Revenue-Based Finance Prices Your Volatility, Not Your Failure

A shop with a card machine or a mobile-money till can raise capital in an afternoon now. A lender reads the last twelve months of turnover off the payment rail, wires a lump sum, and takes a slice of every day’s sales until a fixed total is repaid.⁶ No board seat. No fixed monthly instalment.⁵ On paper it is the friendliest money in the market, because when sales dip, the payment dips with them.

That friendliness hides the real test. A revenue-share facility forgives a bad month. It forgives a bad year. It even forgives a permanent decline, because a smaller share of a smaller number is still a payment it can live with. The lender is not pricing whether you will fail. It is pricing whether your revenue will hold still.

So the decision this piece decides is narrow and it is about shape, not size: can your collection cycle carry a claim that never stops taking, or is your revenue lumpy enough that the same instrument that helps a smooth business quietly strangles yours?

Why these three models, and not one lens

This runs the Wire Model. Score the decision, route it to a small ensemble of formal models, then force the ensemble to produce dated actions. Three different questions sit inside “can my revenue carry this,” and one lens answers none of them alone.

Why a lender offers you a worse rate than the shop next door with the same annual turnover is a pricing question about a signal the lender cannot fully see, so the first model is statistical discrimination: you are priced off the estimated variance of the group you look like. Why a lumpy revenue stream makes the repayment stretch and then snap is a congestion question, so the second model treats the outstanding balance as a queue whose clearance time blows up with variance. What actually happens to businesses that take revenue share, sorted by revenue shape, is a frequency question, so the third model is base rates. On the outcome-type map those are equilibrium, complex and cycle-regime. Three distinct types, so their errors do not all lean the same way.

Two layers are folded in rather than shipped as separate cards, per the three-model cap. The behavioural pull, reading a low daily deduction as low cost, lives inside the cards and the blind spot. Governance is one line in the levers: one owner, one variance number, refreshed monthly.

The framework: three readings of a share of your revenue

1. The price: you are charged for the variance the lender cannot see

A revenue lender wants one thing from your business, and it is not growth. It is the reliability of the next hundred deductions. It cannot observe that directly, so it estimates it from what it can observe: your sector, your card and mobile-money history, the shape of your last year. Then it prices you off the group that history resembles.

This is statistical discrimination, and the everyday example is exact. A December-heavy retailer and a steady grocer can post identical annual turnover, and the lender still hands them different terms, because it reads seasonality as risk to the repayment schedule and charges for it. The charge shows up as a higher factor rate, a bigger daily share, a shorter cap, or a minimum payment clause. Merchant cash advance factor rates run from about 1.1 to 1.5, and the holdback taken from daily card sales typically sits between 5 and 20 percent, with the exact figures varying by your industry, your financials and your sales volume.³ Those inputs are a proxy for one thing. Your variance sets your price.

The lever falls straight out of the model. If the lender is pricing the variance it cannot see, your job is to make your true collection cycle legible, so you are priced on your own data instead of your segment’s reputation. A merchant who lets the lender read the payment rail directly is doing exactly this: the CGAP models where fintechs report non-performing loans as low as 3 percent are the ones with direct visibility into digital sales and automatic deduction at source.¹ Visibility is not surveillance you tolerate. It is the discount you earn.

Statistical discrimination, the pricing lens.

Assumes: the lender cannot see your forward variance and prices you off the estimated variance of your observable group.

Fits because: factor rate, holdback and minimum-payment terms are set from your industry, financials and sales volume, all proxies for revenue volatility.

Breaks when: the lender has direct, verified sight of your actual cash flows and can price you individually.

Counteracts: assuming equal annual turnover buys equal terms.

May reinforce: hiding volatility, which makes the lender assume the worst of your group.

2. The queue: a lumpy river lengthens the line, then breaks it

Think of the outstanding advance as a backlog and your revenue share as the server draining it. Each month, the holdback clears part of the balance. How long the balance takes to reach zero is a waiting-time problem, and waiting time is not governed by your average month. It is governed by the variance around it.

The Pollaczek-Khinchine result makes this precise. In a single-server queue, expected waiting time is proportional to the second moment of the service process over one minus utilisation.⁷ Two forces sit in that expression. The second moment carries the variance, so a lumpy revenue stream lengthens the queue even when the average is healthy. And the one-minus-utilisation term in the denominator means that as the committed share approaches the fraction of revenue you can actually spare, the wait does not rise gently. It runs away.

