Which affiliate partners should you pay, and for what?

Pay each affiliate partner for the orders it adds: sort partners by where they join and who they bring, price their real cost, then set the rates.

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

In short

Pay affiliate partners for the orders they add, not the orders they reach last. Sort partners by where they join the journey, before or after the cart, and whether they bring new or returning customers. Then price each one's commission, network fee and code discount against the share of its orders a test shows it added, and set separate new and returning rates.

It's your third week as head of ecommerce, and the affiliate programme came with the job: a network contract signed years ago, a few dozen active partners, rates nobody has touched since launch, and a monthly network report.

The report is tidy. A handful of partners take most of the commission: two cashback sites, a coupon site, a browser extension and one review site. Return on spend looks excellent, and the account manager suggests a higher rate for the top partner.

Finance reads the same month differently. Their margin view shows discount spend creeping up, and when they matched the affiliate orders against the customer file, most of the buyers had ordered before. One report says the partners found those sales. The other says the partners were standing at the till.

You don't have to pick a side yet. You need a map: which kinds of partner the programme pays, where each joins the shopper's journey, what each does for the business, and what each should be paid for. This guide is that map, with the network settings that build it and links to the deeper guides on rates, creator codes and agreement terms.

What kinds of affiliate partners are there?#

Sort them by when they meet the shopper, and by which customers they bring. Some join before the cart, while the shopper is still deciding: content and review sites, comparison services, creators. Others mostly join after it, when the basket is full and the shopper looks for a better price: coupon sites, cashback and loyalty sites, and the browser extensions many of them run. The first group tends to reach people who don't know you yet; the second, people who were already buying.

More of the money goes to the second group.

35%

of US affiliate spend in 2024 went to cashback, loyalty and rewards publishers, the largest single category in the Performance Marketing Association's industry study.

Affiverse (PMA study), Jun 2025(opens the source)

Coupon and voucher sites fell from 16% to 10% of spend since the PMA's previous study, and content affiliates rose from 15% to 16% [Affiverse (PMA study), Jun 2025] (opens the source). In total, US brands spent $13.62bn on affiliate in 2024, and the channel drove $113bn of ecommerce sales, 9.4% of the US total [Performance Marketing Association, Jun 2025] (opens the source). The PMA is a US-based trade association, and the networks that supplied the data, including Awin, CJ, Partnerize and Rakuten Advertising, are among its top-tier Industry Champion members.

Here's the map. The sections after it show what each type costs and the settings that build each row.

Partner typeWhere it joinsWhat it does for youPay it for
Content and review sitesBefore the cart, often firstIntroduces shoppers who haven't heard of youNew customers at your best rate, returning at a standard one
Comparison servicesBefore the cartWins shoppers weighing you against rivalsSales, at a rate the category's margin allows
CreatorsBefore the cart, often days earlierCreates demand you can't buy elsewhereTheir own link or code, new customers above returning
Cashback and loyalty sitesEither side of the cartBrings members, keeps customers buyingNew customers in full, returning at a lower rate, flat fees after a test
Coupon and deal sitesMostly after the cartCloses shoppers already hunting for a codeOnly codes you issued them, at a low rate on returning customers
Browser extensionsAfter the cart, same sessionApplies a code at the last stepNothing, or a tested deal with stand-down and allowed codes
Card-linked offersAt payment, in a bank's appReaches a bank's customersNew customers, after a test

Browser extensions get their own row because their rules are different, and tightening. Seven of the eight networks in the PMA's January 2026 comparison say they require extensions to stand down when another publisher referred the shopper [Performance Marketing Association, Jan 2026] (opens the source). Awin puts any publisher running an extension on Soft Click by default, so it can't override an earlier partner's click [Awin, Aug 2025] (opens the source). Since 10 June 2025, Chrome's store policy has required a user action before an extension applies any affiliate link, code or cookie [Chrome for Developers, Mar 2025] (opens the source). And fewer US programmes allowed extensions in 2026 than in 2024 [Performance Marketing Association, Jul 2026] (opens the source).

Paying one is still a choice some brands make on purpose.

Why does the network report favour the partners closest to checkout?#

Because the report applies a payment rule, and the rule is last click. Almost 80% of US programmes in the PMA's 2026 brand survey still use last-click attribution, and most rely almost entirely on the network's own reporting [Performance Marketing Association, Jul 2026] (opens the source). Whoever is nearest the payment button gets the credit, whoever started the journey.

Half

the transactions across 2,368 North American retail brands on impact.com in 2025 went to loyalty and rewards partners, from 15% of clicks and for 33% of brand spend.

impact.com, Feb 2026(opens the source)

Content review partners took 24% of spend for 9% of transactions in the same data, which comes from impact.com's own platform [impact.com, Feb 2026] (opens the source).

