A SaaS Affiliate Program: Commission on Revenue You Have Not Collected Yet

The invoice arrives in a month when nobody signed. Twenty dollars is still owed on a customer who joined in the spring, and that number never went through an approval. The first twenty was a rate set at launch. The twenty in month nineteen comes out of a margin since promised to a hiring plan. Between those payments sits the argument about recurring rates, and the terms get signed before the second one is priced. Commission follows the customer after the click, and the structure that looked generous in month one can be the cheaper one by month thirty six. That flip has a month, and it is computable.

Why a Subscription Breaks the Ecommerce Affiliate Model

An ecommerce program pays for a transaction that finishes before the commission is calculated. A SaaS program pays for a subscription that keeps going after the affiliate forgets the click.

US advertisers will spend about $13.81 billion on affiliate marketing in 2026, up 11.3%, according to eMarketer. The playbooks came from ecommerce and brought its habits with them: the PMA 2026 US Brand Survey found 80% of programs still on last-click attribution. That is a fair default when a sale happens once, and a poor one when a customer stays three years.

Retention is no longer the missing idea. The arithmetic that turns it into a rate from the paying side still is.

Three Commission Structures, Priced Over Three Years

One boundary first. A program that pays for introductions between firms that already know each other is a different instrument, covered in referral partner programs. This article prices the public program instead: a tracked link, open to strangers, where the rate has to survive a customer who stays longer than planned. Every figure below is an illustration on a $100 monthly plan.

Churn risk is the column programs ignore; month 36 is the one they read afterward. A percentage structure pays nothing for a customer who leaves in month two. A bounty pays in full.

The Break-Even Month, in One Line

The structures cross at a computable point. Call the bounty B, the rate r, and monthly revenue from one customer M. The recurring structure has paid B by month B / (r x M).

At a $100 bounty, a 20% rate and a $100 monthly plan: 100 / (0.20 x 100) = 5. Month five is the crossing point. Before it the bounty is cheaper; after it the recurring structure is, and stays more expensive every month it pays.

Plan price moves the crossing point hard. On a $500 plan the same bounty is recovered in month one: 100 / (0.20 x 500) = 1. On a $25 plan it takes 20 months. Nineteen months of difference, decided by your price alone. One rate across all your plans hides a different deal on each one.

Why the Rate Follows From Gross Profit

Median software gross margin was 80% across 342 companies reporting FY2025 actuals in the Aleph x Benchmarkit 2026 study. On a $100 plan, 80% margin means $80 of gross profit a month. A 20% commission is $20. As a share of gross profit, that is 25%.

Retention does not change that ratio. It changes how many months you keep paying it. A customer who stays three years is billed thirty six times, each bill taking the same quarter of that month's gross profit. A customer who leaves in month two is billed twice, at the same 25%. Only the total moves.

Recurring commission does cost more when retention is high, because you pay it more times. That is the point of paying it. The error is the conclusion drawn from that: that the rate itself becomes less affordable. Twenty percent at 80% margin is a quarter of gross profit whether a customer stays two months or forty. If a quarter is too much to hand to a link, the rate is wrong on day one and retention will not fix it. Commission is one of the SaaS pricing decisions whose cost keeps landing after the contract is signed, and a rate copied from another program page imports that company's margin and churn into yours.

The Retention Test Before You Set a Rate

An illustration, not a recommendation: plan $100, margin 80%, commission 20%, bounty $100.

• Month 1: $20 against $100, so the bounty is $80 ahead.
• Month 5: $100 against $100. The crossing point.
• Month 12: $240 against $100, so recurring is $140 ahead.
• Month 36: $720 against $100, so recurring is $620 ahead.
• Share of gross profit in each of those months: $20 of $80, which is 25%.

In month 36 the recurring structure has paid $720, which is 25% of gross profit thirty six times. The bounty paid 125% of one month's gross profit once.

Now the part the comparison omits. The median SMB customer does not stay 36 months. CRV's 2026 churn benchmarks put monthly logo churn in the $500 to $5,000 ACV band near 4.1%, roughly 39% annualized. At a constant monthly rate the expected number of paid months is 1 / 0.041, so about 24. That division is mine; CRV publishes the churn rate, not the expected payout. Expected recurring commission is then about 24 x $20, or $480, against a $100 bounty.

