Recurring revenue with bad retention is worse than one-time revenue. That sentence offends people, so let me defend it with arithmetic: a subscription that churns 8-10% a month is not a recurring-revenue business, it is a business that must re-acquire its entire customer base every year and pay the acquisition cost each time. The word "recurring" describes the billing mechanism. It says nothing about whether the revenue actually recurs — and the gap between those two things is where most subscription businesses quietly die.
Founders worship the word. Investors reward the word. A one-time sale gets a lower multiple than a subscription of identical size, on the theory that recurring revenue is worth more because it comes back. It only comes back if the customer stays, and "the customer stays" is a number, not an adjective. The entire premium of the subscription model is a bet on that number. Almost nobody prices the bet before making it.
The one identity that runs the whole model
Start with the fact that decides everything and that most operators can recite but few actually use: average customer lifetime is roughly 1 divided by your churn rate.
If 5% of customers leave each month, the average customer stays about 20 months. At 10%, about 10 months. At 2%, about 50 months. The identity is exact enough for decisions — if you lose a constant fraction each period, the expected number of periods a customer survives is the reciprocal of that fraction. Halve your churn and you double how long every customer stays, automatically, with no change to your product, price, or funnel.
Now chain it to lifetime value. LTV is roughly ARPU times gross margin, divided by churn:
$$\text{LTV} \approx \frac{\text{ARPU} \times \text{gross margin}}{\text{churn}}$$
Churn sits in the denominator. That is the whole story. Everything you can do to LTV through pricing and margin is linear — raise price 10%, get roughly 10% more LTV. Churn is not linear, because it also sets the lifetime you are multiplying by. Cut churn in half and LTV doubles. Cut it to a fifth and LTV quintuples. No other lever in the business has that shape.
Put real numbers on it. Say ARPU is $100/month and gross margin is 80%, so each retained customer throws off $80 of margin a month.
| Monthly churn | Avg lifetime | LTV (ARPU × GM / churn) |
|---|---|---|
| 10% | 10 months | $800 |
| 5% | 20 months | $1,600 |
| 2% | 50 months | $4,000 |
| 1% | 100 months | $8,000 |
Same product. Same price. Same margin. The only thing that moved is retention, and LTV swung 10x across the range. A founder who fights for a 10% price increase while leaving churn stuck at 8% is polishing the linear term and ignoring the one that compounds.
The treadmill, in customers per month
Watch what churn does to the top of the funnel, because this is the part the word "recurring" hides.
You have 1,000 customers and 10% monthly churn. You lose 100 customers every month. To stay flat — not grow, just not shrink — you must acquire 100 new customers every month, 1,200 a year against a base of 1,000. You are replacing 120% of your customer base annually just to run in place. Compounded, 90% monthly retention means a cohort decays to about 28% survivors after twelve months (0.9 to the 12th). Roughly seven of every ten customers you start the year with are gone by December, and you paid CAC to replace all of them.
That is not recurring revenue. That is a bucket with a hole in it, and your growth budget is the water you pour in to keep the level from dropping. The "recurring" label is on the invoice; the behavior is re-acquisition. A one-time-sale business at least knows it has to sell again. The churny subscription tells itself the revenue is annuity-like while spending like a business that resells its whole book every year — and books the resale as "growth."
Where CAC eats the margin
Retention only earns you the premium if the customer stays long enough to pay back what you spent to acquire them and then some. So put acquisition cost against the table above.
CAC payback, in months, is CAC divided by monthly gross margin per customer. At $80 of margin a month, a $600 CAC pays back in 7.5 months. Hold that fixed and compare it to the lifetimes:
- 10% churn, 10-month lifetime. Payback at 7.5 months. The customer leaves 2.5 months after breaking even. You net roughly $200 of margin per customer against $600 spent to get them — an LTV/CAC of about 1.3. You are running a business to move money from your acquisition channel to your cost of goods.