Rankings by return on spend mislead the same way. Partnerize's Americas index for Q2 2025 shows coupon partners returning $20.68 in sales per dollar of commission against $10.59 for content [Partnerize, Sep 2025] (opens the source). We checked the arithmetic: each figure is one divided by the type's average commission rate (4.84% and 9.44%). It shows coupon partners are cheaper per credited sale, not that they added the sale.

New customers are the better signal, with a caveat. In the same index, 25.7% of loyalty partners' revenue came from new customers and 30.5% of coupon partners', against 45.9% for content [Partnerize, Sep 2025] (opens the source). Partnerize sells the platform and says its data isn't a proxy for the market. One quarter is a snapshot; your own split is the number to plan on.

When someone checks where credited orders started, the gap can be wide. PartnerCentric, an agency that sells attribution technology, found coupon, loyalty and cashback partners started 13% of the conversions credited to them across its clients in late 2024 [PartnerCentric, Feb 2025] (opens the source). In a test Haus published in 2023 (Haus sells incrementality testing), one brand switched off its US loyalty and cashback partners for seven weeks and revenue didn't change against comparable countries [Haus, May 2023] (opens the source). The cashback-partner guide shows how to run that test on your own biggest partner before you renew it.

Your partners will have an answer, and it isn't wrong for every order.

Both can be true of the same partner on different orders. A lapsing customer won back by a cashback offer is worth paying for; a customer who was coming back anyway and met the partner at the till isn't. So pay by the order, new or returning, your code or someone else's, rather than cutting a partner type wholesale.

What does a partner really cost you?#

More than its commission rate. Three costs sit on top: the network's own fee, the discount the partner's code carries, and commission on orders that would have happened anyway.

The fee is usually a share of the sale. Awin's US Access plan costs $49 a month plus a 3.5% tracking fee on the sale value, on top of commission; in Awin's own example, a $100 sale at 6% commission pays the partner $6 and Awin $3.50 [Awin, Oct 2026] (opens the source). The discount sits on a different line of the accounts, but you fund it on every order the partner's code touches. How deep any code should go is a discount strategy question; for a partner, it's part of the price.

Some of the cost isn't commission at all.

Nearly £1 in £5

of UK affiliate spend in 2025, out of £1.8bn in total, went outside last-click CPA into tenancies, hybrid deals and other models, according to the APMA, the UK trade body, from data eleven networks and platforms supplied.

Affiliate and Partner Marketing Association (APMA), Apr 2026(opens the source)

A tenancy or flat fee buys placement, not orders, so it needs the same test as a rate.

Put those costs against the orders a partner actually added and one flat rate can lose money on a partner that a split rate pays for.

The commission rates guide adds the discount on the partner's code to that sum, turns it into a rate for each partner, and covers the notice to give before you change one.

How do you pay new and returning customers differently?#

Tell the network which orders came from new customers, then give each group its own rate. No network can work that out alone: your developer sends the customer's status with every conversion.

The catch: "new" means whatever your developer sends. Agree the definition first, such as no previous order on that email, or none in the last year (CJ's Lapsed value exists for that case), and tell partners before their rates change.

Codes have settings too. On Accelerate and Advanced, Awin can set a rate by the voucher code on the order [Awin Help Center, Sep 2026] (opens the source), and impact.com's promo code exception lists pay different rates, including $0, on codes you choose [impact.com Help Center, Oct 2026] (opens the source). That's how a coupon site earns on the code you gave it and nothing on your staff code. A content partner whose click helped but didn't win can get a fixed amount through Awin's Payment on Influence [Awin, Oct 2026] (opens the source).

Creators are the hard row: a code can be read out in a video and a link can't, and a code travels. The creator codes guide covers who gets paid when a creator's code turns up somewhere else.

Should you run the programme through a network, a platform or in-house?#

The choice decides who finds your partners, who polices them and what you pay on top of commission. Most programmes still run on a single network or platform, though the share of US programmes on more than one rose from just under 9% in 2024 to nearly 20% in 2026 [Performance Marketing Association, Jul 2026] (opens the source).

  • A network (Awin, CJ, Rakuten Advertising) brings a publisher base, payments and its own compliance rules. Awin publishes its prices: in the US, Access at $49 a month plus 3.5% of each sale, or Accelerate from $99 a month plus 2.5% [Awin, Oct 2026] (opens the source); in the UK, Access at £99 plus VAT a month plus 3.5% [Awin, Oct 2026] (opens the source). The catch: the fee rises with your sales, and the network's report is also the record you pay against. Networks change, too: Awin closed ShareASale on 6 October 2025 [Awin, Aug 2025] (opens the source).
  • A platform (impact.com, Partnerize) is software for tracking, contracts and payments. impact.com starts at $30 a month plus 2.5% of partner-driven transactions [impact.com, Oct 2026] (opens the source). The catch: you recruit, vet and manage every partner.
  • In-house, through an affiliate app or Shopify Collabs, costs least. Collabs charges a 2.9% processing fee on each automatic commission payment and holds commission for 30 days by default [Shopify Help Center, Apr 2023] (opens the source). The catch: every partner is one you found, and Collabs isn't accepting new creator sign-ups at the time of writing, though merchants can still invite creators [Shopify Help Center, Oct 2026] (opens the source).
  • An agency runs any of the three for a fee. A 2015 study of 4,523 merchants on the three largest US networks, using 2012 data, found programmes managed by the network let in more questionable affiliates overall than in-house ones, mostly in grey areas such as typosquatting, while agency-run programmes did at least as well as in-house teams [Journal of Marketing Research (Edelman and Brandi), Feb 2015] (opens the source). The catch: the partner list is still yours to answer for.

