MRR is not cash: walking a $10,000 month down to what lands

Your billing dashboard reports the price your customers agreed to pay. Your bank account reports what survived settlement delay, processing fees, refunds, disputes, failed charges and the eleven months of service you still owe every annual customer. On a stated $10,000 MRR month those wedges can leave under $5,700 you are genuinely free to spend, and none of the difference shows up anywhere on the dashboard.

This is not an accounting curiosity. It is the specific way small software products die while the chart still points up. Bessemer's own cloud metrics guide puts it plainly: committed MRR "gives you a great sense for the revenue health of the business, but can very often be disconnected from the 'cash health' of the business." ASC 606 and IFRS 15 govern how you report revenue. Nothing at all governs MRR. It is a number your billing tool computes for you, and it is not a claim about your bank balance.

Each wedge below carries its cost in days or in percent, and the reconciliation at the end is a monthly procedure. The worked figures are an illustration (a US founder, one product, no team), not observed data.

Settlement is a queue, and the last week of the month sits at the back of it

A card charge that succeeds today is not money today. Stripe's payouts documentation sets the standard schedule at T+2 business days in the US and T+3 business days in the UK and every supported EU country. Bank debits are slower still: ACH Debit settles in 4 business days, SEPA Direct Debit in 6.

Two consequences follow. First, anything billed in the final two or three business days of the month appears in that month's MRR and on next month's bank statement. On a flat book the tail you lose at month end roughly equals the one you inherit at the start of the next. On a growing book the tail is always bigger than the one before, so you permanently carry a few days of your best week in transit.

Second, for a new account the initial payout is typically scheduled 7 to 14 days after your first successful payment. Your launch week's revenue is real and unspendable for a fortnight, and if you priced your first month's hosting against your first month's sales, that gap is where the overdraft happens. The general version of this problem, the days between work done and money landed, is time-to-cash: where freelancers actually lose money, and subscription billing does not exempt you from it.

The fee is one number in the US and three in the UK and EU

The US headline is a single blended rate: 2.9% + $0.30 per successful card charge, per stripe.com/pricing, with international cards adding 1.5% and currency conversion adding 1%.

The UK and EU rates are unbundled, and the unbundling is where the money goes. Stripe's UK pricing is 1.5% + 20p for a domestic card, genuinely cheaper than the US, but a card issued outside the UK is 3.15% + 20p, and if the charge settles in a currency other than your account's, add 2% for conversion. A UK founder selling to US customers in dollars is therefore paying roughly 5.15% + 20p, not 1.5%. EU pricing works identically with €0.25 in place of 20p.

Then Stripe Billing adds 0.7% of billing volume on top of processing, in every region. Subscriptions are Billing volume, so if you run subscriptions you pay it.

In the illustration, $12,890 of successful charges costs $373.81 in percentage fees at 2.9%. The identical book of business, sold by a UK founder to the same US customers, costs $663.84 at 5.15%. That is about $290 a month, roughly $3,480 a year, entirely invisible in MRR. It is the single largest pricing surprise for a non-US founder, and it is one reason a merchant of record at a flat 5% stops looking expensive — the other reason being tax, which is the subject of VAT and sales tax when you sell software.

A refund returns the customer's money and keeps the fee

Stripe's refunds documentation states it directly: "Stripe's processing fees from the original transaction aren't returned." The support article on refunded payments extends the list — "payment processing, Connect, and currency conversion fees from the original charge are not returned."

So a $100 refund on a US account costs you $100 plus the $3.20 you already paid to collect it. At a 5.15% effective rate it costs $100 plus $5.15. Refunds are not neutral, and a generous refund policy has a real, computable price per unit.

Two exceptions are worth knowing. Canceling an uncaptured authorization is free. And a very early refund processed as a reversal, typically within about two hours of capture, carries no withheld fee. Both are narrow. Neither helps with the cancellation that arrives on day nine.

