How should you set affiliate commission rates for each partner?

Set each partner's affiliate commission from what its orders cost: rate, code discount and network fee, over the share of orders it actually caused.

Author:Benson
PartnershipsUpdated 12 October 202616 min read
Jump to a section

In short

There's no reliable industry rate to copy: published averages come from platforms' own data. Set each partner's rate from what its orders cost you, the commission plus the discount on its code plus the network fee, divided by the share of its orders a holdout shows it caused. Pay more for new customers than returning ones, and test fees, bonuses and tiers like any other cost.

The quarterly rate review is on Thursday. Your deck has one table: each affiliate partner, its commission rate and the revenue the network credits to it. The coupon site is on the lowest rate in the programme and tops the revenue column. The review site you want to move up is on twice that rate.

Finance sends it back with two columns added: the discount on each partner's code, and the share of its orders that disappeared in your spring holdout. Their note is one line. Why does the cheap partner cost more for every order it actually brought in?

Their sums hold. The rate in the contract was only one of the costs on the order, and the revenue column counted orders you'd have had anyway.

This guide covers setting each partner's rate from the whole cost of its orders: the benchmarks, the arithmetic, the structures and network settings, and what happens when you change a rate. It follows from a wider question, which affiliate partners you should pay at all, and for what.

What is the average affiliate commission rate?#

There isn't a reliable one. The averages that circulate are either unsourced or drawn from one platform's own programmes, as its tracking credited the sales.

The most detailed figures come from Partnerize, which sells a partnership platform and reports what its brands in the Americas pay: an average publisher commission rate of 6.9% in Q2 2025, varying widely by partner type. [Partnerize, Sep 2025] (opens the source)

4.84%

effective commission paid to coupon partners by Partnerize's brands in the Americas in Q2 2025, against 9.44% for content partners and 10.76% for comparison sites.

Partnerize, Sep 2025(opens the source)

Partnerize says its index isn't a proxy for the wider market. impact.com, which also sells the platform its data comes from, looked at 2,368 North American retail brands in 2025: loyalty and rewards partners took 33% of brand spend for 50% of transactions, and content review sites 24% of spend for 9% of transactions. [impact.com, Feb 2026] (opens the source)

Content publishers compare your offer with Amazon's US rate card: apparel 4%, home and beauty 3%, luxury beauty 10%. [Amazon Associates, Oct 2026] (opens the source) Shopify's 2026 commission guide passes on an agency founder's view that standard rates for well-known publishers run 10% to 15% a sale, an opinion rather than data. [Shopify, May 2026] (opens the source)

These are what the market pays, not what a partner is worth to you. None of them adds the discount the partner's code carried, or takes out the orders the shopper would have placed anyway.

What does a partner's order actually cost you?#

Three things, and the commission isn't always the biggest: the commission, the discount on the code the partner showed, and the network's fee. You pay all three on every order the partner is credited with, including the ones it didn't cause. So set rates from this:

Cost per incremental order = (commission + the code's discount + network fee) ÷ the share of the partner's orders that wouldn't have happened without it

The network fee sits on top of the commission. On Awin's Access plan in the US it's 3.5% of the sale value, or 2.5% on Accelerate, plus a monthly fee. [Awin, Oct 2026] (opens the source) impact.com charges 2.5% on partner-driven transactions. [impact.com, Oct 2026] (opens the source)

The discount is the cost most rate tables leave out. A coupon partner earns commission on an order that already carries the code it published, and you fund both; one DTC founder told Modern Retail that a leaked-code order is worth less to the brand and still pays the partner out, "a double whammy hit on margins". [Modern Retail, May 2024] (opens the source)

The incremental share is the hard input, and no controlled study publishes it by partner type. PartnerCentric, an agency that sells attribution technology, found coupon, loyalty and cashback partners started only 13% of the conversions credited to them across its clients in late 2024. [PartnerCentric, Feb 2025] (opens the source) In one controlled test measured by Haus, which sells incrementality testing, a DTC brand switched off its US loyalty and cashback partners for seven weeks and saw no change in revenue against comparable regions. [Haus, May 2023] (opens the source) Your share comes from your own test; the holdout guide covers one finance will accept.

The levers that matter are the depth of the code the partner may show, which is a decision about who gets a discount as much as a partner one, and what you pay on the orders it didn't cause.

Should you pay a percentage, a flat CPA or a fee?#

Pay a percentage of the sale by default, a flat amount where basket values vary more than your margin does, and fees only for placements you wouldn't otherwise get. CPA, cost per action, is commission paid when an agreed action happens, usually a sale.

A percentage of the sale. The default on every network: cost moves with the basket. The catch: a big basket earns a big commission whether or not the partner made it big.

A flat amount per order or per new customer. It suits ranges where margin doesn't rise with price, and pays for a customer rather than a basket. The catch: partners push whatever converts most easily.

