Affiliate subnetworks, and what you can check from outside

Across the brands we scanned this year — a self-selected group, not a sample of e-commerce — most publisher accounts we could identify in intercepting coupon ads sat behind a single coupon site. A small group sat behind many. That is also the shape of affiliate subnetworks, and legitimate subnetworks and link-monetisation services are built that way. The structure alone proves nothing.

It matters for one reason: your network pays the account, not the domain. Where one account sits behind several coupon domains, terminating one leaves the account, and the rest, in place.

Anatomy of a Brand-Bidding Operation showed the extreme case: one affiliate ID behind 1,000+ subdomains. This is the register-wide view, and a case for adding one thing to subnetwork disclosure: a per-account declaration, keyed to the identifier in the click link, dated, and testable from outside.

I run AdCrime, and checking click links from outside is close to what we sell, so weigh the argument accordingly. The part that matters most needs no software: one line in a declaration, keyed to the right identifier. Draft that line loosely and no scan rescues it.

How affiliate subnetworks work

The Affiliate & Partner Marketing Association (APMA), the UK affiliate trade body, defines a subnetwork in its draft code as "an organisation aggregating multiple publishers under one account". The network sees one publisher; the subnetwork collects the commission and distributes it among the sites beneath it.

So the identifier in a click link belongs to the account, not the site you clicked from. The site's own identity, if passed at all, travels in a separate field and stays in private reporting.

Read the limits first

Our register is self-selected: brands that asked us for a free scan, brands on rosters we chose to scan, and brands we picked. It says nothing about how common this is across e-commerce, or on your programme.

Everything here comes from public Google results, followed down the redirect chain. Between 19 February and 25 September 2026 we captured 71,977 paid Google ads, 71,482 of them on branded coupon searches such as "[brand] promo code". Scanning was largely paused between 6 and 24 September.

We kept only confirmed or likely affiliate interceptions with a publisher account identifier readable in the click link, excluding the 14 largest mainstream voucher aggregators and brands' own domains.

The distribution

We read identifiers for 1,721 distinct publisher accounts, across 783 coupon sites.

One account behind one site is the normal shape in our register, and a rule written for the tail shouldn't burden those accounts.

The tail is small but concentrated. Of the 2,161 account-to-site relationships across the 1,721 accounts, the 22 accounts behind 5 or more sites hold 217 (10.0% of 2,161).

The five widest accounts

We print no identifiers, not even partially.

"Mostly seen on" is the network label on most, not all, of an account's captures; the bracket shows how many. A capture is one recorded ad. Last-seen dates predate the September pause.

Nothing here shows that any account holder knows what the sites under it do. Each was positioned to collect the commission when a click through its sites converted, and the same structure fits a legitimate subnetwork or link-monetisation arrangement.

Domains go, accounts stay

Terminated the domain. Kept paying the account.

A terminated domain comes off your list, and your case log shows a resolution. Where the account sits behind other domains too, the payment relationship hasn't moved.

That doesn't make domain terminations pointless. It makes them a different action from enforcing at the account rather than the domain. If you're chasing sub-affiliate brand bidding, domain-by-domain termination may never reach the account.

What networks already require

Sub-affiliate network transparency is not a new ask.

• Awin's Subnetwork Code of Conduct requires subnetworks to use the clickref "to identify the Subpublisher responsible for a sale/lead" and the URL field for "the last referrer Subpublisher URL". It holds them responsible for subpublishers "as if those actions or omissions were their own", and lists "Brand bidding/ad hijacking" as non-compliant.
• Rakuten Advertising's sub-network guidelines say sub-networks with websites should "prepare a document for advertisers that lists all the websites they use to drive traffic", and warn: "If you or a member of your sub-network is non-compliant, all traffic on your account may be stopped." That is account-level enforcement, in writing.
• impact.com offers a Sub ID field for the sub-affiliate's ID; we found no published requirement to use it.

All of it is private. Ask for it — but none of it gives an outsider a declaration to test against the identifier in a click link.

The APMA subnetwork code

The APMA published a draft subnetwork code of conduct (version 1.1) on 31 March 2026 and consulted on it until 30 April. It already does most of the work. Clause 1 says "Subnetwork activity must be fully transparent at the publisher level", with the publisher's name, domain or application, contact details and traffic source disclosed "upon request or by default". Clause 7 adds domain-level information and reporting broken down by day; Clause 9, cooperation with "reasonable audit or information requests from Advertisers or Networks".

Kevin Edwards, the APMA's founder, wrote on 1 September that the APMA is launching the code this autumn and asking subnetworks to sign up: "We'll then audit them to ensure they're compliant before adding them to a preferred supplier list." He added that "this is not about punishing good subnetworks; in fact it will elevate them".

We have no relationship with the APMA. Our suggestion is narrow: make the draft's domain disclosure testable by someone other than the subnetwork.

Checkable from outside

Every ad on a branded coupon search has a click link anyone can follow. At the network hop, where it is readable, you'll find the publisher account identifier: the account the network pays. No cooperation from the subnetwork, network or site is needed. See how to read an affiliate redirect chain and where the publisher ID sits in an Awin link.

What it can prove

• That a domain's ad, in a given market on a given date, resolved to a given account identifier.
• That one identifier sits behind several domains.
• That a domain you terminated turned up again under the same account.
• That an undeclared domain resolved to a declared account. Under-declaration is provable from outside.

What it can't prove

• That a declaration is complete: an undeclared domain may not have appeared on your searches.
• Whether the account holder knew about, approved or controls the site.

So the outside check is a falsifier, not a certifier: it can show a declaration is wrong, never that it is right.

What a declaration needs

Clause 1 already puts domains on the table. To make that list testable from outside:

1. Key it to the identifier in the click link, not a partner name or CRM reference. Without that key the list can't be joined to anything observable. This is the drafting choice that matters most.
2. One declaration per account, per network. An identifier means nothing outside its own network.
3. Domains in full, not brand names or general descriptions.
4. Markets named. A domain declared for one market and found in another is a finding.
5. Dated and versioned, with a duty to refresh. An undated list can't be failed.
6. A stated consequence for a discrepancy. The draft says the APMA "is not a regulator and does not enforce commercial decisions", so the consequence must be one advertisers and networks can act on.

You needn't wait for the code. Write the same terms into your own programme when you approve a subnetwork.

Check your own programme

None of this needs a tool.

1. List your subnetworks and multi-site partners with each one's publisher account identifier.
2. Run branded coupon searches, such as "[brand] promo code", in your markets. Follow each coupon site's ad down the redirect chain; note the domain and identifier.
3. Group the results by identifier, not domain. Does any account sit behind more than one site on your programme?
4. For subnetworks, compare against what they've given you: Rakuten's website list, or Awin's clickref and URL data. A domain under their account but missing from their list is a question for them.
5. When you next terminate a domain, ask the network which account it sat under and what else that account runs on your programme.

An account behind several sites isn't proof of wrongdoing. It shows that your termination list and your network's payments are keyed to different things.

If you'd rather not group by hand, the first scan is free.

Mario Vaher is the founder of AdCrime, which takes public Google results on branded coupon searches, follows each ad down the redirect chain to the publisher account behind it, and packages the result as evidence you can file with the network.