What B2B SaaS pays for a referral, stated terms only
Should you launch an affiliate program, and what would you have to offer? We read every public affiliate and referral page across 479 B2B software companies and recorded the stated terms: 32 real programs, commission ranges, recurring periods, cookie windows, and flat bounties.
Executive summary
Most B2B SaaS companies do not run a public affiliate program: 32 of 479 (6.7%, 95% CI 4.8-9.3%) state affiliate or referral terms on their own site. What looks like a program at first glance is usually a channel partnership (22 reseller/agency tracks found), a blog post about referrals, or a job-referral page.
Among the programs that state terms, three patterns hold. First, commissions are ranges, not numbers: the median headline rate is 25% (IQR 20,30%), and most programs publish a band (5-20%, 20-30%, 25-50%) that moves with volume or plan. Second, recurring beats one-off 11 to 4 among programs that state a duration, lifetime and first-year terms dominate; flat one-off bounties (£1,500 at humaans.io, $50 at treble.ai) are the exception. Third, cookie windows are almost never stated: 3 of 32 programs print a number (all three say 90 days), everyone else leaves attribution unstated.
How to use this: if you launch a program, the competitive bar is a 20-30% recurring band with a 90-day cookie and an application gate, that is what the staters state. If you join one as a creator, ask for the cookie window in writing; most programs in this sample do not publish it.
Commission ranges, ranked
Top of the stated range per program. Ranges move with volume, plan tier, or partner level.
| Program | Stated range |
|---|---|
flowglad.com | 50,50% |
mailwarm.com | 25,50% |
webflow.com | 10,50% |
close.com | 30,30% |
jasper.ai | 25,30% |
veryfi.com | 25,25% |
webiny.com | 25,25% |
assembly.com | 20,20% |
cashbook.in | 20,20% |
chatwoot.com | 5,20% |
noloco.io | 20,20% |
storylane.io | 20,20% |
customily.com | 10,10% |
usergems.com | 10,10% |
Recurring vs one-off
Flat bounties on record
| Program | Bounty | Basis |
|---|---|---|
cosmicjs.com | one-time build bounty | stated terms |
humaans.io | £1,500 flat bounty | stated one-off bounty |
treble.ai | $50 on first payment | stated one-off bounty |
usergems.com | $100 per demo held (+10% of closed-won, recurring) | stated terms |
vori.com | $2,000 one-time fee (terms) | stated one-off bounty |
Entry rates vs top rates: the band ladder
4 programs publish two or more distinct rates, the entry commission and the top creators can earn.
| Program | Entry | Top | Width |
|---|---|---|---|
webflow.com | 10% | 50% | 40pts |
mailwarm.com | 25% | 50% | 25pts |
chatwoot.com | 5% | 20% | 15pts |
jasper.ai | 25% | 30% | 5pts |
Single-rate programs
| Program | Rate |
|---|---|
flowglad.com | 50% |
close.com | 30% |
veryfi.com | 25% |
webiny.com | 25% |
assembly.com | 20% |
cashbook.in | 20% |
storylane.io | 20% |
noloco.io | 20% |
usergems.com | 10% |
customily.com | 10% |
Two motions: creator affiliate vs customer referral
14 programs pay percentage commissions to creators and partners; the rest are customer-referral mechanics (credits, bounties, rewards) with or without stated amounts. The two motions price differently: affiliates get bands (5-50%), referrals get fixed bounties (£1,500, $2,000, $50) or account credits.
Fuse them and you misprice both. A creator comparing your 20% against a competitor's $50 bounty is comparing a revenue share against a finder’s fee, the page should say which motion each number belongs to, as tella.com's separate affiliate terms and cosmicjs.com's separate creator bounty do.
Gated vs open: 7 programs screen first
Application or approval gates appear on close.com, webflow.com, deel.com (via PartnerStack), noloco.io and assembly.com, plus approval-at-discretion terms at tella.com (via Dub) and assemblyai.com.
Gates correlate with generosity: every program above 25% in this sample screens applicants. Open-door programs cluster at 10-20%. The trade is explicit, higher rates buy selectivity, and founders setting rates should set the gate in the same decision.
