Most influencer reporting I see is a screenshot of impressions next to a number of discount code redemptions, and a conclusion drawn from the gap between them. Both numbers are real. Neither answers the question anyone actually has, which is whether the money produced sales that would not have happened anyway.
Creator marketing is genuinely hard to measure. That is not a reason to measure it badly.
Why the default numbers fail
Impressions and engagement measure the creator, not the campaign. They tell you the post was seen and liked. They correlate weakly with whether anyone bought anything, and they are the numbers creators report because they are the numbers that flatter creators.
Discount codes undercount, badly. They only capture people who both remembered the code and used it at checkout. A large share of people who buy because of a creator do not use the code — they search your brand, land on your site, and buy at full price days later. Code redemptions are a floor, not a measurement.
Tracked links undercount too. The link is in a story that expired, or in a bio that changed, or the person saw it on their phone and bought on a laptop. Link clicks capture the immediate, same-session buyer and miss the rest.
Last-click attribution hides almost all of it. The classic path: story on Monday, brand search on Thursday, purchase from the branded search ad. Google books the revenue, the creator programme shows nothing, and someone concludes influencer marketing doesn’t work. It is the textbook failure case for last-touch attribution.
Put together: your creator programme is almost certainly performing better than your dashboard says, and you have no idea by how much.
A stack that works, cheapest first
1. Codes and links, understood as directional. Still worth doing. Give each creator a unique code and a unique link, and treat what they capture as a lower bound and a relative ranking between creators — not as the campaign’s return.
2. Branded search and direct traffic, watched around post dates. This is the cheapest genuinely useful signal available and almost nobody runs it. Mark the dates creators post. Look at branded search impressions and direct sessions in the following 72 hours against a baseline of the preceding weeks. A real lift is visible without any special tooling.
Keep this out of your paid search reporting, though — a creator-driven spike in branded searches will inflate your branded campaign’s apparent performance, which is the same problem Performance Max creates.
3. A post-purchase survey. One question at checkout: “How did you hear about us?” with a free text or a creator list. Self-reported attribution is biased and imprecise, and it is still one of the best tools available for channels that cannot be clicked. It catches exactly the buyers that codes and links miss. Every ecommerce brand should have one and most do not.
4. Matched-market or holdout testing. The only method that establishes causation. Run creator activity in some geographies and not others, with comparable baselines, and compare total revenue rather than attributed revenue. Expensive in the sense that you deliberately forgo activity in the control — and the only way to answer the question honestly.
5. Marketing mix modelling, once spend is large enough. It handles unclickable channels by design, which is exactly the problem. Google’s open-source Meridian added brand-equity signals like branded query volume in version 2.0, which is directly relevant here: branded search volume is often the clearest fingerprint a creator campaign leaves.
Most brands should run levels 1–3 permanently and level 4 once or twice a year on the largest programmes.
The metrics worth reporting
Replace the impressions screenshot with:
- Cost per thousand relevant impressions, not raw reach. If half the creator’s audience is in a country you don’t ship to, reach is a vanity number.
- Incremental revenue estimate, with the method stated. “Code revenue USD 4,200; survey-attributed revenue USD 11,800; branded search lift 34% over baseline in the 72 hours following” is a defensible report. A single confident ROI number is usually fiction.
- Cost per creator, all-in. Fee plus product plus usage rights plus the hours your team spent. The last one is routinely excluded and routinely the largest.
- Content performance as paid media. Often the real return: how did the creator’s asset perform when you ran it as an ad, against your existing best creative? This is measurable precisely, unlike everything else here.
- Repeat rate of creator-acquired customers. Do these buyers come back? A cheap first purchase from an audience with no affinity is worth less than a more expensive one that repeats. Read alongside LTV and CAC together.
Set it up before the campaign, not after
Three things that have to exist before the first post goes live, because none of them can be retrofitted:
- A baseline. Four to six weeks of branded search, direct traffic and total revenue before any creator activity. Without it, you cannot claim a lift.
- A post calendar with real dates. Creators post late. If you do not log the actual date, your 72-hour windows are pointing at nothing.
- The survey, already running. Turning it on the week of the campaign gives you no comparison and a response profile skewed by the campaign itself.
What good looks like
A brand running creator marketing seriously has a baseline, unique codes and links per creator, a post-purchase survey running permanently, a log of actual post dates, usage rights negotiated up front so winning content can be amplified, and one holdout test a year on the biggest programme.
That is not an expensive setup. It is mostly discipline, and it is the difference between a channel you can defend in a budget meeting and one that gets cut because nobody could prove it did anything.
The short version
Codes and links are a floor. Engagement is noise. Last-click is blind to most of the effect. Build from the cheap signals up: baseline, branded search lift, a post-purchase survey — then holdouts when the spend justifies them. Report a range with the method attached instead of a confident number nobody can defend.
And remember that the most measurable return in creator marketing is usually the least discussed one: taking the content that worked, buying the rights, and running it as paid media. More on picking the right partners in macro vs micro influencers.
If your creator programme cannot prove its value and you suspect it deserves better, measurement is usually the first thing I audit.