Tenancies, flat fees and hybrids. A tenancy is a fixed fee for a placement over a period, such as a homepage slot; a hybrid adds commission on top. Close to one pound in five of UK affiliate spend in 2025 went outside last-click CPA, according to the APMA, the UK trade body. [Affiliate and Partner Marketing Association (APMA), Apr 2026] (opens the source) Webgains, a network, says cashback and loyalty partners usually expect tenancy payments for extra promotion. Adtraction's test is the right one: a fixed fee should always buy additional placements or activity. The catch: the placement's sales still earn commission, so on a hybrid you can pay twice for one order.

Bonuses and peak boosts. impact.com found brands raised commission rates by 7% in November 2025 to compete for placement. [impact.com, Feb 2026] (opens the source) The catch: a boost for the partner closest to checkout pays more for orders the season was bringing anyway, so tie bonuses to placements or new customers.

Tiers. Here a tier is a rate that rises with the partner's volume in a period, not commission on a second level of affiliates. The catch: last-click volume sits with the partners nearest checkout. In impact.com's 2025 data, loyalty and rewards partners had 15% of clicks and 50% of transactions, [impact.com, Feb 2026] (opens the source) so a ladder keyed to transactions hands your top rate to the partners whose orders are least likely to be extra. If you tier, tier on new customers.

How do you pay a different rate for new and returning customers?#

Set up a commission group or payout rule per customer status, and send the status with every order. No network can tell who is new on its own.

Awin. Commission groups apply a rate to transactions that match a condition; anything else falls back to Default. Access plans get three groups including Default; Accelerate and Advanced get unlimited. Awin's developer docs have you create groups coded NEW and EXISTING, then pass the group code with the order value on the conversion tag:

AWIN.Tracking.Sale.parts = "NEW:12.49";

A separate customerAcquisition field takes NEW or RETURNING. Awin's developer docs describe it as reporting, but on Awin Classic you can also set a rate on a Customer acquisition condition, so confirm with your account contact which one your rates read.

CJ. Situational Commissioning pays different rates on attributes passed with the order, including customerStatus with the values New, Lapsed and Return; CJ asks you to consult its team before adding them.

impact.com. The customerStatus parameter takes New and Existing by default, and impact.com says it can't identify new customers unless you send it. Payout groups set the rates, and only one applies to an action: the first that matches, from the top.

Shopify affiliate apps vary: Refersion's new and returning rates read earlier affiliate conversions unless you capture every order's email, and can't be combined with tiers.

The catch in all of them: "new" is whatever your tag says. And new isn't incremental: a first-time buyer who clicked your ad, then a coupon site's link, is new to you and still not the partner's doing. Partnerize's figures show how the mix can differ: in Q2 2025, new customers were 30.50% of coupon partners' revenue and 45.92% of content partners'. [Partnerize, Sep 2025] (opens the source) That's one quarter, so set a new-customer rate from your own split, not a benchmark.

That works only if the code really is limited to new customers, which the first-order discount guide covers. Networks also price by code: impact.com's Promo Code Exception Lists pay different rates, including nothing, on chosen codes, and Awin's voucher code condition does it on Accelerate and Advanced. That's how you pay a creator less when their code turns up on a coupon site; see the creator codes guide.

What happens when you cut a rate?#

The network limits how fast you can move, and the partner sees the change at once. Awin's help centre says a commission rate can be cut by at most 20% each time, no more than once every 30 days, with at least 7 days' notice to partners, unless your contract says otherwise. A partner on 5% can go to 4% in one step, and each further cut waits another 30 days. [Awin, Apr 2026] (opens the source)

Pay on what sticks, too. On Awin, pending sales approve themselves when the auto-validation period ends; you can't switch it off, it runs up to 67 days on Access, and Awin suggests your returns window plus seven days. [Awin Help Center, Jul 2026] (opens the source) NRF and Happy Returns, which sells returns services, estimated that 19.3% of US online sales would be returned in 2025. [National Retail Federation and Happy Returns (a UPS company), Oct 2025] (opens the source) Approve early and you pay commission on orders that come back. The catch: every extra week of holding is a week the partner waits to be paid.

The second-order effects land in three places. The partner sees the cut in its earnings per click, the figure publishers compare programmes by, and your placement may move; if a holdout says most of its orders would have happened anyway, that costs you little. Your reports show the partner's effective rate falling while its revenue doesn't, so tell finance which number to watch. And some of a late-funnel partner's orders will arrive through the next partner closest to checkout, so read the whole programme after a change.

Won't partners leave if you pay them less?#

Some will, so decide in advance which ones you can afford to lose. The first objection is that creators and publishers go where commission is highest. It's true, and brands compete on it: Modern Retail reported the supplement brand Obvi offering influencer partners commissions of up to 50%. [Modern Retail, Apr 2025] (opens the source)

Pricing by cost per incremental order pays those partners more, not less. A creator or review site that introduces customers has a high incremental share, so it can carry a higher rate, funded by what you stop paying on orders others joined late.