At that churn level the expected payout falls from $720 to about $480, a third lower than the month-36 row suggests. The gap to the bounty narrows from $620 to $380, and it does not close: the structures meet only at about five expected paid months, which means churn near 20% a month. At 4.1% the recurring structure stays more expensive in expectation. What it buys for that extra $380 is that you pay only for customers who stayed, while the bounty is paid on a customer who may leave in month two. That is a risk decision sitting on the arithmetic. And the curve is steep. CRV's own example has monthly churn doubling from 1.5% to 3.0% and cutting lifetime value from about $106,700 to about $53,300. Median gross revenue retention fell from 88% to 84% in Benchmarkit's 2026 study, as reported by The SaaS CFO.

Attribution When the Trial Is Free

Everything above assumes you know which customer the affiliate delivered. On a free trial, you often do not.

ChartMogul's SaaS Conversion Report from January 2026, covering 200 products, puts median free-to-paid conversion at 8%. The spread is wide: opt-in trials without a card convert at 8.9%, opt-out trials with a card at 31.4%, freemium at 5.6%. A 2025 series puts those at 18.2% and 48.8%, so the sample moves the figure. Use your own number, but use one.

Timing matters too. Most B2B conversions happen at trial expiry, day seven is the hinge, and after day 14 conversion falls to about 1%. ChartMogul puts it plainly: one percentage point of conversion is worth roughly 15% more new revenue per cohort.

Then the cookie. TrackRev's Q2 2026 synthesis of partner-industry reports puts the median cookie window at 60 days, with 41% of programs at 60 days and only 7% at 180 days or longer. Optifai's 2026 pipeline study of 939 B2B SaaS companies puts the median sales cycle at 84 days, with a buying committee that has grown to 6.8 stakeholders. Subtract one published figure from the other and the median program's attribution window closes 24 days before the median deal closes. That subtraction is mine; neither source makes the claim.

Walk one visitor through it. The click lands on day one, the trial converts on day 26, the contract closes on day 84, or later in enterprise, where cycles run 90 to 180 days. On a 60-day window the affiliate is paid for a conversion inside the window and unpaid for the deal that closed outside it. The program then reports that the affiliate brought a trial, not a customer, and affiliates who bring durable accounts look exactly like the ones who bring tire-kickers.

A longer window is not automatically the fix. Match it to your trial model. With a card-required trial converting at 31.4%, a short window is defensible: the payment event sits close to the click. With an opt-in trial converting at 8.9% and a cycle 22% longer than in 2022, a 60-day window measures the wrong event.

What Belongs in the Terms

I am not your lawyer, and this is not legal advice. These three clauses matter once a stranger holds a link, and each is expensive to add later.

The clawback window should be at least as long as your refund period. If commission is paid on a payment that can still reverse, the window is the only mechanism that gets the money back. Without it you have paid a bounty on revenue that did not survive its guarantee.

Brand bidding is the clause that protects your search budget. FirstPromoter's Q1 2026 detection data attributes up to 76.6% of detected affiliate fraud cases to Google and names coupon stuffing as the largest leak, because last-click always credits the most recent cookie. The clause can be short: no brand-term bidding, no coupon interception, a named consequence. It needs a monthly check, because a clause nobody reads is a wish.

Disclosure is the clause the advertiser carries. Under the FTC Endorsement Guides, 16 CFR Part 255, in the 2023 revision current in 2026, a commission creates a material connection that requires clear and conspicuous disclosure, and the advertiser shares liability with the endorser when disclosure is missing. Put the requirement in the terms and hand affiliates the sentence.

None of this replaces the unpaid layer. Customers who recommend you without a tracked link cost nothing a month, covered in customer advocacy. A program that counts only the paid one measures half its referrals.

Conclusion

Recurring commission is a claim on revenue you have not collected yet, and its cost is decided in three places: the month two structures cross, the share of gross profit the rate takes every month after, and whether your attribution window reaches the close. Get those right and the rate becomes a decision, and what you then track against it belongs to your [marketing KPIs](/blog/marketing-kpis/). Skip them and it is a copy of somebody else's.

That invoice will arrive again next month. The only question worth answering before it does is whether the number on it came from your margin or from a page you read at launch.