A dispute costs two fees and counts against you whether you win or not

Since 17 June 2025, Stripe's dispute fees are $15 in the US, £20 in the UK, €20 in the EU for the dispute received, plus the same amount again for a dispute countered. The disputes documentation is explicit about what comes back: "Stripe returns the dispute countered fee if you win the dispute. Unless otherwise stated in your Stripe contract, we never return the dispute received fee."

Winning returns half the fee stack. Losing returns none of it and takes the transaction amount as well.

The part almost nobody writes about is the monitoring program sitting behind all this. Visa's consolidated VAMP program has been live since 1 June 2025 (fact sheet, thresholds as rendered by Stripe). Its ratio counts TC40 fraud alerts plus TC15 disputes, divided by settled transactions — fraud alerts and disputes in the same numerator. You are non-compliant at just 5 events and a 0.5% ratio. The Excessive tier sits at 1,500 events and 1.5%, tightened from 220 basis points on 1 April 2026.

The count is unaffected by whether you won, and unaffected by whether you refunded the customer before the dispute landed. Run the arithmetic for a small product: at 500 settled transactions a month, five events is a 1.0% ratio, so the count floor and the ratio threshold trip in the same moment and the ratio is already double what it needs to be. A founder with a few hundred customers and one bad week of card testing can be non-compliant without ever losing a dispute.

An annual plan is cash today and an obligation for eleven months

A customer paying $1,200 on 1 January for twelve months gives you $1,200 of cash and $0 of revenue that day. You hold a $1,200 contract liability and release $100 to revenue each month as you deliver. Under ASC 606, a contract liability is "an entity's obligation to transfer goods or services to a customer for which the entity has received consideration," and revenue is recognized over time where "the customer simultaneously receives and consumes the benefits" (606-10-25-27), which is the definition of a subscription. IFRS 15 applies the same five steps.

The practical consequence is a trap, not a technicality. The month you sell a lot of annual plans is the month you most overestimate what you can spend. Cash spikes, MRR barely moves, and the spike is money you owe as service. If you spend it and then have a churn month, you are funding delivery for customers who already paid, out of revenue from customers you no longer have.

This is also why a runway figure built on your bank balance flatters you. Subtract the unearned portion of every annual plan before you compute months of cover in the runway calculator, or you are counting other people's prepaid service as your own cash.

Involuntary churn is a cash line, not a support problem

Stripe's own engineering write-up on Smart Retries reports that 25% of lapsed subscriptions are purely due to payment failures, and that subscriptions recovered from involuntary churn "continue on average for seven more months." A recovered subscription is not a saved ticket. It is seven months of MRR you would otherwise have written off.

Two operational details decide whether you get it. The default recommended retry policy is 8 attempts within 2 weeks. And hard decline codes are not retryable at all — lost_card, stolen_card, incorrect_number, revocation_of_authorization, authentication_required, transaction_not_allowed and their siblings need a new payment method, not another attempt (revenue recovery docs). Retrying a hard decline is time spent on an outcome that cannot occur. Split your failure list by code before you send a single email.

There is no trustworthy public figure for what share of recurring card payments decline — every version of that statistic traces back to a vendor blog. Measure your own; your billing dashboard already knows it.

Nothing is withheld from a payout, so the tax bill arrives whole

No tax comes out of a Stripe payout. Not income tax, not self-employment tax, not VAT. The full gross lands and every obligation against it is yours to reserve.

For a US founder this creates a specific first-year trap. The 2026 estimated payment dates are 15 April 2026, 15 June 2026, 15 September 2026 and 15 January 2027, and the Form 1040-ES safe harbor requires withholding and estimated payments of at least the smaller of 90% of your 2026 tax or 100% of your 2025 tax — 110% if your 2025 AGI exceeded $150,000 ($75,000 if married filing separately). You owe estimated tax if you expect to owe $1,000 or more after withholding.

Read what that means when your first profitable year is year one. Prior-year tax was near zero, so 100% of it is near zero, so the safe harbor costs you nothing to satisfy and you pay nothing in during the year. The entire bill then arrives on 15 April the following year, in one installment, on money you have already spent. Self-employment tax alone is 15.3% of 92.35% of net profit — about 14.1% of profit as a floor, before any federal or state income tax. Compute your own monthly set-aside with the tax set-aside calculator and move it to a separate account on payout day, not at filing time.

