Home / Blog / The Leaky Bucket
Day 13The Leaky Bucket: Churn, Cohorts and the Growth Ceiling You Cannot Out-Market
Your churn rate has already decided how big your company gets. Acquisition only decides how long it takes to get there.
There is a number in your business that sets a hard maximum on your size, and most founders have never calculated it. It is not your market size or your ad budget. It is new customers per month divided by monthly churn rate, and it produces a ceiling you cannot spend your way past.
Add 100 customers a month, lose 5 percent of your base each month, and you stop at 2,000 customers. Not because demand ran out. At 2,000 customers, 5 percent is exactly 100, so every new customer replaces one walking out the back. You can double your sales team and reach the ceiling faster. You cannot raise it. Only retention raises it.
Acquisition sets your speed. Retention sets your destination.
This is why "we just need more leads" is the most expensive sentence in a stalled company. More leads at 5 percent churn buys you a faster approach to the same wall. Halving churn to 2.5 percent doubles the ceiling to 4,000 customers without a single extra sale, and it does it permanently.
Fred Reichheld's work at Bain established the first number: raising retention by 5 percentage points raised profits by 25 to 95 percent across the industries they studied. The second is the figure Harvard Business Review popularised from the same body of research, that winning a new customer costs five to twenty five times what it costs to keep one you have. Neither is a marketing slogan. Both fall out of the arithmetic above.
Find your ceiling before you fund another campaign
Pull your real numbers. New customers per month is the average of your last three months. Monthly churn is customers lost divided by customers at the start of the month.
Under 1 percent monthly. Enterprise-grade retention. Spend on acquisition with confidence. 1 to 3 percent. Healthy. Your ceiling is high enough that growth is a distribution question. 3 to 6 percent. Typical for small-business software and survivable, but retention work now beats any ad budget. Over 6 percent monthly. Your customers last under 17 months and the bucket empties faster than you can fill it. Stop acquiring, start interviewing the leavers.
Ceiling equals monthly new customers divided by monthly churn. Average lifetime is one divided by churn. At 100 new customers a month, 5 percent churn and 120 a month in revenue, you top out at 2,000 customers and 2.88 million a year, with customers lasting an average of 20 months. Drag churn to 2.5 percent and both numbers double.
Drag the churn slider and watch the revenue ceiling move. Then drag the acquisition slider and watch it move by the same proportion but at much greater cost. One of those levers you buy with money every single month. The other you buy once, with product and service work, and it keeps paying.
Blended churn hides the truth. Cohorts show it.
A single company-wide churn number is an average of every customer you have ever had, which means it is dominated by your past. If you fixed onboarding in March, the March cohort might retain twice as well as the December one, and your blended number will take a year to notice.
Cohort analysis fixes this. Group customers by the month they joined. Track what percentage of each group is still active at month one, month two, month three. Read down the columns to compare cohorts and across the rows to see the shape of decay.
| Cohort | M1 | M2 | M3 | M4 | M5 | M6 |
|---|---|---|---|---|---|---|
| January | 74% | 58% | 47% | 41% | 38% | 36% |
| February | 76% | 60% | 50% | 45% | 42% | 41% |
| March · onboarding fixed | 88% | 79% | 72% | 69% | 68% | 68% |
| April | 89% | 80% | 74% | 71% | 70% | 70% |
Illustrative cohorts. January and February decay toward zero with no flattening, which is a rental. March and April flatten near 68 to 70 percent by month five, which is a base. The blended churn number for this company would still look terrible, because it is dominated by pre-March customers who are still leaving. Cohorts show the fix worked ten months before the average would.
Two things to read in any cohort table. First, does the curve flatten? A curve that flattens means you found a group for whom the product became part of the routine. A curve that keeps sliding means every customer eventually leaves and you are running a rental business with subscription pricing. Second, are newer cohorts above older ones? If yes, your work is compounding. If no, whatever you shipped this year did not touch the reason people leave.
Revenue churn is a different animal
Logo churn counts customers. Revenue churn counts dollars, and the two can point in opposite directions. Lose 5 percent of customers who each pay 20 a month while your 2,000-a-month accounts all upgrade, and you lost logos while gaining revenue.
Net revenue retention captures this. Take a cohort's revenue twelve months on, including upgrades, add-ons and price increases, and divide by what they paid at the start. Above 100 percent means the existing base grows faster than it leaks, so revenue compounds with zero new sales. That single property is what lets a company survive a bad quarter of acquisition without shrinking, and it is why investors weight it heavily.
The math is worth sitting with. A company at 120 percent net revenue retention that sells nothing new for a year still grows 20 percent. A company at 85 percent has to win 15 percent of its revenue in new sales just to stand still, which means its sales team is on a treadmill and its founder is confused about why the effort never shows up.
Three companies each add new revenue equal to 20 percent of their base in a year. The one leaking 15 percent nets 5 percent growth. The one expanding 20 percent nets 40 percent. Identical sales teams, eight times the outcome, decided entirely by what happens after the sale.
Where churn actually comes from
Founders treat churn as a product quality problem. It usually is not. Break your leavers into four buckets and the fixes stop being vague.
