TL;DR: A one-person software product is not a small version of a venture-backed company, and the metrics that govern it are not scaled-down versions of the venture metrics either. Three things decide whether it survives. First, the effective take rate of your payment layer, which is not the headline percentage: at Stripe’s published 2.9% plus 30 cents with Billing at 0.7%, a $9 subscription loses 6.93% to fees while a $99 subscription loses 3.90%, and on Paddle’s published 5% plus 50 cents the same two prices lose 10.56% and 5.51%. Fixed per-transaction fees make the cheap product 78% more expensive to collect on through Stripe and 92% more expensive through Paddle. Second, monthly churn, which sets everything downstream: at 5% monthly churn a $29 subscription is worth about $553 net over its life, at 10% it is worth about $277, and the difference is not a 5 point gap, it is half the business. Third, the replacement treadmill: at 10% monthly churn a 200-customer product must acquire 20 new customers every month to stay exactly where it is. Below: a derived take-rate table across five price points, a survival grid across six churn levels, the breakeven arithmetic against published infrastructure list prices, and the reason retention outranks both margin and acquisition cost in the published research.
You can build a working software product in a weekend now. That part genuinely changed, and the people telling you so are not wrong. What did not change is the arithmetic that decides whether the thing is still alive in eighteen months.
The gap between those two facts is where most small software products die. Not at the build stage. At the month-fourteen stage, when the founder finally does the sums and discovers the product has been running a deficit against its own churn since roughly month three.
So this is the sums. Not benchmarks borrowed from companies a thousand times your size, and not a survey of what other founders claim their churn is. Published prices, published research, and arithmetic you can check.
Why MRR is the wrong dashboard
Monthly recurring revenue is a lagging, aggregated, sign-ambiguous number. It tells you the sum of what happened. It does not tell you which of the four things that move it actually moved.
MRR can rise while the business deteriorates. You add fifteen customers, lose twelve, and the twelve you lost were your longest-tenured cohort while the fifteen you added came from a discount campaign. MRR is up. Every underlying quantity is worse.
The four movements underneath are new, expansion, contraction and churned. Only the fourth compounds against you, and only the fourth is close to a constant of the product rather than a constant of your marketing effort. That is why it deserves the top of the dashboard.
There is published evidence for that ranking, and it is unusually specific.
The retention finding that reorders the priority list
Gupta, Lehmann and Ames Stuart published “Valuing Customers” in the Journal of Marketing Research in February 2004. They modelled firm value as the discounted sum of expected future customer earnings and then ran the sensitivity analysis that matters here: what happens to firm value when you improve retention, margin or acquisition cost by one percent each.
Their answer: a 1% improvement in retention improves firm value by about 5%. A 1% improvement in margin improves it by about 1%. A 1% improvement in acquisition cost improves it by about 0.1%. They also found that a 1% improvement in retention had almost five times the impact of a 1% change in the discount rate or cost of capital.
Retention beat margin by a factor of five and beat acquisition cost by a factor of fifty.
That result did not arrive from nowhere. Reichheld and Sasser had made the case fourteen years earlier in “Zero Defections: Quality Comes to Services”, published in Harvard Business Review in September 1990, reporting that cutting the defection rate by 5% raised profits by 85% in one bank’s branch system, by 50% in an insurance brokerage and by 30% in an auto-service chain. The spread across those three industries is the useful part: the direction was consistent, the magnitude was not, which is a fair warning against treating any single retention multiplier as transferable to your product.
That result is usually quoted as a slogan. It is more useful as arithmetic, because the mechanism is not mysterious and you can reproduce it on your own numbers in one line.
Take a product retaining 95% of customers each month, so monthly churn is 5%. Under the standard geometric retention model, expected customer lifetime is one divided by the churn rate, which is 20 months. Now improve retention by one percent in relative terms: 95% becomes 95.95%. Churn does not fall by one percent. It falls from 5.00% to 4.05%, a relative drop of 19%, because churn is the small complement and a small relative move in the large number is a large relative move in the small one. Expected lifetime rises from 20.0 months to 24.7 months.
Our derived reading: a 1% relative improvement in retention produced a 23.5% increase in expected customer lifetime, and therefore in lifetime value at constant margin. That is the leverage Gupta and colleagues measured, expressed at the scale of a single small product.
This is the whole reason retention sits above acquisition on a solo founder’s list. Acquisition is linear in effort. Retention is leveraged.
What the payment layer actually costs you
Before any of the retention arithmetic runs, some fraction of every payment leaves before it reaches you, and almost nobody building a small product computes that fraction correctly.
