How do I find out if creator sales are incremental or would have happened anyway?

An affiliate programme usually credits a purchase to the creator whose link was clicked last or whose code was used. That tells you who was involved in the sale. It does not tell you if the sale would have been lost without the creator. Two examples: a regular customer has already filled her basket, sees a story with a link sticker in the evening and orders through the link. Or someone looks for a discount code just before paying and finds a creator’s code on a coupon site. In both cases you pay commission on a purchase that would have happened anyway.

It works the other way round, too. Someone watches a video, later searches for your shop on Google and buys there without a link or a code. The creator triggered that sale, but your affiliate programme never sees it. Incremental therefore means revenue that would not have existed without the campaign, measured across all sales of the promoted products. Google researchers noted in a 2011 paper that clicks and conversions do not show how people would have behaved without the advertising.

For ads, there are tests with a control group. In Google Ads Conversion Lift, which according to Google Ads Help isn’t available for every account, a treatment group sees the ads and a control group does not, and according to the same help page the difference in conversions is the lift. You can’t split a community like that: someone who follows a creator either sees the post or doesn’t. So you need a different comparison, over time, across products, across regions or across groups of creators.

A comparison region is the idea behind the geo experiments Google researchers described in 2011. Regions are randomly assigned to test or control, and before the test there is a period in which every region runs the same way. The precondition there is that ads can be served by region. Creator posts, by contrast, usually reach the whole community wherever it lives. This route only fits if a creator’s community is clearly concentrated in one region, such as one country, and you have separate sales figures for it.

The new-customer share shows how many attributed orders come from people who have never ordered from you, or not for a long time. If almost all orders through a creator come from your regular customers, a large part is probably sales you would have made regardless. The figure can’t prove it, though: a first order might also have happened without the creator. Two more signals work without a link: a question after checkout (“How did you hear about us?”) and searches for your brand name during the campaign. Both show effects that links miss, but neither proves anything on its own.

In a staggered start, one group of creators is a holdout at first, meaning a group that does not take part yet. That takes more effort than a simple before and after comparison, but it gives you two measurements instead of one. A holiday week or a newsletter can then hardly explain both jumps in revenue. Google researchers described a similar idea for geo experiments in 2012: regions rotate between test and control across several test periods.

The test stands or falls on creators posting at a known time. An affiliate programme alone doesn’t make that happen. Many partners in a programme never become active, and the creators you want most rarely post without a reason. You can only steer the start through a reason you provide, such as samples, a brief, an agreed date or a paid campaign. So ask more creators than you need for two groups.

1. Set the metric: net revenue after returns from all orders of the promoted products, across all channels. Not just revenue through creator links, because incremental sales only show up in total revenue. If you also sell through retailers or marketplaces, their sell-out belongs in it as far as you get the figures.
2. Measure the baseline: weekly figures for the last eight weeks and the same weeks last year. Note how much the weeks vary without a campaign. If they range between €4,600 and €5,400, say, a campaign that adds €200 a week gets lost in the noise.
3. Choose comparison products: similar products with the same seasonality that are not part of the campaign. They show how revenue would have developed without the campaign.
4. Draw lots for the creators: take the creators who have agreed to take part in the campaign. Anyone who already features your products on their own stays out, because you can’t hold them back. Sort the rest by reach, form pairs and draw lots in each pair for who goes into group A. If you start the strongest creators first, you can’t tell later if a difference came from the timing or from the creators.
5. Start at different times: group A gets samples and the brief in week 1, group B in week 4, for example. You control the start through shipping and an agreed date. If someone in group B still posts early, record when. Tell everyone in advance why there are two start dates.
6. Keep everything else the same: no new discount campaigns, ad budgets for the products unchanged, enough stock in every variant. If anything else happens during the test, such as a newsletter, note it down with the date.
7. Evaluate once returns are complete, and by the draw, not by who actually posted: a creator in group A who doesn’t post stays in group A. If total revenue rises at each start by roughly what is attributed to that group, the revenue was more likely incremental. If only attributed sales rise and total revenue stays flat, the links mostly picked up sales you would have made regardless.

