How do you measure influencer marketing ROI accurately?

Measure the change in total company orders during the activation window rather than the orders tagged to the creator's code or link. Attributed orders answer the question of which touchpoint got credit, not which orders would have been missing without the partnership. The most practical approach for most brands is a geographic or time-based holdout combined with a baseline comparison: establish the expected order volume from the surrounding weeks, run the activation, and treat the lift above baseline as the incremental result. Then evaluate that lift on contribution margin after the discount, the commission, the product cost, and the shipping cost.

Why is affiliate ROAS misleading?

Affiliate platforms credit any order where the code or link was used, which systematically captures demand that already existed. Coupon sites, browser extensions, and loyal customers searching for a discount before checkout all intercept orders at the last moment and claim them. The result is a reported ROAS that can look extremely strong while the underlying revenue would have arrived anyway. The error compounds because the affiliate order also carries a discount, so you pay twice on a sale you already had.

What is a good commission rate for an affiliate or influencer program?

Set it from contribution margin, not from an industry benchmark. Start with the gross margin on a typical order, subtract the customer discount the creator offers, subtract shipping and payment processing, and the remainder is the total budget available for commission and profit combined. On a 60 percent gross margin product with a 15 percent customer discount, the fully loaded room is often under 20 percent of revenue, which means a 20 percent commission rate is a break-even or loss-making order even before considering whether the sale was incremental.

Influencer and Affiliate ROI: Measuring What Was Incremental

Attributed revenue vs. caused revenue

Every affiliate and influencer platform reports the same thing: revenue where the creator's code or link was present in the customer's journey. That is an attribution statement. It answers "who gets credit," which is a billing question, not a business question.

The business question is causal: how many orders would not have happened if this partnership hadn't existed? Those two numbers are never equal, and in affiliate they can differ by a factor of two or more — because the channel's structure actively attracts orders that were already in flight.

Consider what a discount code does. It sits in a public place, indexed, searchable, and permanently available. A customer who found you through a Meta ad, browsed for four days, added to cart, and then opened a new tab to search "[brand] promo code" before checking out will find a creator's code and use it. The affiliate platform records a conversion. The creator gets a commission. The customer gets a discount. And nothing about that order was caused by the creator — it was intercepted at the final step of a journey your paid media already paid for.

Affiliate doesn't have an attribution problem so much as an interception problem. The channel is structurally positioned at the last moment before checkout, which is exactly where demand created elsewhere passes through.

This is the same fundamental error that makes platform-reported ROAS untrustworthy in general — credit assigned to the touchpoint closest to the purchase rather than to the spend that generated it. The broader version of this failure, and why it distorts your blended acquisition cost, is laid out in why your ROAS is lying about rising CAC. Affiliate is the most acute case because the interception is built into the mechanism.

The discount-plus-commission stack

Here's what makes affiliate misattribution more expensive than the paid-social version: you pay for it twice.

When Meta over-claims a conversion, you've wasted the ad spend but the order arrives at full price. When an affiliate over-claims a conversion, the order arrives at a discount and you owe a commission on the discounted revenue. A customer who would have paid $100 now pays $85 and costs you an additional $12.75 in commission. You've converted a $100 order into a $72.25 order to pay for a sale you already had.

On most DTC gross margins, that's not a diluted win — it's a negative-margin transaction. And because the affiliate dashboard reports it as $85 of attributed revenue against $12.75 of cost, it books as a 6.7x ROAS. The worst orders in the channel report as the best ones.

What the order actually earns, worked

Run a single order all the way down. Take a $100 list-price order on a product with 60% gross margin — $40 of COGS.

Direct order, no creator involved:

Revenue $100, COGS $40 → gross margin $60
Payment processing at 2.9% + $0.30 → −$3.20
Shipping and fulfillment → −$9.00
Contribution margin: $47.80

Same order through an affiliate with a 15% customer discount and a 15% commission on net revenue:

Revenue $85, COGS $40 → gross margin $45
Payment processing → −$2.77
Shipping and fulfillment → −$9.00
Commission (15% of $85) → −$12.75
Contribution margin: $20.48

The affiliate version of the identical order earns 57% less. Which is fine — if the order is incremental. A new $20.48 is strictly better than no order. But if that customer would have converted anyway, the partnership didn't earn $20.48; it destroyed $27.32.

Now blend it. Suppose 40% of code redemptions are genuinely incremental and 60% are interceptions:

40% incremental — +$20.48 each → +$8.19 weighted
60% intercepted — −$27.32 each → −$16.39 weighted
Net: −$8.20 of contribution margin per attributed order

A program reporting 6.7x ROAS is losing $8.20 an order. The break-even incrementality rate here is about 57% — below that, the program destroys margin no matter how large it gets. Scaling it makes the hole deeper, which is precisely what happens when a brand sees a strong reported ROAS and pushes more volume through the channel.