The headline rate is not the effective rate, because the fixed per-transaction component does not scale. At the time of writing, Stripe publishes 2.9% plus 30 cents per successful domestic card transaction, plus 1.5% for international cards and 1% where currency conversion is required, with Billing on pay-as-you-go at 0.7% of billing volume. Paddle, operating as merchant of record, publishes 5% plus 50 cents per checkout transaction, with tax compliance, registration and fraud protection included in that number and custom pricing required below $10.
Those are the published inputs. Here is what they mean per transaction.
| Monthly price | Stripe fee | Stripe rate | Stripe + Billing | Effective rate | Paddle fee | Effective rate | Paddle premium |
|---|---|---|---|---|---|---|---|
| $9 | $0.561 | 6.23% | $0.624 | 6.93% | $0.950 | 10.56% | $0.33 |
| $19 | $0.851 | 4.48% | $0.984 | 5.18% | $1.450 | 7.63% | $0.47 |
| $29 | $1.141 | 3.93% | $1.344 | 4.63% | $1.950 | 6.72% | $0.61 |
| $49 | $1.721 | 3.51% | $2.064 | 4.21% | $2.950 | 6.02% | $0.89 |
| $99 | $3.171 | 3.20% | $3.864 | 3.90% | $5.450 | 5.51% | $1.59 |
This table is derived by CEOtudent from the published fee schedules of each provider; the fee schedules are theirs, the per-price-point arithmetic is ours. Domestic card assumed, no currency conversion, monthly billing.
Three things fall out of it that change decisions.
Cheap products are structurally more expensive to run. The effective rate on a $9 subscription is 6.93% against 3.90% on a $99 subscription through the same provider, a 78% higher toll for the identical service. The fixed 30 cents is 3.3% of $9 and 0.3% of $99. If you are choosing between a $9 tier and a $19 tier, the fee structure is quietly voting for $19.
Annual billing is a fee decision, not just a cash-flow decision. One annual charge pays the fixed component once instead of twelve times. On a $29 monthly plan the fixed component alone costs $3.60 a year. Charging $290 annually pays 30 cents.
The merchant-of-record premium is a real number you can compare against a real cost. Paddle costs $0.61 more per transaction than Stripe with Billing at the $29 price point, which is $7.27 per customer per year. Whether that is expensive depends entirely on what you would otherwise spend handling VAT registration and remittance across jurisdictions yourself. It is a build-versus-buy decision with a price tag on both sides, and at low volume the tag on the buy side is small. The wider version of that decision is covered in digital product pricing in the AI era.
The survival grid
Now the central table. Fix the price at $29 a month, take the net after Stripe plus Billing fees of $27.656 per customer per month, and vary only the monthly churn rate.
| Monthly churn | Expected lifetime | Net LTV | Max CAC at 3:1 | CAC payback | New customers per month to hold 200 | 12-month retention |
|---|---|---|---|---|---|---|
| 2% | 50.0 months | $1,382.80 | $460.93 | 16.7 months | 4.0 | 78.5% |
| 3% | 33.3 months | $921.87 | $307.29 | 11.1 months | 6.0 | 69.4% |
| 5% | 20.0 months | $553.12 | $184.37 | 6.7 months | 10.0 | 54.0% |
| 7% | 14.3 months | $395.09 | $131.70 | 4.8 months | 14.0 | 41.9% |
| 10% | 10.0 months | $276.56 | $92.19 | 3.3 months | 20.0 | 28.2% |
| 15% | 6.7 months | $184.37 | $61.46 | 2.2 months | 30.0 | 14.2% |
This table is a CEOtudent editorial framework. It combines one verified input, Stripe’s published fee schedule, with the standard geometric retention model from the customer-lifetime-value literature. Expected lifetime is one divided by monthly churn. Net LTV is monthly net revenue divided by monthly churn, with no discounting and no expansion revenue, which makes it a deliberately optimistic ceiling rather than a forecast. The 3:1 LTV to CAC ratio is a conventional planning heuristic, not a law. Read the rows as comparisons, not as promises.
The comparisons are what matter.
Churn does not degrade the business proportionally. Going from 5% to 10% monthly churn is a 5 point move that halves lifetime value, halves affordable acquisition cost, and doubles the acquisition volume required to stand still. Going from 5% to 2% multiplies lifetime value by 2.5. The relationship is a reciprocal, not a line, and human intuition is bad at reciprocals.
The replacement treadmill is the number that ends most small products. At 10% monthly churn with 200 customers you must find 20 new paying customers every month before you grow by one. That is a permanent, non-negotiable marketing job, and it is the job a solo founder is least equipped to hold down for years while also maintaining the software. At 3% churn the same product needs 6. The difference between those two products is not effort. It is that one of them has a retention problem it never diagnosed because it was watching MRR.