A pure loss does not do this. If sales step down and stay down, a percentage share simply takes a smaller amount for longer, and the payment moves with the business rather than against it. The Kenyan merchants on Kopo Kopo’s Grow product describe the relief directly: they repay as a percentage of the transactions they receive, so they are “not on the hook for set weekly or monthly payments, which is especially helpful when business is slow.”² That is the instrument absorbing a loss, and it is genuinely kind.

Variance is the different animal. A seasonal business does not decline. It swings. It earns in the December rush or the harvest window and empties out for the months between, and it is in those empty months that the queue turns dangerous. Because the lender priced your variance in step one, your facility is the one most likely to carry a minimum payment or a fixed daily debit that runs “regardless of how much you earn in sales.”⁴ Some turnover-assessed lenders make the floor the whole structure: Lula qualifies a South African business on annual turnover but then collects over a flat three-to-twelve-month term, a fixed schedule that does not flex with a slow month at all.⁹ Now the trough does not just lengthen the line. A fixed floor drawn against near-zero revenue is a solvency event, arriving at the one moment your buffer is thinnest. A percentage of a lumpy river is not a repayment plan. It is a queue that lengthens every time the river drops, until a fixed floor snaps it.

Queueing and congestion, the waiting lens.

Assumes: the advance is a backlog and your revenue share is the server, with a floor beneath it.

Fits because: clearance time scales with the variance of revenue, not the mean, and explodes as the committed share nears the cash you can spare.

Breaks when: the share is uncapped, purely proportional and floorless, so a trough only stretches the term and never forces a payment.

Counteracts: judging affordability off an average month.

May reinforce: stacking a second advance to cover a trough, which raises utilisation toward the runaway zone.

3. The base rate: which class of borrower are you joining

Match on structure, not on sector. The reference class is not “African SMEs” or “fintech-funded shops.” It is businesses that pledged a share of revenue against a fixed repayment, split by the shape of their collections.

One class does well. Merchants with steady digital sales and deduction at source, the Kopo Kopo and PayPal Working Capital pattern, sit at non-performing rates around 3 percent, because the model was built for exactly their cash shape.¹ The other class is the merchant cash advance horror file: effective annual costs that the published range puts anywhere from 40 percent to 350 percent, on advances repaid in three to eighteen months.³ The APR is not a scandal by itself. It is a symptom. That range is what variance costs, and the businesses at the top of it are the ones whose revenue could not keep the queue short.

So the base rate for you is not the market average. It is the average of your shape. A smooth-collections business reads the low-default class as its own future. A seasonal or lumpy one that treats the friendly-sounding average as its base rate has picked the wrong reference class, and the mispricing lands entirely on the founder.

Base rates and reference class, the frequency lens.

Assumes: the right comparison group is revenue-share borrowers sorted by collection shape, not by industry.

Fits because: outcomes split cleanly on variance, with smooth digital collections near 3 percent default and lumpy revenue at the costly tail.

Breaks when: a new structure, such as an uncapped floorless share, changes the payoff and voids the old frequencies.

Counteracts: reading the market-wide average as your personal odds.

May reinforce: anchoring on the one peer who survived it and ignoring the shape difference.

What to check before you sign, cheapest move first

Turn the free, reversible dials before the expensive, permanent ones.

1. Compute one number: your revenue coefficient of variation. Take twelve months of collections, divide the standard deviation by the mean. This is an afternoon in a spreadsheet and it is the whole decision. Under roughly 0.3 is smooth. Above 0.5 you are the seasonal case the instrument is built to punish.
2. Find the floor in the term sheet. Read for a minimum daily or monthly payment, a fixed-debit clause, a maximum term, or a “true-up.” A purely proportional share with no floor absorbs your bad months. Any floor converts your worst month into a fixed bill. The floor is the single line that decides survivability.
3. Make your cycle legible to earn the good price. Offer the lender direct read access to your payment rail and a clean twelve-month collections history. You are converting your variance from an assumption into a fact, and the discount is real.
4. Convert the cost to an annualised rate yourself. Take the factor or cap, divide by your realistic repayment months, annualise it. Compare that number to term debt and to the dilution the same cash would cost in equity. Do not let “a small daily percentage” stand in for a price.
5. Own the variance number. One person, one figure, refreshed monthly, sitting next to runway. That is the governance layer, and it is one line.