Where programs live, and who runs one
Program URLs by path shape. No single convention dominates; affiliates hide under partners, legal, and solutions as often as under /affiliates.
| Path shape | Programs |
|---|---|
/terms/redirect page | 17 |
/partners | 4 |
/affiliates | 2 |
/partner-program | 2 |
/affiliate-program | 2 |
/referrals | 1 |
/legal/affiliates | 1 |
/affiliates-program | 1 |
/partner | 1 |
/affiliate | 1 |
| Segment | Program rate |
|---|---|
| 1-10 employees (n=129) | 5.4% |
| 11-50 employees (n=190) | 6.3% |
| 51-200 employees (n=111) | 9.0% |
| 200+ employees (n=41) | 7.3% |
| Per-seat (n=63) | 11.1% |
| Flat / tiered (n=117) | 8.5% |
| Usage-based (n=50) | 8.0% |
| Contact-only (n=115) | 3.5% |
What the "partner" pages actually offer instead
22 excluded pages are reseller, agency, or technology tracks, a different motion from affiliate. Three shapes cover nearly all of them: certified-agency directories (whalesync.com, revenuecat.com), co-sell/implementation tracks (docker.com channel partners, rippling.com referrals-inside-ecosystem), and technology/embed tracks (revenuecat.com technology program, observe.ai). No public commission appears on any of them.
The naming lesson for partnership teams (see who-for below): if you pay per referral, call it affiliate or referral. "Partners" alone reads as certification, and 22 vendors in this sample prove that affiliates bounce off certification walls.
Who this is for
Launch at 20-30% recurring, 90-day cookie
That is the modal stated offer in this cohort. Anything less generous needs a justification on the page (brand pull, high ticket); anything more generous should say what the affiliate must do to earn the top of the band.
Get the cookie in writing
Only 3 of 32 programs publish a window. Duration and attribution decide your earnings more than the headline rate, a 50% one-off with no stated cookie loses to 20% recurring with 90 days.
Do not confuse partner tracks with programs
22 "partner" pages in this sample are reseller, agency, or tech tracks with no public commission, a different motion (co-sell, certification) from affiliate (pay-per-referral). Name yours correctly or affiliates will bounce off a certification wall.
Stated terms only, per-domain union
One program per domain (best-evidence URL kept); blog posts, templates, job referrals and integration docs excluded with reasons in the data. Reuse the CSV with those exclusions attached.
Methodology, deviations & reproducibility
What we read: all 479 company domains checked for affiliate, referral and partner-program paths (direct paths plus homepage- and sitemap-linked pages), each URL robots-checked before fetching. 89 domains showed program-like pages; content validation (homepage-redirect, thin-page and terms-language checks) plus commission-context verification (% figures only count near commission/payout/earn language) and hand review of all 26 ambiguous pages left 32 genuine programs. Excluded with reasons: 22 reseller/agency tracks, 20 job-referral pages, 13 articles about programs, plus templates, integration docs, glossaries and press releases.
Hard limits kept: no sign-ups, no applications, no link testing, stated terms only. No payout-behaviour claims, no generosity rankings.
Downloadable dataset: programs.csv (per-program kind, rates, duration, cookie, notes).
python3 scripts/research/s6_discover.py # program discovery (resumable)
python3 scripts/research/s6_fetch_extract.py # fetch + term extraction
python3 scripts/research/build_s6_page.py # rebuild this page
Corrections
Factual corrections are published on the page; send the URL to amit@datadab.com.
Covering this research? Please link the primary source so readers can verify the data and methodology:
Markdown: [B2B SaaS Affiliate & Referral Benchmarks 2026 by DataDab](https://www.datadab.com/research/b2b-saas-affiliate-benchmarks/)
Spot an error? Send the URL to amit@datadab.com; corrections are published on the page.
What this study does not show
- Private and network-hosted programs. Programmes behind logins, applications, or affiliate networks are unmeasured; only public on-domain terms count.
- Actual payouts. Stated terms, never tested behaviour.
- Single reviewer. Extractor-assisted coding with full hand review of ambiguous pages; no independent double-coding yet.