Low churn makes long CAC payback survivable, and only low churn does. At 2% churn a 16.7-month payback is fine because the customer is expected to stay 50 months. At 15% churn a 2.2-month payback looks fast and is still marginal, because the customer is gone in under 7 months and you are back at the top of the funnel. Payback period in isolation tells you nothing. Payback period as a fraction of expected lifetime tells you everything, and in this grid it is a constant one third by construction.
Annual retention collapses faster than monthly churn suggests. A 10% monthly churn rate does not mean you keep 90% of a cohort for the year. It means you keep 28.2%. A 5% rate keeps 54.0%. Founders quote monthly churn because it sounds small and then plan as though it were the annual figure.
The breakeven floor is lower than you think, and that is the trap
Infrastructure for a small product is genuinely cheap now, and it is worth naming a real published price rather than gesturing at one. DigitalOcean publishes Basic shared-CPU droplets at $4.00 a month for 1 vCPU with 512 MiB RAM and 10 GiB SSD, $6.00 for 1 GiB RAM, $12.00 for 2 GiB, and $18.00 for a 2 vCPU, 2 GiB configuration with 3,000 GiB of transfer.
So how many $29 subscribers cover the fixed monthly cost?
| Fixed monthly cost | Subscribers needed | Gross MRR at that count | Net after fees |
|---|---|---|---|
| $18 (one published 2 vCPU droplet) | 1 | $29 | $27.66 |
| $50 (small stack plus a tool or two) | 2 | $58 | $55.31 |
| $200 (realistic tooling and services layer) | 8 | $232 | $221.25 |
| $500 (larger footprint, paid support tiers) | 19 | $551 | $525.46 |
This table is a CEOtudent editorial framework. The $18 row is anchored to DigitalOcean’s published Basic droplet list price; the higher rows are illustrative cost bands, not measured figures, and are included to show the shape of the requirement rather than to predict your bill. Subscriber counts are rounded up.
The point of this table is not that infrastructure is cheap. It is that infrastructure being cheap is precisely what makes the failure mode invisible.
If your product costs $200 a month to run and you have 40 customers, you are covering costs eight times over. Nothing hurts. Nothing signals. And if churn is 10%, you are losing 4 customers a month, replacing 3, and the entire deterioration is buried under a comfortable margin on a small absolute number. The infrastructure bill will never be the thing that tells you.
This is the single most important difference between a micro-SaaS and a funded company. A funded company has a burn rate that forces the arithmetic into view every month. A solo product on a $18 droplet can decay for two years without a single financial alarm going off. You have to run the numbers deliberately, because nothing will run them for you.
That is the CEO half of the job. The building half got easy, which is exactly why the managing half became the constraint. The same asymmetry shows up across the one-person business, and it is why recurring revenue models for solo operators differ so sharply in how much management overhead they impose.
The five numbers to actually track
If you replace the MRR dashboard with five numbers, use these.
1. Effective take rate. Total payment and billing fees divided by gross revenue, computed monthly from your actual statements rather than from the headline percentage. If it is drifting up, your mix is shifting toward cheaper plans or more international cards.
2. Monthly logo churn, by cohort. Customers lost divided by customers at the start of the month, tracked by the month they joined. Aggregate churn hides the pattern that actually diagnoses the product: whether people leave in month one because onboarding failed, or in month nine because the value ran out. Those are different problems with different fixes and they look identical in a single blended number.
3. Net revenue retention. Starting MRR of a cohort, plus expansion, minus contraction, minus churn, divided by starting MRR. This is the one number that can exceed 100%, and if it does, the replacement treadmill stops. It is also the only realistic route to a durable one-person product, because it is the only mechanism that does not require you to keep finding strangers forever.
4. CAC payback as a fraction of expected lifetime. Not payback in months. Payback divided by expected lifetime. Under one third is comfortable, over one half is fragile, and the absolute month count on its own is close to meaningless.
5. Months of runway at zero new customers. Take current customers, apply your churn rate forward, and find the month where net revenue falls below fixed costs. This is the number that converts an abstract churn percentage into a date, and a date is the only form in which most people act on it.
What this arithmetic does not tell you
Being clear about the limits is part of the method.
| Question | What the model above can say | What it cannot say |
|---|---|---|
| Will my product survive? | What churn rate it would need to survive at a given cost base | Whether you will achieve that churn rate |
| Is my churn normal? | What each churn level implies mechanically | Nothing. There is no credible published benchmark for one-person software products, and quoted figures for larger SaaS companies do not transfer |
| What should I charge? | How the fee toll and required customer count change with price | What your market will bear |
| Is 3:1 LTV to CAC right? | What that ratio implies for payback | Whether the heuristic applies to you. It is a convention, not a finding |
| Should I use Stripe or Paddle? | The exact per-transaction cost difference at your price | What your tax compliance burden would cost you to handle directly |
This table is a CEOtudent editorial framework, separating what follows from the arithmetic from what would require data this model does not contain.
