Methodology

Where the data comes from

Every performance figure on this site is pulled from affiliate network APIs and reporting exports — Awin, CJ Affiliate, Impact, Rakuten Advertising, Skimlinks, FlexOffers, Mavely, AvantLink, Partnerize and LinkConnector. Nothing is transcribed from a merchant's recruiting page. If a network doesn't report a metric for a program, the cell shows a dash; we never fill gaps with estimates.

Networks refresh on a rolling schedule — the current vintage per network:

Network-reported metrics move, and commission terms get renegotiated — treat every figure as a strong signal, not a contract, and confirm current terms inside the network before you apply.

What the metrics mean

EPC (earnings per click) is the network-reported average earned per click sent to the program, usually over the trailing one to three months. It bakes in commission rate, conversion rate and basket size at once, which makes it the single most honest number in affiliate marketing — and the one merchants least like to publish. A high advertised commission with a low EPC means clicks aren't turning into money.

Commission is shown as reported by each network, per network. The same brand frequently pays different rates — and different structures, percentage versus flat — on different networks. We show every row rather than averaging them.

Cookie window is the referral period attached to the tracking cookie. A 2-day Skimlinks cookie and a 30-day direct-program cookie are different businesses; the tables make that visible.

Conversion rate and reversal rate come straight from network reporting where available. Reversal rate — the share of commissions later cancelled — is the quiet killer in verticals like travel; a program paying 8% with a 40% reversal rate pays less than one at 5% with clean tracking.

Ghost rate is ours: the share of live programs in a vertical that publish no performance data at all. Ghosts are counted and flagged, but excluded from rankings — a program with no data can't earn a rank.

House rules

• No blended scores. We never merge networks' figures into an invented composite. Data is presented side by side; you weigh it.
• Data is never edited for commercial reasons. Reported figures appear as the networks report them. Any sponsored or featured placement on this site is labeled as such and never alters a program's data, its position in data-ranked tables, or its flags.
• Outliers stay out of rankings. A tiny program can post a spectacular EPC on thin volume. Values far outside a vertical's norm — EPC above roughly 10× the vertical median, conversion rates above 25%, cookie windows longer than a year — are usually small-sample or data-entry artifacts on the network side. We exclude them from ranked and "at a glance" tables, while the complete lists show every value as reported.
• Editorial is human-reviewed. Analysis drafts are produced with the help of AI tooling against our database, then reviewed and edited before publication.

Change tracking

Every data refresh is snapshotted. Between snapshots we record three kinds of change per program: launches (new in the network's catalogue), closures (delisted — dead programs are removed from rankings and listings, not left to mislead), and material term changes (EPC, commission rate or cookie window moving more than 5%). The alerts digest — weekly at launch, currently in waitlist — will report the notable ones; between our March and August 2026 snapshots that meant 17,451 launches, 16,639 closures and 24,673 term changes across the database — roughly a third of all programs moved in six months. Change history starts from the first snapshot pair — it cannot be reconstructed retroactively, which is why we keep it.

How matching works

The Browse tool scores every program in the database against the programs (or niche) you enter, on six weighted factors: niche fit, commission model, EPC proximity, price point, network overlap and cookie window — plus an editorial co-occurrence layer built from 1,100+ independent "best affiliate programs" articles across 130+ niches, which tells us which programs are actually promoted together. Matching runs in your browser against the public dataset; your picks are not linked to your identity.

We validate the engine continuously: it is stress-tested with batches of test portfolio submissions — over 1,000 to date — each run's recommendations reviewed and graded for audience fit, with systematic failures fixed before they reach you. Recommendations are starting points, not guarantees; confirm every program's current terms before applying.

Disclosure

Some outbound links are affiliate links, meaning this site may earn a commission if you sign up through them. This does not influence rankings — see the rules above — and most programs listed carry no link at all.