The wedges side by side

The reconciliation to run on the first working day of each month

Copy this, replace the figures with yours, and run it monthly. The numbers shown are an illustration: a US founder, $10,000 stated MRR, 250 successful charges, three annual plans sold this month, $3,000 of monthly operating costs.

MONTHLY MRR-TO-CASH RECONCILIATION — month of ____

1 Stated MRR (dashboard) 10,000.00
2 Less MRR attributable to annual plans -300.00
3 Plus annual plans collected in full this month +3,600.00
= BILLED THIS MONTH 13,300.00

4 Less charges that failed and are in retry -410.00
= SUCCESSFULLY CHARGED 12,890.00

5 Less refunds issued -200.00
6 Less disputed amounts withdrawn -37.00
= CUSTOMER MONEY RETAINED 12,653.00

7 Less processing fees (2.9% + $0.30 x 250) -448.81
8 Less Stripe Billing (0.7% of 12,890) -90.23
9 Less dispute fees (2 received, 2 countered, -45.00
1 won and returned)
= NET SETTLED 12,068.96

10 Less charges still in transit at month end -1,150.00
= CASH THAT ACTUALLY LANDED 10,918.96

11 Less unearned portion of annual plans -3,300.00
12 Less tax reserve (30% of recognized profit) -1,924.79
= UNRESTRICTED CASH FROM THIS MONTH 5,694.17

Stated MRR 10,000.00 -> unrestricted 5,694.17
Gap: 43.1%

Why each line is there:

• Lines 1–3 separate revenue from collection. Annual plans contribute one twelfth to MRR and twelve twelfths to the bank, so both adjustments are needed or you double-count.
• Line 4 is involuntary churn caught at the moment it is still recoverable. Split it by decline code: soft codes go to retries, hard codes go to a "new card please" email.
• Lines 5–6 are money that left. They are separate from line 7 because the fees on refunded and disputed charges stay inside line 7 and never come back.
• Line 7 is computed on successfully charged, not on retained, for exactly that reason.
• Line 9 is the fee arithmetic, not the transaction amounts. Two disputes received at $15 each, both countered at $15 each, one won and one countered fee returned: $30 + $30 − $15 = $45.
• Line 10 is timing, not loss; it arrives within days. Track it anyway, because the difference between the tail you carry out and the one you carried in is permanently absent from your balance.
• Line 11 is service you owe. Spending it is borrowing from your own delivery obligation.
• Line 12 uses this illustration's 30% assumption on recognized profit of $10,000 − $3,000 costs − $584 fees. Your rate is your own; the sourced floor for a US sole proprietor is self-employment tax at 15.3% of 92.35% of net profit.

Line 10 cannot be computed from the billing dashboard alone, which is why the reconciliation only works against the actual bank line. Before reaching for a spend-management product, note the shape of the problem: this is a read-only reconciliation question rather than a card-issuing one, which is the argument on the Ramp alternative page. Worklyn's bank feeds are read-only and available on Pro and above, and they exist to match what landed against what was billed.

What to change this week

1. Run the reconciliation above for July, then for June. Two months gives you your own gap percentage. Use that percentage, not MRR, as the input to any runway or hiring decision — the mechanics of forecasting a thin quarter are in how to plan for a slow quarter before it arrives.
2. Split your failed-payment list by decline code today. Hard codes get an email asking for a new card; soft codes get the retry schedule. Anything you retry that is lost_card or stolen_card is time you have decided to waste.
3. Move the unearned portion of every annual plan and your tax reserve into a separate account on payout day. Not monthly, not at quarter end. On the day the money lands, before you have a chance to look at the balance and feel wealthy.

The wider version of this argument, that shipping stopped being the constraint and everything downstream of it became one, is in the bottleneck moved twice. Worklyn's CFO Mode computes Safe to Spend after deferred revenue and tax reserve, and its 13 Weeks Out forecast runs on cleared bank lines rather than on MRR, which is the only version of the number that can be spent.