Know what a customer is actually worth before you set a commission — free, no login

The LTV Estimator turns your repeat rate, order value, and margin into a defensible lifetime contribution figure — the ceiling any creator payout has to fit under.

Open the LTV Estimator →

The three leakage channels

Interception isn't random. It concentrates in three predictable places, and all three are auditable this week.

Coupon and deal aggregators. Sites and browser extensions that surface codes at checkout. They contribute effectively zero demand and capture a disproportionate share of commissions, because they sit at the exact moment of purchase intent. Pull your affiliate partner list and sort by revenue — if a partner with no audience and no content is in your top five, that's not a partner, that's a toll booth.

Your own existing customers. Repeat buyers who know the brand, know they want the product, and search for a code out of habit. These orders were your highest-margin orders, and the program converts them into your lowest-margin ones. Segment attributed orders by new versus returning customer — a high returning-customer share in affiliate is a direct measure of cannibalization.

Codes leaking out of their context. A creator's code posted in a deal subreddit or a Discord stops being a creator's code and becomes a public discount, permanently. Volume on a code that keeps climbing months after the post went live is the signature — real creator-driven volume decays sharply within days.

How to measure incrementality without a data team

Full geo-experiments are the gold standard and most brands can't run them cleanly. Three practical approximations get you most of the way.

Baseline lift. Before an activation, establish the expected order count from the four surrounding weeks, adjusted for any promo. Then compare total company orders in the activation window against that baseline. The gap — not the code redemptions — is your incremental result. If total orders didn't move while 300 codes were redeemed, you have your answer.

Unique-code discipline. Give every creator a code that exists nowhere else and expires. This won't stop leakage, but it makes leakage visible, because you can watch redemption decay against the content's actual reach.

New-customer-only economics. Evaluate the program exclusively on first-time buyers. This is a deliberately conservative read — it undercounts genuine reactivation — but it removes the largest source of inflation and gives you a floor. If the program is unprofitable on new customers alone, the returning-customer orders aren't rescuing it; they're the thing dragging it down. What those new customers are worth over their lifetime is the number that decides whether an unprofitable first order is acceptable, which is the calculation in how to calculate customer lifetime value.

Flat-fee creators and the CPM comparison

Flat-fee influencer deals dodge the discount-stack problem entirely but introduce a different one: there's often no measurable response at all, so the spend gets defended on impressions.

If you're going to buy on impressions, price it like impressions. A $6,000 post reaching 180,000 people is a $33 CPM. Compare that honestly to what the same $6,000 buys in paid social — usually several times the reach, fully targetable, and measurable. The creator premium has to be justified by something the ad platform can't deliver: credibility, a specific niche audience, or usage rights to content you'll run as paid media afterward.

That last one is frequently the real value and it's rarely priced in. A creator deal that includes perpetual whitelisting rights for ads that outperform your in-house creative can be worth the fee even if the organic post converts nothing. But that's a creative-sourcing decision, not an acquisition channel, and it should be measured as one.

Designing a program that survives the math

The changes that move a creator program from margin-negative to margin-positive are mostly structural, not tactical:

Commission on contribution margin, not on revenue. Paying a percentage of revenue means the payout is indifferent to what you actually earned. Paying against margin aligns the creator with the same constraint you have.
New-customer-only commissions. Pay full rate on first-time buyers, a reduced rate or nothing on returning ones. This removes the single largest cannibalization channel in one policy change.
Exclude coupon and extension partners from the network, or move them to a rate that reflects their actual contribution — which is close to zero.
Expire codes on a fixed window tied to the content, so leaked codes stop earning.
Block stacking between creator codes and sitewide promos so an activation during a sale can't compound into a negative-margin order.
Cap the fully loaded payout at a share of contribution margin you've actually calculated, and reprice the discount down if the commission has to come up.
Run one holdout per quarter — a region or a two-week pause — and reset your incrementality assumption against the result rather than carrying last year's guess.

Audit your creator spend

The tell is almost always the same: affiliate shows the highest ROAS of any channel, a large share of its attributed orders are returning customers, and blended CAC hasn't improved in the two quarters since the program scaled. Each of those facts is defensible on its own. Together they describe a channel that is reallocating credit rather than creating demand — and charging a discount plus a commission for the service.

Want to know what your affiliate and influencer program is actually contributing?

I rebuild creator economics on contribution margin, size interception against your new-versus-returning mix, and set commission and discount structures your unit economics can carry. Starts with a free diagnostic call.

Request a free audit →