What should you change first in a programme you've inherited?#

Five steps, in order.

  1. Sort every paid partner by where it joins. Export last quarter's orders by publisher, match them to your landing sources and customer file, and mark each partner before or after the cart, mostly new or mostly returning. The affiliate manager owns this. The catch: analytics rarely sees the extension itself, only the referrer change it caused.
  2. Send customer status, then split the rates. Your developer adds the status to the conversion tag; the affiliate manager builds the groups. The catch: partners on the old flat rate will notice, so give the notice your terms require.
  3. Price your biggest partners properly. Run the sum above for each: rate, network fee, code discount and the share of orders it added. The catch: that share is an estimate until you test it.
  4. Test the biggest after-the-cart partner. Switch it off in one market against a similar one, or hold out a random slice of the sessions it reaches; the holdout guide covers the design. The catch: you need comparable markets or enough traffic, and the partner has to accept the method.
  5. Write down what each partner may do. Which codes it may show, when an extension must stand down, whether it may bid on your brand name, and how it discloses the relationship; the agreement terms guide has wording. Disclosure is your responsibility too: the FTC's Endorsement Guides can hold the advertiser liable for an affiliate's missing disclosure [US Federal Trade Commission, Jul 2023] (opens the source), and in the UK the CAP Code makes the brand and the affiliate both responsible [Committee of Advertising Practice (AdviceOnline), Mar 2023] (opens the source). The catch: terms work only if someone checks them.

How Benson fits a partner strategy#

Benson covers the part of this map your analytics can't see: what happens in the shopper's session after the cart. The partner-deal playbook sets it out in order.

Record the deal, see where the partner joined, then keep, renegotiate or cut it.
  • Where each partner joined. Attribution integrity notices when the affiliate id, click id or UTM source a shopper arrived with changes after the cart, names the extension detected in the same session, and estimates the commission moved at your configured rates.
  • A report the network can check. The network report lists those late-funnel overwrites by network, with the rates and method stated, for your network contact.
  • Your deals, recorded. Tell Benson which extension you pay, how and at what rate; it stays private to your team. The partner's page shows its orders, the commission you paid and how often it joined late in the funnel.
  • One rule for the extensions you don't pay. On your storefront, from home page to cart drawer, hide pop-ups from extensions you have no deal with and let your partners through, with rules by schedule, country, traffic source, device and page. Keep a holdout, and the Value page shows revenue per session for both groups, with a range.
  • Your partner's code keeps working. No app can change what an extension shows inside Shopify checkout, so Benson's protection at checkout works on codes. With the strict code policy on, only codes on your allowlist work in coupon-extension sessions: allow your partner's own code and its shoppers keep their discount, while other codes are refused with a short message.
Sample store
The affiliate click trail: where the shopper arrived from, and where the credit changed. Sample data.
Sample store
The network report: late-funnel overwrites by network, with the rate and method shown. Sample data.

Seeing which extensions and AI agents are in your sessions, and what Benson would hide, is free on every plan. Attribution integrity, Block with its holdout, and checkout code policies start on Stop; pricing has the detail.

Before your next rate review, start with the evidence: see Benson on your store, or open the partner-deal playbook.

Questions merchants ask

What types of affiliate partners are there?
Content and review sites, comparison services, creators, cashback and loyalty sites, coupon and deal sites, browser extensions, and technology partners such as card-linked offers. The useful split is when they join: before the cart, or after it, when the basket is already full.
Should coupon and cashback partners get the same commission as content partners?
Usually not. Pay each partner for what it adds: your full rate on new customers, a lower one on returning customers, and commission only on codes you issued it. Test the biggest after-the-cart partner before you renew its rate.
Do browser extensions count as affiliate partners?
Yes, when your network pays them. They usually join after the cart, so most networks require them to stand down when another partner referred the shopper. You can pay one for a say over the codes it shows, or pay none and test the difference.
Should I use an affiliate network or run the programme in-house?
A network brings publishers, payments and compliance rules for a fee on top of commission. A platform is software, so recruiting is yours. In-house costs least, and every partner is one you found.
How do I know if an affiliate partner drives incremental sales?
Test it. Switch the partner off in one market against a similar one, or hold out a random slice of the sessions it reaches, and compare revenue per session over the same weeks. A last-click report can't answer the question.

Sources#

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