Does an affiliate program pay for itself? The payback maths, with every assumption labelled

By Jimi Barkway · Published 7 September 2026 · Part of the blog

Does an affiliate program pay for itself?

In shortThe software does, in the first month, if even three of a hundred approved partners send one customer. The commission is a share of sales you would not otherwise have had, so it pays for itself by definition. The conditions are that those customers would not have come anyway, and that the rate sits under your margin. The line that determines the answer is not the plan price. It is how many partners ever send anyone.

The research method was straightforward. I took the activation rate from the largest referral dataset in the category. The commission is the reward printed on 358 public program pages. The software and fee lines are from the calculator on this site. Then I ran one worked example, with every assumption written down so you can substitute your own numbers. I discovered that the payback question is settled in month one on the software side. Everything after that is a question about margin and activation.

Three words first. Activation is the share of approved partners who ever send you a single person. A commission is what you pay a partner when someone they sent buys. And payback is the month in which what the program brought in passes what it cost. Those three determine the entire sum.

What does the program cost before anyone sells?

In shortThe software, and nothing else. From $0 to $90 on the calculator's entry plans: $49 on Rewardful or FirstPromoter, $69 on Tolt, $90 on Dub. No commission is owed until a sale, and no payout fee until a payout. The activation data says the first week is when partners who will ever refer mostly begin. So the wait is short or it is long.

One more line belongs on the before side, and it is not money. The activation data says half of the partners who ever refer do so within seven days of joining. So the first week after an approval is the one that decides whether that partner counts. A founder who approves a hundred partners and then goes silent has spent the plan price on a list. One who writes to each of them in the first week has spent it on a program.

That is the entire downside of a program that does not work: the plan price. Twelve of the twenty platforms on the calculator charge nothing on payouts, so on those the fixed cost is the only cost. A year of $29 is $348. A year of $49 is $588. A year of $99 is $1,188. Nobody has lost a business to that line, and the cost article puts it beside the others.

What a founder actually spends before the first sale is time. Reviewing applications, writing the portal copy, setting the rate. The one-week guide sizes that at an afternoon for the software and a week for the decisions, the terms and the tracking check. The activation data says three quarters of the partners who ever refer have done so by day 49. So by the end of the second month you know whether the program has a pulse.

How fast does the software pay back?

In shortIn the first month with two sales on a $49 plan, or one sale on the plans under $35. Let's say one partner sends one customer on a $50 plan. That is $50 of revenue against $49 of software, $12.50 of commission and $0.25 of PayPal. Not quite, by $11.75. Two customers and it is back. In the worked example above, six partners send six customers in month one, so $300 covers $125.50 with room.

Run the low case. Let's say activation is at the bottom of the middle half, 6%. Only three of a hundred partners ever deliver a buyer, at one customer a month each. Month one is $150 of revenue against $49 plus $37.50 plus $0.75. Still paid back. Year one is $11,700 of revenue, $2,925 of commission, $588 of software and $9 of fees. That is $8,178 net. Half the example's figures, identical answer.

Run the empty case. Nobody refers in month one. The dataset says a program with no referral by day 49 is unusual. The cost is the plan price. That is the bet. The upside case in the table is $16,944 net in year one from six working partners. The reading-the-numbers module of the manual has the dashboard view of the same calculation.

Does the commission pay for itself?

In shortBy definition, if two conditions hold. The customers would not have come without the partner, and the rate is under your margin. At 25% you retain 75% of a sale you did not have. The commission fails only when a partner is paid for a customer who was coming anyway. That is what the self-referral, coupon-site and brand-bidding rules exist to prevent.

The first condition is a rules problem, not an arithmetic one. A customer who finds a coupon site at checkout and clicks a partner's code was already buying. A partner who bids on your brand name is intercepting your own traffic. A customer who signs up through their own link is paying themselves. The program rules module covers all three in plain English, and the coupon article covers the checkout case.

The second condition is margin. A margin is what is left of a sale after the cost of serving it. The commission has to fit inside it with room. The dataset's 25% for 12 months is what programs publish. Whether it suits your product is the commission module's question. For an AI product with a model bill behind every seat, the AI SaaS page covers paying on the margin rather than the sale.

When should you decide it is not working?

In shortDay 49, on the activation data. Three quarters of the partners who will ever refer have done so by then, and half within the first week. So a program with no referrals after seven weeks is not waiting for a slow beginning. It has the wrong partners, the wrong offer, or the wrong rules. The cost of discovering that was two months of software.

That is the useful shape of the dataset for a founder deciding whether to continue paying. The bet is bounded on both sides. On the downside, $0 to $90 a month for two months. On the upside, six working partners out of a hundred and the table above. The middle half of programs sit between 6% and 28% activation. So a program where 3 of 100 bring a buyer is at half the median, and one where 6 or 7 do is ordinary. Both pay for the software in month one.

What should you measure to know?

In shortFour numbers, monthly. Partners approved. Partners who have referred anyone. Customers the program delivered, and what they paid. Commission and software paid. Divide the last by the third and you have the program's cost per dollar of revenue. At 25% and a $49 plan that settles near 27 cents on commission and software in the worked example by month twelve.

For example, imagine month three. Forty partners approved, five have referred, eleven customers on the books paying $550 a month. That is $137.50 of commission and $49 of software. The program cost $186.50 to generate $550 that month, and those eleven customers renew. Now imagine month three with forty approved and none referring. The program cost $49 and generated nothing. The activation data says that is unusual by day 49. Review the applications again, and the activation article for where the referring partners come from.

A word on this site's own product, with its trade-offs. AffiliateRail is $29 a month founding to $10,000 of affiliate revenue, and charges 0% of every payout. From the $59 plan it pays partners from your own PayPal or Wise on a schedule. So the fixed cost in the empty case is $29, and the fee line in the working case is PayPal's or Wise's alone. Its day-one flow is 20% with a 14-day hold. It has no marketplace, so the partners in the table are yours to recruit.

Every dataset figure on this page is from the linked sources on the dates beside them. The assumptions are labelled as assumptions. Substitute your own numbers before you decide anything.

Common questions