What to settle before the facility goes live

DO NOW, by T+3. Reversible, and correct whatever you decide.

1. Compute the coefficient of variation on the last twelve months of collections. Put it at the top of the funding memo.
2. Highlight every floor, minimum payment and maximum-term clause in the draft agreement. If there is a floor, model the worst trough month against it.
3. Annualise the true cost and set it beside your term-debt and equity alternatives on one page.

HEDGE, by T+14. Cheap insurance against a cycle you cannot perfectly forecast.

1. Negotiate the floor out, or cap the term, before you negotiate the amount. Trade a smaller advance for the removal of a fixed-payment clause. That trade is almost always worth it for a lumpy business.
2. Grant payment-rail visibility to move your price off your segment and onto your data.
3. Size the advance so the committed share stays well below the cash you can spare in your thinnest month, not your average one.

DEFER AND TRIGGER. Taking revenue share against seasonal revenue is the move that can trap you, so pre-commit the signal instead of deciding in the moment.

1. Trigger to take the facility: coefficient of variation under 0.3, or a term with no floor of any kind. Then revenue share is cheap, forgiving capital and you should use it.
2. Counter-trigger, by T+28: if your variation is above 0.5 and the lender will not remove the floor, stop pursuing revenue share for working capital. Route to an overdraft, invoice finance, or a customer prepayment instead, and revisit only after two quarters of smoother collections.

From the lender’s side of the table. If you are the one pricing revenue share, DO NOW: underwrite on the variance of collections, not the level, because the level tells you almost nothing about repayment risk. HEDGE: prefer direct-deduction, floorless structures for volatile merchants, which lower your default rate more than a minimum payment protects it. DEFER: before you attach a fixed floor to a seasonal borrower, price the correlation you are creating, because your floor bites hardest in the exact month their revenue is lowest, and a default you manufactured is still a loss you carry.

What usually happens to a business that pledges its revenue

The base rate matched on shape is the honest forecast, and it is not the market’s advertised average. Revenue share against smooth, digitally-collected sales performs well, which is why the model spread across mobile money at all: credit is now the most popular adjacent service offered by mobile-money providers, growing faster than savings or insurance.⁸ Where the collection rail is steady and deduction is automatic, defaults sit low and the instrument does its job.

Present-state modifiers push the seasonal borrower the wrong way. The same rails that make the good version cheap also make the aggressive version easy to stack. A trough tempts a second advance to cover the first, utilisation climbs toward the runaway zone in the queueing model, and the effective cost marches into that 40-to-350 percent band.³ The instrument did not change. The revenue shape underneath it did the damage.

Subtract the counterfactual before you blame the facility for everything. Some of a seasonal business’s cash pain is just seasonality, and would exist with no lender at all. Charge revenue share only for the incremental harm: the fixed floor drawn in the empty months, and the price premium you paid for variance you could have made legible. That is the honest number, and for a lumpy business it is still large enough to decide the question.

Matrix-break flag. Real-time, transaction-linked lending is starting to dissolve the floor entirely. When a lender reads your till live and deducts a pure percentage per transaction with no minimum, the instrument stops pricing variance and starts only pricing level, and the seasonal penalty in this whole analysis softens toward zero. Watch for it. Until your term sheet is genuinely floorless, though, you are in the world the three models describe, and the variance test stands.

What these three models cannot price

Three things sit outside the ensemble, and none of them lets you skip the variance number.

A demand shift disguised as variance. The models read your last twelve months as the distribution of your future. A structural change, a new competitor, a regulation, a lost anchor client, is not noise around a stable mean. It is a new mean, and no coefficient of variation computed on the old data will warn you.

The lender’s own funding cycle. Revenue lenders borrow too. When their capital tightens, floors and minimum payments appear in terms that were flexible a quarter earlier, for reasons that have nothing to do with your business. The pricing model assumes a stable counterparty. It is not always one.

The behavioural trap under a small daily number. A 12 percent holdback feels like nothing next to a loan repayment, and that feeling is the danger. It hides the annualised cost and it hides the floor. The models can price the facility. They cannot stop a tired founder from signing on the daily number instead of the annual one.

