If you run marketing at a $10M–$100M brand, you have almost certainly been pitched an affiliate program as "free growth — you only pay when you make a sale." That is half true, and the half that is missing is exactly the half your finance team will ask about. Affiliate is genuinely pay-for-performance at its core, but a program carries a real, layered cost structure, and the headline commission rate is the smallest part of the story for some brands and the largest for others.
This guide is for the person who is going to build the model. We will walk the full cost stack layer by layer, reframe the channel around the metric that actually matters to a finance reader, and run an illustrative program through the math so you can see where it lands relative to your other paid channels. Every figure here is an example, clearly labeled — your verticals, margins, and partner mix will move the numbers — but the structure is what you can take to a budget meeting.
Why how much does it cost is the wrong first question.
For most channels, you set a budget and spend it. Affiliate inverts that. You set commission terms, and cost scales with revenue the channel produces. A program that does $50k in attributed revenue this month and $500k next month costs roughly ten times as much in commissions — but the cost of sale, the percentage, stays roughly flat. That is why a finance-literate operator should ignore the raw dollar figure and anchor on two ratios from the start: cost of sale (total program cost divided by attributed revenue) and the channel's contribution to contribution margin.
The second reframe is harder and more important: the revenue your platform reports is almost always inflated by last-click attribution. Some share of those "affiliate" sales would have happened anyway — the customer was already going to buy and a coupon site caught the final click. Pay on that and you are buying sales you already owned. The only honest denominator is incremental revenue, and the only honest cost-per-customer is incremental CAC, not blended CAC. We will come back to this; it is the single biggest place finance and marketing talk past each other.
Layer 1: Commissions, the variable core.
The commission is what you pay a partner when they drive a sale or a qualifying action. It is structured as cost-per-sale (a percentage of order value) or cost-per-action (a flat bounty per conversion). This is the only cost layer that is purely variable and purely performance-based — no sale, no commission — and it is the layer everyone thinks of as the cost of affiliate.
Rates vary widely by vertical, and the logic is margin. Retail and general DTC commonly sit around 5–15% of sale because margins are thinner and order values are well understood. High-margin verticals — supplements, beauty, some apparel — can support 15–25% or more because the contribution margin absorbs it. Subscription and SaaS programs often pay flat bounties per signup or a recurring share (commonly 20–30% recurring for a defined window) because the lifetime value justifies a larger up-front share. These are ranges, not quotes; your own numbers depend on margin, AOV, and how aggressively you want to recruit.
| Vertical | Typical commission structure | Why |
|---|---|---|
| Retail / general DTC | ~5-15% of sale (CPS) | Thinner margins; well-understood AOV |
| High-margin (supplements, beauty) | ~15-25%+ of sale | Contribution margin absorbs a higher payout |
| Subscription / SaaS | Flat bounty, or ~20-30% recurring for a set window | LTV justifies a larger up-front share |
| Considered / high-AOV goods | Lower % or flat CPA | Large order values make % payouts expensive |
Layer 2: The network override or platform fee.
If you run your program on an affiliate network, you pay a network override on top of every commission — commonly in the ~20–30% range of commission paid. Pay a partner $100 and the network takes roughly $20–$30 for tracking, partner marketplace access, payment rails, and compliance. It is a real, recurring surcharge that scales with your commissions, and it is the line budgets most often drop.
The alternative is a SaaS affiliate platform, where you pay a flat monthly or annual fee instead of a percentage override and bring (or recruit) your own partners. At low volume the override is cheaper; past a crossover point the flat fee wins because it does not scale with revenue. Working out where your program crosses that line is its own analysis — we cover it in network vs platform costs — but for modeling, the point is simple: this layer is real, and it is either ~20–30% on top of commissions or a fixed fee you must amortize over expected revenue.
Layer 3: Management, the line everyone underestimates.
A program does not run itself. Someone recruits partners, negotiates terms, polices fraud and coupon abuse, produces creative, and optimizes the partner mix toward incremental sources. You pay for that capability one of two ways.
In-house: a dedicated affiliate manager is a real salaried line — often six figures fully loaded once you include benefits and overhead — and a program of any size needs meaningful fractions of that person's time long before it justifies a full headcount. Agency / OPM: an outsourced manager runs the program for a monthly retainer, a performance percentage of program revenue, or a blend of both. The retainer model is predictable; the performance model aligns incentives but adds another percent-of-revenue layer you must stack into the cost of sale.
Layer 4: Setup, creative, tooling, and fraud.
The smaller lines add up. Network or platform onboarding sometimes carries a setup fee. Creative — banners, text links, custom landing pages, refreshed seasonal assets — is real production work. Fraud and brand-safety monitoring (catching trademark bidding, cookie stuffing, coupon-extension abuse, and unauthorized discount leakage) is either a tool subscription or a chunk of your manager's time. Tracking and attribution tooling, if you go beyond what the network provides, is another subscription. None of these is large on its own; together they are a line item, not a rounding error, especially in the first year.