Never activated
They signed up, never reached the moment where the product does its job, and left within 60 days. This is the biggest bucket in most early companies, and it is an onboarding problem, not a feature problem. Day 14 covers the fix in detail.
Wrong fit from the start
Sales sold to someone outside the profile. They were always going to leave. Fix this at the top of the funnel, in qualification, not with a save offer at the end.
Value faded
They used it, then usage dropped as a project ended or a champion changed roles. This is where usage-based health scoring earns its keep: a customer whose logins halved is telling you three months before the cancellation email.
Involuntary
Their card expired. Nobody decided anything. Card failures are a meaningful slice of subscription churn and are fixed with dunning emails, card updater services and retry logic, which is a week of engineering that pays back permanently.
Export every customer who cancelled in the last 90 days. Sort by revenue. Call the top twenty and ask one question: "What was happening in your business the week you decided to stop?" Not "why did you cancel," which produces the polite lie ("budget"). Ask about the week, and you get the trigger event. Tally the answers into the four buckets above. The largest bucket is your entire retention roadmap for the quarter, and it is almost never the feature request list you were about to build.
The retention fixes that actually move the number
Ordered by effect per hour spent, based on what consistently shows up in cohort data.
Fix involuntary churn first. It is pure engineering, has no product debate attached, and recovers revenue from customers who never wanted to leave. Smart retry logic and card-update flows are a known, bounded piece of work.
Move the activation moment earlier. If customers who complete a specific action in week one retain at twice the rate of those who do not, your job is to get more people to that action, faster. In a subscription business this is the product work with the highest return per engineering hour, and Day 14 takes it apart step by step.
Bill annually where it fits. An annual contract removes eleven monthly decisions to leave and improves cash conversion at the same time. Offer two months free and watch how many take it. You are buying retention and working capital with the same discount.
Add expansion paths. Seats, usage tiers, modules. Net revenue retention above 100 percent is built here, and it is usually easier than reducing gross churn further.
Then, and only then, spend more on acquisition. Filling a bucket you have not patched is how a company burns a funding round and ends the year the same size it started.
The two churn numbers a board will ask for
When someone experienced looks at your business, they ask for gross churn and net revenue retention, and they ask for both by segment. Gross churn tells them how leaky the bucket is. Net revenue retention tells them whether the water left in it is rising. Reporting one without the other is how founders accidentally hide their best news or their worst.
Segment matters because a single company usually contains two businesses. Your smallest plan might churn at 8 percent monthly while your top tier churns at 1 percent. Blended, that reads as a mediocre 4 percent and prompts a vague conversation about product quality. Split out, it says something specific: the small plan is either a marketing cost that feeds the top tier or a distraction that consumes support hours, and you now have to decide which. Founders who never split the number spend years optimising an average that describes no actual customer.
One more discipline. Report churn on a consistent definition and never quietly change it. Counting a downgrade as churn, or excluding customers who paused, or measuring against start-of-month rather than average base, all shift the number by a point or two. Pick a definition, write it down next to the metric, and keep it through the bad quarters. A churn number you can trust across two years is worth more than a flattering one you cannot compare to anything.
The takeaway
- Your ceiling is new customers per month divided by monthly churn. Calculate it today.
- Blended churn hides direction. Build a cohort table and watch whether the curve flattens.
- Net revenue retention above 100 percent means the base grows without new sales. Below it, your sales team is running a treadmill.
- Split leavers into never activated, wrong fit, value faded and involuntary. Each has a different fix.
- A 5 point retention gain moved profit 25 to 95 percent in Bain's research. No acquisition channel does that.
Frequently asked questions
How do I calculate my growth ceiling from churn?
Divide monthly new customers by monthly churn as a decimal. 100 new at 5 percent churn gives 100 / 0.05 = 2,000 customers. At that size losses equal gains and growth stops even though acquisition never slowed. Halving churn to 2.5 percent doubles the ceiling without one extra sale.
What is a cohort analysis?
Grouping customers by join month and tracking what share of each group is still active at month one, two, three and so on. It reveals direction that a blended average hides, and the key question it answers is whether your retention curve flattens or slides to zero.
What is a healthy churn rate?
It depends on the buyer. Small-business software commonly runs 3 to 5 percent monthly logo churn; enterprise is typically under 1 percent. The better test is whether a cohort stabilises. Flattening at 60 percent is a base. Continuing to slide is a rental.
What is net revenue retention?
Revenue from an existing cohort a year later, including expansion, divided by what they paid at the start. Above 100 percent means revenue compounds with zero new sales. It is what separates companies that survive a bad quarter from companies that shrink in one.
Nobody dies of churn. They die of ignoring it for four quarters.
Kill My Startup shows how retention problems disguise themselves as growth problems until the funding runs out.
Buy on Amazon →Sources
- Frederick Reichheld, Bain & Company, research linking a 5 point retention increase to 25 to 95 percent profit gains.
- Harvard Business Review, "The Value of Keeping the Right Customers," on the 5 to 25 times cost of acquisition versus retention.
- David Skok, "SaaS Metrics 2.0," on churn benchmarks, cohort curves and negative churn.
- Standard cohort analysis method and the steady-state formula: base size equals acquisition rate divided by churn rate.