Guide
Tracking & ROI
Brands
You can't track every influencer-driven sale — in-app browsers strip UTMs, cookie lifespans are short, dark social hides clicks, and people switch devices between seeing a post and buying. The fix isn't better tracking alone: layer direct tracking with proxy indicators, periodic incrementality tests, and (at scale) marketing mix modeling to estimate the full effect instead of just the trackable slice.
Most brands discover quickly that tracked sales — coupon-code redemptions, click-throughs on a tracking link — only capture part of what an influencer campaign actually drives. Five structural gaps explain most of the difference: in-app browsers on Instagram and TikTok that strip or shorten UTM parameters and referrer data when someone taps through from a post; short cookie lifespans in browsers running Intelligent Tracking Prevention (see attribution windows in influencer marketing); "dark social" — a follower who screenshots a post, forwards it in a DM, or just mentions the product out loud instead of clicking a trackable link; device switching, where someone sees the content on their phone but buys later from a laptop in an unlinked session; and plain brand recall, where someone searches the brand name directly instead of clicking through.
None of these is a setup mistake. They are structural limits on what a single tracking link or discount code can capture, regardless of how correctly your UTM parameters are built.
The expensive error isn't the tracking gap itself — it's treating the gap as proof the campaign didn't work. If a campaign generates 50,000 kr. in coupon-code sales and nothing else is measured, the natural but wrong conclusion is that 50,000 kr. is the campaign's total value. The tracked figure is a floor, not a ceiling.
Instead of chasing one perfect tracking setup, work with four complementary layers. None is sufficient alone; together they give a realistic picture.
Unique discount codes and tracking links per creator. Captures the share of sales where the buyer clicks through directly or redeems a code. Cheap and precise for what it catches — but it is systematically a floor, not a full count, for the reasons above.
Signals that move with the campaign without being directly attributable: branded search volume during the campaign window (Google Search Console or Google Trends), direct-traffic lift on the days a creator posts, and follower growth on the brand's own channels. None of these proves causation alone, but a consistent pattern — search volume rising every time a specific creator posts — is a strong signal.
A geographic or time-based holdout experiment: run the campaign in some regions or periods and pause it in others, then compare sales. It's the same logic behind Meta's own Conversion Lift tests for ads. It's the most reliable way to estimate the full effect, including the share direct tracking never sees. See influencer marketing attribution explained for how to set up a holdout test.
At higher advertising and influencer budgets, a statistical model — regression across time-series sales and marketing-activity data by channel — can estimate each channel's contribution without relying on cookies or user-level tracking at all. Google's own Marketing Mix Modeling guidance frames this as MMM's core advantage: it can estimate marketing effectiveness without cookie-based tracking. That makes it especially robust against exactly the gaps Layer 1 hits — but it typically needs months of historical data and rarely makes sense before influencer spend is substantial.
| Layer | What it measures | Requires | Best for |
|---|---|---|---|
| Layer 1 — Direct tracking | Click/code-redeemed sales | Unique codes/links per creator | Any budget size, day-to-day optimisation |
| Layer 2 — Proxy indicators | Directional signals (search, direct traffic) | Search Console/GA4 access | Ongoing, low-cost checks between tests |
| Layer 3 — Incrementality test | Full incremental effect, including untracked sales | Holdout group/period, enough volume for statistical power | Quarterly validation of the tracked share |
| Layer 4 — MMM | Channel contribution across the whole marketing mix | Months/years of historical sales and media time-series data | Mature, multi-channel advertisers with substantial spend |
The following is an illustrative worked example, not an actual Make Influence customer case. Imagine a brand that ran a geographic holdout test last quarter: the influencer campaign ran in half the country and was paused in the other half. Sales in the test region were 60% higher than in the control region during the campaign window — the incremental lift. Direct tracking (codes + links) captured the equivalent of 100,000 kr. in the same period, while the full incremental effect measured by the holdout test was 160,000 kr. That gives a "tracking coverage ratio" of 100,000 ÷ 160,000 = 62.5%.
In the next campaign, without running a fresh holdout test, the brand tracks 150,000 kr. in direct sales. Using the same 62.5% coverage ratio as a rule of thumb, they estimate the full incremental effect at 150,000 ÷ 0.625 = 240,000 kr. That's an estimate, not a newly measured fact — and it needs revalidating with a new holdout test if the media mix, creator types, or season change materially.
| Situation | Recommended layer |
|---|---|
| Just starting, low budget | Layer 1 alone, plus a simple proxy check (search volume) |
| Running influencer marketing on an ongoing basis but never validated the coverage ratio | Add Layer 3 (one holdout test) |
| Spending across several channels and need to allocate budget between them | Layer 4 (MMM), once enough history exists |
| Need to justify budget to leadership without full data | Combine Layers 1 + 2, and be explicit that the number is a floor |
Our experience is that brands who steer purely by code and link tracking systematically under-invest in influencer marketing, because they're only looking at the tip of the iceberg. We recommend running at least one incrementality test in a new brand's first quarter specifically to calibrate how large the gap typically is between tracked and real effect — then reusing that number as an acknowledged margin of uncertainty, not a permanent truth.
No. They remain the cheapest and most precise tool for day-to-day optimisation — which creator, which post, which product converts best. They just aren't a complete picture of total effect.
Quarterly is a reasonable default for most brands, or whenever the media mix, creator types, or season change enough that an old coverage ratio can no longer be assumed to hold.
Rarely as a first step. MMM needs months to years of historical cross-channel data to produce reliable estimates — most brands get more value from Layers 1-3 before a model is worth building.
Branded search volume (Google Search Console is free) cross-referenced against the dates a creator posted. It needs no new setup and often gives a fast directional signal.
That's exactly what an incrementality test answers: if sales in the control group (without the campaign) are as high as in the test group, there's no incremental effect to explain — a real signal the budget isn't working, not just a tracking problem.
Make Influence
Find creators with real audience data, run collaborations in one place, and see clicks and sales per creator while the campaign is live.
Book a demoCreate accountMake Influence
Apply to campaigns from brands that are actively looking, follow your own clicks and sales, and get paid without chasing invoices.
Create creator profileMore creator guidesMake Influence
Briefs, agreed terms, tracking links and results sit together — so brands and creators see the same numbers.
See how it worksBrowse the Academy
