Affiliate Commission Rates

There is no single normal, which is most of why this study exists. Rates move with the vertical, with whether the commission recurs, and with the price of what is being sold, low-priced products tend to carry a higher percentage and a lower absolute payout. Read the vertical table rather than the headline number, and read the sample size next to it.

A percentage where order values vary, a flat amount where the price is fixed. Flat amounts have a second advantage that is easy to miss: a partner can repeat the number out loud. A creator who can say what they earn per sale promotes more consistently than one who has to compute it.

Not on its own, the structure decides it. 20% once on a $49 subscription and 20% on every renewal for the life of the account are the same percentage and very different businesses. If the recurring version does not survive your margin, cap it by number of payments rather than dropping the rate.

No, and most that work do not. Capping by number of payments or by months keeps the recruiting argument, a high rate for a defined period, while bounding what you owe. Referly expresses this directly: a rate, a cap in payments or months, and optional step-down periods after it. Step-down schedules start on the Business plan.

Every program in our sample runs the 60-day default, so we do not publish that as an industry finding: it measures how many people changed a setting, not what the market prefers. As a working rule, 60 days is a fair standard and anything under 14 reads as bad faith to experienced partners.

When your margin changes, when your partner mix changes, or when recruiting stalls for a reason you can trace to the offer. Not on a schedule. Cutting a rate after launch is the single most damaging thing you can do to a partner relationship, so set one you can keep paying at ten times the volume.

It comes from live affiliate programs running on Referly, aggregated so that no individual program is identifiable, and nothing drawn from fewer than fifteen programs is published. And yes, this study, its tables and its charts may be republished with attribution and a link back to this page. The aggregate data is downloadable as a CSV, ungated.