• Season and dates: Black Friday, Christmas, school holidays or the start of the month, just after payday, can change revenue even without a campaign.
• Parallel discounts: a sale in your shop or a newsletter voucher in the same week makes the campaign look better than it is.
• Ads at the same time: if ad spend rises during the test, the two effects mix. The same goes for retargeting, meaning ads for people who have already visited your shop, for example through a creator link.
• Stock and price: sold-out variants push revenue down, and a lower price in other shops pulls purchases away.
• Follow-up period too short: in Germany, private buyers can usually withdraw from an online purchase within 14 days of receiving the goods and then have up to 14 days from the withdrawal to send them back (§ 312g(1), § 355(2) and (3), § 356(2), § 357(1) BGB). If you evaluate earlier, you count revenue that will partly be refunded. A longer return policy in your shop pushes the date back further.

All values are made up, and the example uses the baseline period with comparison products. Assume the promoted products brought in an average of €5,000 net revenue after returns per week in the eight weeks before the test. The comparison products grew by 10% during the four-week test period, which is the season. Without the campaign you would therefore have expected around €22,000.

Incremental ROAS (iROAS) divides only the extra revenue by the cost. Google Ads calculates it the same way in Conversion Lift: incremental conversion value divided by total spend. With a ROAS of 5.0, the campaign in the example looks strong. But incremental revenue is only €5,000, equal to 42% of the attributed revenue.

Your break-even ROAS tells you if that pays off. It is the ROAS at which the campaign just covers its cost: 1 divided by your contribution margin ratio. That is the share of net revenue left after cost of goods, shipping and payment fees. At an assumed 40% the break-even ROAS is 2.5, and the iROAS of 2.08 is below it. The €5,000 of extra revenue brings €2,000 in contribution margin, while the campaign costs €2,400. At €50 per order, that is 100 extra orders at €24 each, although each one only brings in €20.

Put every rule into your terms before the start, with a definition. For example: a first order is an order from an email address that has not been used to order in the last 12 months. Rules that only appear afterwards cost you the creators’ trust.

Run the numbers with the example above. Assume 40% of the €12,000 came from first orders. If you pay 20% on those and 5% on the rest, commission costs €960 + €360 = €1,320 instead of €1,800, or €1,920 with samples. If the extra revenue stays the same, the iROAS rises to €5,000 ÷ €1,920 = 2.60 and sits above the threshold of 2.5. Only the next test shows if creators take part just as much at these rates.

With wyrld Boost, you put your budget behind creators who feature your products and give them a reason to post that an affiliate programme alone does not. You set the products, the time frame and the KPIs the payout is based on, such as purchases in a given week, and you decide where people buy, for example only in your online shop. For measurement, that means the campaign has a fixed start and end, and with that choice the purchases from the campaign show up in your own shop data. Leave enough comparison weeks before the time frame, and the baseline period with comparison products from above fits well. A campaign is meant for many creators at the same time. Taking part is up to each creator, so how many join is not fixed in advance. Creators label campaign videos as advertising, and after the campaign you see which creator delivered which results.

If you have an affiliate programme with one of the major affiliate networks, wyrld appears in it as a single partner, through which creators can usually already pick your products. Being available to pick doesn’t mean the creators you want will post, though. The brand and shop subscription shows you which creators behind it feature your products, who brings clicks and who sells. wyrld measures clicks through the links, and sales where the shops report them. Before a staggered start, that tells you whom to leave out of both groups: creators who already feature your products. With creators who sell, wyrld sets up individual terms, so you don’t have to pay everyone the same commission.

You pay for results on your terms: per click, per sale or both. Anyone who buys through a link to your shop orders from you, and the order, payment, shipping and customer relationship stay with you. Outside Boost, if other shops also stock your product, their offers and prices appear alongside and the shopper decides where to buy. Those purchases then won’t show up in your shop data, and for the comparison you need those shops’ sell-out figures, as far as you can get them. Prices for the subscription and Boost are available on request. If you are planning a test with creators, wyrld is happy to work out with you what fits first: the connection, the subscription or a Boost campaign.