An affiliate report tells you how many sales a publisher was credited with. It does not tell you how many of those sales would have happened anyway. The difference is incrementality, and it is the most useful question to ask of any performance channel.
Why reported sales overstate impact
Most programmes credit the last click before purchase. A shopper who had already decided to buy, then searched for a coupon or clicked through a cashback app, is recorded as an affiliate sale. Some of these sales are truly influenced by the publisher. Others are not.
Signals you can check without a test
- New-customer share: a publisher bringing mostly first-time buyers is more likely adding sales.
- Time from click to order: clicks seconds before checkout suggest the shopper was already buying.
- Order value: compare with your site average; higher baskets suggest real influence.
- Overlap with other channels: how often the same order also had a paid-search or email click.
Simple tests
- Pause test: pause a publisher or a publisher type for a set period and compare total sales, not only affiliate sales, with a similar period.
- Geo test: run the activity in some regions and not in others, and compare.
- Rate test: change the commission or cashback rate and see whether total orders move.
Keep other marketing steady during a test, and run it long enough to smooth out normal weekly swings.
Using what you learn
Incrementality is rarely all or nothing. The practical step is to pay more where partners add sales and less where they mostly capture them: higher rates for new customers and for content that introduces your brand, lower rates for last-moment clicks.
In short: judge publishers by the extra sales they bring, not the sales they are credited with. Check new-customer share and click timing, test by pausing, and set payouts to match.