Affiliate and Partnerships
Date: 2026-08-16
Paying a commission for a referred sale. The model looks risk-free — you only pay on conversion — and that structure is exactly what makes it so vulnerable to paying for sales you’d have had anyway.
What it is
Affiliate marketing pays partners a commission on sales they refer, usually tracked by a link and attributed on last click within a window.
The appeal is genuine: no upfront cost, payment only on outcome, and it scales without headcount.
The structural problem
Commission is paid on attribution, and attribution rewards the last touch. So the tactics that win affiliate commission are the ones closest to the purchase — which are the ones least likely to have caused it.
The recurring pattern:
customer decides to buy
→ searches "yoursite discount code"
→ lands on a voucher site
→ clicks through, buys
→ voucher site earns commission
what the affiliate contributed: a code the customer
went looking for after deciding
Voucher and cashback sites intercept demand at the moment of purchase. They’re not wrong to exist — they serve a real customer need — but paying an acquisition commission for them is paying for a conversion you already had, and often paying a discount on top.
This is last-click attribution failing in a way that costs actual money, because commission is real spend rather than a reporting artefact.
The types, and how they differ
| Type | Typically | Incrementality |
|---|---|---|
| Content and review sites | Genuine discovery | Often high |
| Comparison and aggregators | Consideration stage | Moderate |
| Loyalty and cashback | Intercepts at checkout | Low |
| Voucher and code sites | Intercepts at checkout | Usually very low |
| Brand partnerships | Genuine reach | Depends entirely |
Commission rates are usually flat across all of these, which means you pay the same for demand creation and demand interception. That’s the thing worth fixing first.
Testing it
The same method as everywhere in this domain, and it’s rarely applied here:
- Suppress a partner type — voucher sites, say — for a period or in a set of regions, and measure total orders rather than affiliate orders. If total orders hold, the commission was buying nothing — Geo Holdout Tests, Incrementality Testing
- Remove the code field from checkout in a test. Watch conversion and watch how many customers leave to search for a code. This is a genuinely informative and cheap test
- Check the path. Affiliate conversions with a very short time from click to order are usually interceptions rather than referrals
Managing it properly
- Tier commission by partner type. Content and discovery partners earn more; voucher and cashback earn less or nothing. This is the single highest-value change and it’s a policy decision, not a technical one
- Set attribution rules that don’t reward last click alone. Many networks support de-duplication against other channels
- Control code distribution. Codes intended for a specific audience end up on aggregator sites within hours, which turns a targeted offer into a public discount — Discounting Strategy
- Watch the code field in checkout. An empty code box is an invitation to leave and search, and that’s a conversion leak with a commission attached — Checkout Design
- Monitor for brand bidding. Affiliates bidding on your brand terms are competing with you for traffic you’d have had, then charging you commission for it
The honest position
Affiliate can be genuinely valuable — content partners reaching audiences you can’t, and genuine discovery. It’s also the channel where the payment model and the attribution model combine to reward interception, and the only way to tell the two apart is to withhold and measure.
Judge it on incremental orders and on contribution after commission and any code discount — not on the network’s reported revenue — Contribution Margin.