Most influencer marketing ROI numbers are not lies, but they are not always measurements either. Some revenue can be read directly: a real order, attributed to a real creator. Much of the rest has to be estimated, because the buyer saw a video, searched the brand days later, and checked out on another device. The problem is not estimation; it is passing a guess off as measurement, like treating earned-media value as if it were cash. No wonder that in Linqia’s 2026 State of Influencer Marketing, 79% of marketers said they struggle to measure ROI and 48% pointed straight at attribution as the biggest gap. This is the honest guide: know which part you are reading, which part you are modeling, and do each one rigorously.
It matters more every year. The influencer marketing industry reached roughly $24B in 2024 and ~$32.5B in 2025 (that is total industry value, not pure brand spend), and US brand spend alone passed $10.5B in 2025. The more you spend, the more the measurement gap costs you.
How do you measure influencer marketing ROI?
Measure influencer marketing ROI as (attributed revenue − total cost) ÷ cost. The honest catch is the word attributed: part of that revenue can be read directly from real orders, and part can only be estimated. The accurate approach is layered — read realized revenue wherever you can, and model the rest rigorously instead of leaning on a single proxy like promo codes or earned-media value.
The formula is simple. The integrity of the inputs is everything. Before you trust the result, be honest about what each input actually is: a weak proxy, or a rigorous method.
The proxies, and where they quietly miscount
These four are how you estimate when you cannot see the sale. They are legitimate and useful, but each has a failure mode that is rarely disclosed, and none should be mistaken for ground truth.
- Promo / discount codes. Easy, creator-friendly, and the most-used method in practice. But they undercount: many buyers never type the code, find a sitewide coupon instead, or buy after it expires. Code-only ROI is a floor, not the full picture.
- UTM links + last-click analytics. Clean when the journey is one click. But creator discovery is rarely one click. People see a product in a video, search the brand, then buy on desktop days later. Last-click hands that sale to Google, not the creator who started demand.
- Earned-media value (EMV). A modeled equivalent-ad-spend number. It is a reach proxy dressed up as money; it correlates with impressions, not purchases, and different vendors compute it differently. Fine for comparing campaigns to each other, dangerous as a stand-in for revenue.
- Brand-lift / post-purchase surveys. The only way to catch the dark-social sales the trackers miss, but they are sampled, self-reported, and lagging. Directional, not precise.
The danger is not using a proxy; it is treating one as a measurement. Email marketing learned this the hard way when Apple’s Mail Privacy Protection made open rate meaningless overnight — a metric everyone trusted became noise. Reach-based influencer metrics like EMV are on the same path. It is the same trap as why engagement rate is a vanity metric.
The two rigorous tiers: read what you can, model the rest
Two methods are different in kind from the proxies above, and the strongest measurement setups use both, layered.
Tier one — read the actual orders. Where you can ingest a brand’s real sales data and attribute orders to the creators who drove them, that slice of ROI needs no modeling at all. You are not estimating; you are reading realized revenue, creator by creator.
Tier two — model everything you cannot read. Most of the journey is not directly readable: view-through, dark social, the cross-platform path between first touch and purchase. Modeled / multi-touch attribution (MMM/MTA) is the rigorous way to account for it. Done well, it is the opposite of guesswork. It weights real signals to estimate each creator’s contribution across the whole path. It needs volume and clean inputs to behave, and a model is never a receipt, which is exactly why it belongs next to realized revenue: each one covers what the other cannot.
That layering is the approach Hyperstar is built on: connect your store, and it attributes realized revenue to specific creators and ranks them by the dollars they actually drove. Rigorous modeling accounts for the rest, so the estimated part is estimated carefully, anchored to real sales rather than hand-waved with EMV.
Here is the honest scope, because a guide that overclaims is just another inflated number. Hyperstar’s realized-revenue attribution covers Shopify and Amazon-US merchants today (Shopify ingestion is newly shipped), and the modeling / multi-touch attribution layer that ties the rest of the journey together is in active development. There is no first-party TikTok Shop attribution yet — the widely cited ~$64B in 2025 TikTok Shop GMV is a single-firm estimate, and creator-level attribution there is still on the roadmap — and Instagram GMV is not available as a creator signal at all. Where your sales live in Shopify or Amazon-US, you can read realized ROI today; everywhere else the rule is the same: estimate rigorously, label estimates as estimates, and never pass a model off as a receipt or a receipt off as the whole picture.
The takeaway
There is no single method that solves every attribution problem. There is a layered hierarchy of confidence. Read realized revenue where you can. Model everything you cannot with rigorous multi-touch attribution, not a vanity proxy. And never wave an EMV number around as if it were cash. Read the sales that happened, model the rest honestly, label the estimates as estimates, and your ROI stops being a story you tell and becomes a number you can defend.
Want to read realized revenue on your own store data, with rigorous modeling for the rest instead of another EMV guess? Get started.