- 5% churn, 20-month lifetime. Same 7.5-month payback, but now 12.5 profitable months after it. LTV/CAC around 2.7. The same CAC and the same price is suddenly a real business — solely because the customer stayed twice as long.
- 2% churn, 50-month lifetime. Payback is a rounding error against the lifetime. LTV/CAC above 6. You could double CAC to win better customers and still print money.
Nothing changed except the denominator. The churny model does not have a marketing problem or a pricing problem you can spend your way out of — it has a retention problem that makes every dollar of CAC a worse investment, because the asset you bought evaporates before it pays. On that treadmill, CAC does not eat some of the margin. Past a certain churn it eats all of it, and the faster you grow the faster you lose money, because growth is just buying more short-lived customers at a loss.
A "one-time" sale can beat a churny subscription
Here is the frame that should reorganize how you think about the multiple. A one-time product with durable, organic re-purchase can be worth more than a subscription that churns, because the re-purchase carries no CAC.
Think about what actually happens when someone rebuys a consumable they love — a coffee they reorder, a tool they replace, a supplier a merchant keeps calling because it works. Revenue recurs, but you did not pay to make it recur. Demand is durable; the customer comes back on their own. Compare that to a subscription where the billing recurs automatically but you spend on retention, incentives, and win-back to stop the base from draining. Which revenue is worth more — the kind that comes back because the customer wants it, or the kind that comes back on a credit-card timer until they notice and cancel?
I watch this at Kommerce, where the market is cash-on-delivery and trust is scarce. There is no auto-renew to hide behind — a merchant reorders because the last order arrived, the cash reconciled, and the thing worked. Every repeat is a fresh, voluntary decision to trust us again. That is harder than an auto-charge. It is also honest: the revenue recurs only when the value recurred, so I can never mistake a billing cycle for a satisfied customer. A lot of SaaS dashboards let you make exactly that mistake for a year, right up to the renewal cliff.
Subscriptions are not the problem. "Recurring" is simply not the source of the value. Durable demand is. A subscription is one way to package durable demand; when the demand is not durable, the subscription is a financing trick that front-loads revenue recognition and back-loads the reckoning.
Which customers you retain is a pricing decision
Churn is not a single number that happens to your business uniformly. It is a blend, and the mix is set upstream — by price. Your price decides who walks in the door, and the cheapest customers arrive with the shallowest version of the problem, which is to say the version they abandon the moment it stops being nearly free. Segment your churn by the price and discount at which each customer was acquired and the pattern is almost always the same: the worst-retaining cohort clusters at your lowest realized prices. You did not catch a churn problem. You recruited one.
Discounting is the fastest way to recruit it. A discount does not just cost you margin on this sale; it imports the price-sensitive customers who churn hardest, because you selected for the people who only showed up for the price. Then you fold their churn into your blended number, watch LTV sag, and go looking for the fix in your onboarding flow. The onboarding was never the problem. The casting call was.
What to actually do
Stop grading the model on the word. Grade it on two numbers.
First, compute your real monthly logo churn — customers lost over customers at the start of the period, honestly, no dead accounts scrubbed from the denominator — and convert it to average lifetime with 1/churn. Say the number out loud. "At 9% monthly churn my average customer stays 11 months" lands very differently than "we have recurring revenue," and it is the same fact.
Second, put that lifetime against CAC payback. If payback is 7 months and lifetime is 11, you have four months of profit per customer and no room for error. If payback exceeds lifetime, every sale loses money and scaling accelerates the loss. Judge the business on net revenue retention and payback, never on the presence of a recurring invoice.
Then cut. Segment churn by acquisition price and discount depth, find the cohort dragging the blended number, and either reprice it until it retains or stop acquiring it. Killing your worst-retaining segment can raise LTV and lower blended CAC in the same move, because you stop paying to refill the fastest-draining part of the bucket. That is the rare decision that improves two numbers at once.
Recurring revenue is a bet that retention beats churn-adjusted CAC payback. Run the arithmetic before you place the bet, because the market runs it for you at the renewal, and it does not grade on the word.