium:  U = A / (c - k)

Two consequences worth understanding. First, virality effectively reduces your churn rate: a k of 0.02 against a c of 0.07 behaves like a churn rate of 0.05, raising the ceiling by 40 percent. Second, if k ever exceeds c, the denominator goes negative and the model has no finite ceiling — that is the mathematical description of a product that grows without bound. In reality it always saturates as the addressable market fills, but the regime is real and it is what people are describing when they say a product went viral.

Segmented carrying capacity

A blended, company-wide capacity number is directionally useful and strategically misleading. Different segments have different churn rates and different acquisition volumes, so they have different ceilings — and they hit them at different times.

This table tells a clear story that the blended number would have buried. Paid social is at 95 percent of its ceiling — spending more there produces almost nothing, because the users churn out nearly as fast as they arrive. Referral has enormous headroom and the lowest churn, so investing in referral mechanics has a far better expected return. Organic search has both headroom and volume, which makes it the most valuable channel to expand.

Where the model breaks down

Every model is wrong somewhere, and knowing where keeps you from over-trusting it.

• Churn is not constant across tenure. Real retention curves are steep early and flatten later, so a single c overestimates long-run churn and therefore underestimates your true ceiling. For more accuracy, split into a new-user churn rate and a mature-user churn rate and model the two populations separately.
• Market saturation is not in the model. The formula assumes new users are always available at rate A. Every market has a finite addressable population, and as you penetrate it, A falls on its own.
• Seasonality distorts short windows. Fitness apps in January and tax apps in April will produce nonsense capacity figures if you sample the wrong month. Use trailing twelve-month averages when seasonality is strong.
• Product changes reset the parameters. A major release can shift both A and c. Treat capacity as a current-state diagnostic, not a permanent property.
• It says nothing about revenue. A high ceiling composed of users who never pay is not obviously better than a lower ceiling of users who do. Run the model on your paying population as well as your active population.

Using it for planning

Carrying capacity is most valuable as a planning instrument, because it converts strategy arguments into arithmetic ones.

1. 1Sanity-check targets. If the annual plan calls for 500,000 MAU and your current capacity is 130,000, then the plan is not an acquisition plan — it implicitly requires roughly a fourfold change in the ratio of acquisition to churn. Better to discover that in planning than in Q3.
2. 2Compare initiatives on the same axis. Convert every proposed project into its effect on A or c, then compute the resulting ceiling. A retention project that moves churn from 7 to 6 percent and a campaign that adds 1,200 new users per month can now be compared directly.
3. 3Time your investments. Below roughly half of capacity, acquisition is efficient because there is room to fill. Above roughly 80 percent, additional acquisition mostly replaces churned users and retention work is almost always the better investment.
4. 4Set honest expectations. Sharing the capacity math with leadership before growth flattens converts a future crisis into a current strategy discussion.

How this connects to the aha moment

Carrying capacity tells you that churn is the dominant variable. It does not tell you what to do about churn. That is where activation analysis comes in: the largest single block of churn in nearly every consumer app is users who never reached the core value in the first place, and the aha moment is the operational definition of reaching it.

The two models fit together cleanly. Carrying capacity is the diagnosis — it tells you how much headroom you have and which lever matters. The aha moment is the treatment — it tells you exactly what to change to move the lever that matters most. Teams that run both stop arguing about whether to focus on growth or product, because the arithmetic makes it obvious that those were never separate things.