The second objection: pay coupon and cashback partners only for new customers and they'll drop you. Some might. Loyalty and cashback partners' revenue skews to existing customers: new customers were 25.68% of loyalty partners' revenue in Partnerize's Q2 2025 data, [Partnerize, Sep 2025] (opens the source) and some retailers pay for that on purpose: MandM Direct's head of digital told a 2024 Partnerize panel the retailer uses cashback partners for retention, as Hello Partner reported it. If that's your reason, pay a returning-customer rate chosen for retention, not an acquisition rate by default. And if a holdout shows most of a partner's orders would have happened anyway, losing it costs you few sales.

How do you set each partner's rate yourself?#

Six steps, each with an owner and a catch.

  1. Agree the ceiling with finance. The most you'll pay for an incremental order: first-order gross profit, plus any repeat value finance will count. Use your own margin; listed US apparel companies averaged 56.88% gross margin in NYU professor Aswath Damodaran's January 2026 data, general retailers 33.18%. [Aswath Damodaran, NYU Stern, Jan 2026] (opens the source) The catch: repeat value invites optimism, so agree the payback period in writing.
  2. Price each partner's orders. The affiliate manager pulls them from the network export with the code used, and adds the discount and fee to the commission. The catch: the network only knows the code and customer status your tag sends, so the developer may have to add them.
  3. Measure the incremental share. Hold shoppers back from the partner's pop-up, or switch the partner off in one region; the partnership review guide sets out both. The catch: a storefront holdout measures the in-cart part of the relationship, not shoppers who start on the partner's site.
  4. Set each rate below its break-even, as in the example. The catch: when the code alone costs more than the partner's incremental orders earn, no rate fixes it; renegotiate the code.
  5. Build it in the network and the terms. Rules for new and returning customers, code-level rates, an approval period matched to your returns window, and written rules on rate changes; the affiliate agreement guide covers the clauses. The catch: plan limits and the network's rate rules apply whatever your terms say.
  6. Review quarterly against cost per order, incremental share and cost per incremental order. The catch: the share moves with your own marketing, so re-test after a big change.

How Benson helps you set and defend each rate#

For the coupon and cashback extensions in your programme, Benson supplies the inputs a network report can't, and the evidence to take to the partner. The steps are in the partner-deal playbook.

Record the deal, read how often the partner joins after the cart, then keep, renegotiate or cut.
  • Record the deal. Tell Benson which extension you pay, how and at what rate. It stays private to your team, and every estimate uses your rate.
  • Attribution integrity (Stop and above). Benson notices when the affiliate id, click id or UTM source a shopper arrived with changes after the cart, which extension it detected in the same session, and the estimated commission that moved at your configured rates, including commission a content partner lost to a later click.
  • The deal readout. The partner's page shows its orders, the commission you paid and how often it joined after the shopper reached the cart: the case for a lower rate on those orders, or a new-customer rate.
  • A holdout for the incremental share (Stop and above). Block hides the partner's pop-up on your storefront for most extension shoppers and leaves a random slice unchanged, so the denominator comes from your own traffic.
  • Checkout code policies (Stop and above). On Shopify, no app can change what an extension shows inside checkout, so Benson works on the codes: with the strict code policy on, only allowlisted codes work in coupon-extension sessions. Allow the partner's own code and its shoppers keep their discount, so the discount in your sums is the one you agreed.
  • The network report. A printable list of late-funnel overwrites by network, with your rates and the method stated, for your network contact.
Sample store
The network report: late-funnel overwrites by network, with the rate and method shown. Sample data.

Seeing which extensions are in your sessions is free on every plan; Attribution starts on Stop. Before the next rate review, see Benson on your store.

Questions merchants ask

What is a good affiliate commission rate?
One that costs less per incremental order than that order is worth to you. Published averages come from platforms' own tracked sales and vary widely by partner type, so use them to see what the market pays, then set each partner's rate from the cost of its own orders.
Should coupon and cashback partners get a lower commission?
Often, once you count the discount on their codes and the orders they join late. A lower rate on returning customers, or on codes you didn't issue to them, is usually fairer than a flat cut, and a holdout tells you how far to go.
Should you pay affiliates on returning customers?
You can, deliberately, at a rate you've chosen for retention. Send customer status on the conversion tag and set separate commission groups or payout rules, so returning orders don't earn an acquisition rate by default.
Should CPA be a percentage or a flat amount?
A percentage keeps commission in line with the basket and suits most stores. A flat amount suits ranges where margin doesn't rise with price, or paying a set amount for each new customer.
How much notice do you have to give before lowering commission?
It depends on your network and contract. Awin's help centre asks for at least seven days' notice and limits both the size of each cut and how often you can make one. Put your own notice period in your programme terms as well.
What is a tiered commission?
A rate that rises as a partner's sales in a period pass set thresholds, not commission on a second level of affiliates. Volume tiers tend to reward the partners nearest checkout, so if you tier, tier on new customers.

Sources#

Keep the decision moving

Find out what's leaking. 14 days of Own, no card.

Start free trial