None of that changes the first move. This week, compute the coefficient of variation on your last twelve months of collections, and read the draft agreement for a floor. If your revenue is smooth and the share is floorless, take the money. If your revenue swings and the floor will not come out, walk, and fix the collection cycle before you rent against it. Match the instrument to the shape of your revenue, not the size of it.

Sources and notes

1. CGAP, “Digital Credit Models for Small Businesses.” Digital merchant cash advance lets “businesses and lenders dynamically set up repayment schedules based on the volume of sales as opposed to the traditional predefined monthly schedule,” and “fintechs using this model reported nonperforming loans ratios as low as 3 percent.” Named players include PayPal Working Capital, Kopo-Kopo Grow Loan and Amazon Lending. Some models “use direct payment integrations to gain visibility into the MSE’s digital sales,” enabling “automatic deductions.” Cited for the low-default class, the data-visibility mechanism and the sales-linked schedule. Publication.
2. CGAP, “Responsible Digital Credit for Merchants: Insights from Kenya.” Kopo Kopo’s Grow lets “merchants repay their advances as a percentage of the digital transactions they receive,” so “merchants are not on the hook for set weekly or monthly payments, which is especially helpful when business is slow.” One merchant: “you hardly feel the burden of paying the advance.” Cited for revenue share absorbing a loss, the kind case. Blog.
3. NerdWallet, “What Is a Merchant Cash Advance (MCA)?” MCA fees are charged as a factor rate that “typically ranges from 1.1 to 1.5”; the holdback or “retrieval rate” is a percentage of sales that “typically ranges from 5% to 20%,” with “the exact rate” varying by “your lender, the advance amount and your sales volume”; and the cost is “expensive,” with “effective APRs may range from 40% to 350%,” on advances “repaid within three to 18 months.” Cited for the terms-set-from-observables mechanism and the cost range. Guide.
4. NerdWallet, “What Is a Merchant Cash Advance (MCA)?” On the fixed-debit variant: “fixed repayments are made daily or weekly from your account regardless of how much you earn in sales, and the fixed repayment amount is determined based on an estimate of your monthly revenue.” Cited separately from note 3 for the floor that converts a trough into a fixed bill. Guide.
5. re-cap, “Revenue-Based Financing (RBF): Terms, Cost and Guide.” “You repay it as a percentage of your revenue until you hit a repayment cap. No equity. No board seat. No fixed monthly installment.” Worked example: a €500,000 raise at a 10 percent monthly revenue share, capped at €1,000,000 of total repayment. Cited for the instrument’s core shape and the “repayments naturally slow” property. Guide.
6. Merchant Capital (South Africa), “SME Cash Advances.” A merchant cash advance provides “a lump sum of cash in exchange for a percentage of future sales,” where “repayment is made through a percentage of your turnover until the advance is paid off.” Cited as a concrete African revenue-share lender pricing off card and mobile-money turnover, distinct from the fixed-term structure in note 9. Product page.
7. Eytan Modiano, “M/G/1 Queues,” MIT 6.263 lecture 8. Derives the Pollaczek-Khinchine formula, W = λE[X²] / 2(1 − ρ), where ρ is line utilisation and E[X²] is the second moment of service time. Expected waiting time therefore rises with the variance of the service process and runs away as utilisation approaches one. Applied here to the outstanding advance as a backlog drained by a variable revenue share. Lecture slides PDF.
8. GSMA, “Maturing Global Mobile Money Market Hits $1.4tn in Transaction Value.” “More providers now offer adjacent financial products, such as credit, savings, and insurance. Credit is the most popular adjacent financial service offered by mobile money providers (MMPs), with a 73% increase” in the number of credit products year on year. Cited for the scale and growth of mobile-money-linked lending in the African context. Press release.
9. Lula (South Africa), “Business Funding.” Lula’s funding is repaid “over 3, 6, 9 or 12 months,” a fixed-term schedule, with eligibility from a “minimum annual turnover of R500 000” and “one year trading history.” Cited as the fixed-schedule foil: a turnover-qualified lender whose repayment does not flex with a slow month, in contrast to the pure revenue share in notes 2 and 6. Funding page.