The honest summary: the survival grid is a constraint map, not a forecast. It tells you which regions of the churn and price space are survivable for a one-person operation and which are not, and it does that reliably because the arithmetic is closed. It cannot tell you where in that space your product sits. Only your own cohort data can, which is why number two on the list above is a measurement instruction rather than a target.
Where to start if you have never done this
Take last month’s payment statement. Compute total fees divided by gross revenue and compare it to the headline rate you assumed. That is the first surprise.
Then take the customers who were active on the first day of a month twelve months ago and count how many are still paying. Divide, take the twelfth root, subtract from one. That is your real monthly churn, measured rather than estimated, and it is almost always worse than the number founders quote.
Then find your row in the survival grid.
Most of the decisions that follow, whether to raise prices, whether to move to annual billing, whether to spend the next quarter on acquisition or on the retention problem you just found, resolve themselves once you know which row you are in. That is what a constraint map is for. It does not make the decision. It removes the ones that were never available.
For the layers around this: micro-SaaS as a category covers what these products are and how they get built, the AI wrapper market analysis examines whether the category you are entering has defensible margins at all, the cold start problem handles the acquisition side before you have an audience, and the one-product business argues the case for concentrating rather than stacking offers.
Frequently asked questions
What is a good churn rate for a micro-SaaS?
There is no credible published benchmark for one-person software products, and figures quoted for larger SaaS companies do not transfer because the customer mix, contract length and support model are all different. The useful framing is not comparative but mechanical: at 10% monthly churn a 200-customer product must replace 20 customers a month to stand still, at 3% it must replace 6. Ask whether you can sustain your replacement rate indefinitely, not whether your churn beats someone else’s.
How do I calculate LTV for a subscription product?
Under the standard geometric retention model, expected customer lifetime is one divided by the monthly churn rate, and lifetime value is monthly net revenue multiplied by that figure. Use revenue net of payment fees, not gross. Note that this version excludes discounting and expansion revenue, which makes it an optimistic ceiling. Treat it as a comparison tool across scenarios rather than as a valuation.
Is Stripe or Paddle cheaper for a small software product?
On published rates, Stripe with Billing costs less per transaction at every price point examined: $1.344 against $1.950 on a $29 subscription, a difference of $0.61, or $7.27 per customer per year. Paddle’s higher rate includes acting as merchant of record, which means it handles global tax registration and remittance. The comparison is therefore between a known fee premium and an unknown compliance cost, and which wins depends on how many jurisdictions you sell into.
Why does everyone say retention matters more than acquisition?
Because it is leveraged rather than linear. Gupta, Lehmann and Ames Stuart found a 1% improvement in retention improved firm value roughly five times as much as a 1% improvement in margin and roughly fifty times as much as a 1% improvement in acquisition cost. The mechanism is that churn is the small complement of retention, so a small relative move in retention is a large relative move in churn. Improving retention from 95% to 95.95% cuts churn from 5.00% to 4.05% and raises expected lifetime by 23.5%.
Should I charge monthly or annually?
Annual billing pays the fixed per-transaction fee once instead of twelve times, which on a $29 plan saves $3.30 a year in fees alone, and it converts twelve churn decisions into one. The cost is a lower conversion rate at checkout and cash you may have to refund. For a product with monthly churn above about 5%, the churn reduction usually dominates the conversion loss.
How many customers do I need to break even?
Fewer than you expect, and that is the danger. At $29 a month net of Stripe fees, a single subscriber covers DigitalOcean’s published $18 droplet, and 8 cover a $200 monthly cost base. Because the breakeven point is so low, a small product can decay for years without any financial signal. Breakeven is not the milestone that matters. Net revenue retention is.
Sources
- Gupta, Lehmann and Ames Stuart, Valuing Customers, Journal of Marketing Research, volume 41, issue 1, February 2004, pages 7 to 18
- Stripe, published pricing documentation, fee schedule for domestic and international card payments and Billing, accessed August 2026
- Paddle, published pricing documentation, merchant-of-record checkout transaction fee schedule, accessed August 2026
- DigitalOcean, published pricing documentation, Basic shared-CPU droplet list prices, accessed August 2026
- Reichheld and Sasser, Zero Defections: Quality Comes to Services, Harvard Business Review, September to October 1990 issue
This content was compiled with the support of AI following in-depth research, then written and prepared for publication by the CEOtudent editorial team.
This post is also available in:
