In the ~$1.4M-a-year affiliate program of a DTC brand we worked with, affiliate ROAS read $8.3 per $1 in the brand's MTA, far above Google PMax (~$4) and non-brand search (~$3) on the same basis and beaten only by brand search, which is why the instinct was to keep pushing budget into it. Setting aside clicks made within 30 minutes of checkout took it to ~$5.6, and paying those clicks a coupon rate instead of a media rate would save ~$510k a year.
The brand, masked at its request with the figures rounded, sells a considered purchase at ~$300 an order and spends ~$39M a year on paid media. Its affiliate program is mostly media partners (large general-interest publishers, product-review sites, deals and roundup pages, a sub-affiliate network and two small blogs), plus one coupon site paid a coupon rate. To check what the two tests catch, we went past the platform's report to click and order timestamps, landing pages, the post-purchase survey and an MMM, which between them show whether a buyer had already decided before they clicked.
The clearest way to see it is one publisher's two kinds of page, paid at the same media rate. On the brand's largest partner, ~28% of spend, review articles that link to a product page ran 32% of paid orders within 30 minutes of the click with a median of ~11 hours from click to order, and its discount-codes page, which links to the homepage, ran 81% with a median of ~12 minutes.
Same publisher, same media rate, opposite behaviour
- Paid orders, click ≤30 min
- 32%
- Median click to order
- 11 hrs
- Share of the publisher's credit
- 67%
- What to pay
- Media rate
- Paid orders, click ≤30 min
- 81%
- Median click to order
- 12 min
- Share of the publisher's credit
- 33%
- What to pay
- Coupon rate
Why the affiliate platform and the MTA agree
They land in the same place for different reasons, and partly by coincidence. The affiliate platform is last-click with its own cookie tracking, which is weaker than the brand's, so it reliably catches the coupon click at checkout and loses many buyers who read a review and come back days later. Its own report said ~$8.7 per $1. A first-party MTA tracks the journey better and gives real credit to the long-tail editorial partners that start journeys, which the platform under-counts.
But the MTA still rewards the last touch, because closing the journey is always part of how it splits credit, and it can't tell a genuine closer from a coupon grab 10 minutes before checkout. In this program 27% of media-partner orders came through clicks like that, and in the post-purchase survey those buyers mostly named search, social or a friend as how they found the brand, not the partner. That gap, between closing a journey and collecting credit for a click minutes before checkout, is the flaw in MTA this post is about, and changing the attribution model doesn't close it.
The useful question is not whether an affiliate's clicks convert, it's whether they arrived before or after the customer had decided.
The 30-minute test
For each partner, take its paid orders and measure the share where the affiliate click came 30 minutes or less before checkout, and the median time from click to order. The coupon site sits at 81% within 30 minutes with a median of ~12 minutes, because people arrive with the cart already built. The product-review sites sit at 35-39% with medians of ~4-9 hours, because people read, leave and come back. Most partners fall clearly on one side, and the ones in between are usually mixing both kinds of page.
The two dedicated deals and roundup pages are the expensive finding. Both are paid at media rates (12% and 11%) and both behave like the coupon site, at 78% within 30 minutes with a median of ~12 minutes. On reported ROAS you could not tell them apart from the review sites ($7.8 and $8.4 per $1 against $7.1-7.5).
The 30-minute test, partner by partner
| Partner archetype | Spend | Reported $/1 | Click ≤30 min before order | Median click to order | Lands on product page | What to pay |
|---|---|---|---|---|---|---|
| Coupon / cashback site | 1% | $46 | 81% | 12 min | 12% | Coupon rate (already) |
| Deals/roundup page A | 13% | $7.8 | 79% | 13 min | 14% | Coupon rate |
| Deals/roundup page B | 4% | $8.4 | 78% | 12 min | 14% | Coupon rate |
| Sub-affiliate network A | 7% | $8.6 | 67% | 16 min | 35% | Coupon rate |
| Large general-interest publisher A | 28% | $7.2 | 52% | 27 min | 57% | Split by page (SubID) |
| Its review pages (to product page) | 16% | $8.3 | 32% | 11 hrs | 88% | Media rate |
| Its deals pages (to homepage or sale) | 12% | $5.6 | 81% | 12 min | 9% | Coupon rate |
| Small blog B | 2% | $5.9 | 43% | 2 hrs | 72% | Split by page (SubID) |
| Product-review site C | 4% | $7.5 | 39% | 4 hrs | 76% | Media rate |
| Large general-interest publisher B | 11% | $8.2 | 37% | 6 hrs | 79% | Media rate |
| Product-review site B | 9% | $7.4 | 36% | 7 hrs | 82% | Media rate |
| Product-review site A | 18% | $7.1 | 35% | 9 hrs | 84% | Media rate |
| Small blog A | 3% | $7.9 | 34% | 10 hrs | 83% | Media rate |
| Sub-affiliate network BTracking flood | 0% | $74 | 0% | 23 days | n/a | Pause and ask |
A rising affiliate budget behind a flat reported ROAS is the tell: most of the year's growth in commission came from existing partners' deals pages (~62%) and rate increases on the same placements (~20%). The four partners given a raise drove ~12% more clean orders while their commission rose ~65%, because more money on the same articles buys no new reach.
How often each channel opens the journey
Multi-visit journeys only (2+ visits): share of each channel's MTA credit that comes from the first visit (the big number), with middle and last visits muted.
If the platform won't pass the page, your own first-party pixel records the landing URL and referrer, which is enough to run both tests. Across the program, coupon-like clicks grew from ~14% of the credit in Q1 to ~42% in Q4, so the reported number only drifted from ~$8.9 to ~$8.1 while the clean number fell from ~$7.6 to ~$4.5.
The tracking-flood signature
One partner had a different problem. It reported $74 per $1 through 6 click IDs spread across ~8,200 browsers, 98% of its visits carried a Google ad click ID, and none of its paid orders came within 30 minutes of the click. That is what it looks like when a partner bids on your brand name and routes the clicks through its link. At 0.5% of spend it barely moves the channel, but I would pause it and ask before the next invoice.
The adjusted number is still an MTA number
~$5.6 is the best number an MTA can give you, not proof the channel is incremental. A standard weekly MMM put affiliates at ~$2.7 per $1 with an 80% range of ~$0.6 to ~$5.5, wide because the channel is under 4% of spend. A post-purchase survey implied ~$3.6, with 88% of buyers who named a partner publication coming through review pages, so the honest statement is a range of ~$2.7 to ~$5.6, not the reported $8.3.
What to do with it
Each step changes either what you pay a partner or how much budget the channel gets.
Pay coupon-like traffic at coupon rates. Ask the platform for a SubID per page so review pages keep the media rate and deals pages get the coupon rate, and where a partner can't split, set the rate by its 30-minute share. Here that saves ~$510k of ~$1.4M a year.
Pause partners with the tracking-flood signature and ask about brand bidding before paying again.
Hold the channel's budget at the adjusted number. Before approving a rate increase, check whether the partner's clean orders grew or only its commission did.
Chase more review and "best of" placements that link to product pages. That is the part of the program that opens journeys, and where a higher rate is worth negotiating.
Fix the reporting so everyone sees the adjusted number. Flag a click as coupon-like when
click_to_order_minutes <= 30and it lands on the homepage or a sale page, and show reported and adjusted side by side.
If your affiliate ROAS looks better than everything except brand search, the 30-minute test is the quickest way to find out why. We rebuild channel measurement for DTC brands, and I'm happy to walk through how we'd approach it for your setup.
Common questions
Should coupon sites get affiliate commission?
Yes, at a coupon rate, typically 1-3%, because most coupon clicks come from buyers who had already decided. In this program, 81% of the coupon site's paid orders came within 30 minutes of the click. The bigger cost is usually media partners whose deals pages behave the same way but are paid 11-12%.
What is a good affiliate ROAS?
Compare affiliates to other channels on the adjusted number, not the reported one. For this brand, the MTA said $8.3 per $1, the 30-minute and landing-page checks brought it to ~$5.6, and an MMM and a survey put it at ~$2.7 to ~$3.6.
How do I tell if an affiliate is bidding on my brand?
Look for a handful of click IDs spread across thousands of browsers, almost every visit carrying a Google ad click ID, a return far above the rest of the channel, and almost no orders within 30 minutes of the click. Pause the partner and ask before paying the next invoice.
We rebuild channel measurement for DTC brands, from click and landing-page capture to adjusted affiliate reporting your team and your partners can both check against the order data.
See how it works →Figures are from one brand's affiliate program over 12 months, rounded, with the brand and its partners masked at the brand's request.