- Setup / onboarding fees (one-time, network- or platform-dependent)
- Creative production and ongoing refreshes
- Fraud, compliance, and brand-safety monitoring
- Tracking, attribution, and reporting tooling beyond the platform default
A worked model: a $200k/month program.
Put the layers together on an illustrative program doing $200,000/month in attributed revenue, on a network, at a 12% blended commission, a 25% override, and a performance-based OPM. These are example inputs to show the shape of the stack — not benchmarks to copy.
| Cost layer | Amount | As % of attributed revenue |
|---|---|---|
| Commissions (12% blended) | $24,000 | 12.0% |
| Network override (25% of commissions) | $6,000 | 3.0% |
| Management (OPM, ~3% of revenue) | $6,000 | 3.0% |
| Tooling, creative, fraud monitoring (amortized) | $2,000 | 1.0% |
| All-in program cost | $38,000 | 19.0% |
So the headline 12% commission becomes a ~19% all-in cost of sale once the override, management, and tooling layers are stacked in — roughly 60% higher than the commission line alone. That is the number to compare against your other channels. If your blended paid-social-and-search CAC implies a cost of sale in the mid-20s to 30s percent for comparable new-customer revenue, a ~19% affiliate cost of sale looks strong. If it implies low teens, affiliate has to win on incrementality and reach rather than on raw cost. The comparison only means something when both sides are measured the same way.
One more adjustment a finance reader will insist on: that 19% is a blended cost of sale across all partners. Strip out the non-incremental coupon and loyalty volume — the sales you would have closed anyway — and recompute against incremental revenue only. The incremental cost of sale (and the incremental CAC behind it) is higher than the blended figure, sometimes materially. That gap is not a reason to kill the program; it is the number that tells you which partners to grow and which to cap.
ROAS, payback, and realistic timelines.
On a pure-ratio basis the channel looks great: a 19% cost of sale is roughly a ROAS of about 5.3:1 on attributed revenue. But two realities temper the headline. First, that ROAS is gross and last-click; the incremental figure is lower. Second, a program does not arrive at steady state on day one.
Programs ramp over months: recruit partners, activate them, then compound as the productive ones scale and you prune the rest. The first quarter is mostly investment — management cost and setup against thin revenue, which means an ugly cost of sale and a longer payback period early. By months six to twelve a healthy program typically settles into its mature cost of sale, and a well-run one becomes one of the most efficient channels you have on a cost-per-incremental-acquisition basis, because you are paying a curated set of partners only for results.
“Affiliate is the rare channel where the cost structure is mostly variable and mostly self-correcting, but only if you measure it on incremental lift. Measure it on last-click and you will scale exactly the partners you should be capping.”
Is an affiliate program really free because you only pay on sales?
No. Commissions are pay-for-performance, but the override, management, tooling, and fraud-monitoring layers are real costs that exist regardless of any single sale. The commission line is often only about two-thirds of the all-in cost of sale once everything is stacked in.
What all-in cost of sale should we expect?
It depends entirely on your vertical, margin, and partner mix, so treat any single number with suspicion. As an illustrative shape: a retail program at a low-teens commission, on a network, with managed support often lands in the high teens to mid-20s percent of attributed revenue all-in. Recompute against incremental revenue for the number that actually matters.
Why does our platform report higher affiliate revenue than finance credits us with?
Because the platform uses last-click attribution and counts every sale that touched an affiliate link, including ones that would have closed anyway. Finance wants incremental revenue — sales the channel actually caused. The gap is normal; the fix is incrementality testing, not arguing over the dashboard.
In-house manager or an agency (OPM)?
In-house gives control and is a fixed cost; an OPM is variable or semi-variable and brings existing partner relationships, which helps most in the early recruiting-heavy phase. Many brands start with an OPM and bring management in-house once volume justifies a dedicated headcount.
How long until the program pays back?
Plan for months, not weeks. The first quarter is largely investment against thin revenue and an ugly early cost of sale. Most healthy programs reach their mature, efficient cost of sale somewhere in the six-to-twelve-month window as productive partners compound and weak ones are pruned.
How do we avoid overpaying coupon and loyalty partners?
Measure incrementality with holdout or test-and-control. Partners who only catch the final click on sales you already owned show low incremental lift; cap their rates or move them to a lower commission tier, and reallocate budget toward partners with proven incremental reach.
The bottom line.
Affiliate is one of the few channels whose cost is mostly variable, mostly performance-based, and — run correctly — mostly self-correcting toward efficiency. But run correctly carries weight: budget all four cost layers, not just commissions, and judge the program on incremental CAC against your other channels rather than on inflated last-click revenue. Do that, and a mature program often becomes one of the most defensible lines in your acquisition mix. If you want a partner to build that model and stand the program up against it, tell us about your brand.