Influencer Marketing Attribution Explained: Last-Click, Multi-Touch and Incrementality

What attribution actually measures

Attribution is the method you use to decide which piece of marketing gets credit for a sale. In influencer marketing there isn't one correct answer — there are three different questions, and the three common methods each answer a different one. Last-click attribution asks "what did the customer click right before buying?" Multi-touch attribution asks "which touchpoints were part of the journey, and how much should each one get?" And incrementality testing asks "how many of these sales would not have happened without the campaign at all?" The three methods can give markedly different answers for the same campaign, and none of them is objectively "correct" — they measure different things.

Last-click attribution

Last-click is the default in most analytics tools and the easiest to understand: 100% of the credit for a sale goes to whatever the customer clicked last before completing the purchase. A customer clicks an influencer's link, browses, leaves, then comes back three days later through a Google ad and buys — Google Ads gets full credit, and the influencer gets none, even though the influencer created the original interest.

Strength: simple to implement, needs no modelling, and works well when the customer journey is short and involves a single channel.

Weakness: it systematically undercounts channels that appear early in the journey — which is exactly where influencer content usually sits. Influencer marketing typically drives awareness and consideration, while a different channel (paid search, direct traffic, email) captures the final click. Last-click moves credit away from the influencer and toward whichever channel the customer happened to use to come back.

Multi-touch attribution (MTA)

Multi-touch attribution splits credit for a sale across several of the touchpoints in the customer's path, instead of giving it all to the last one. The most common models are:

• Linear: every touchpoint in the path gets an equal share of the credit.
• Time-decay: touchpoints closer to the purchase are weighted more heavily than early ones.
• Position-based (U-shaped): the first and last touchpoints get the most credit (typically 40% each), with the remaining 20% split across whatever happened in between.
• Data-driven (algorithmic): a machine-learning system calculates each touchpoint's actual contribution from patterns across converted and non-converted paths. This is now the default in several major analytics platforms, but it typically needs a meaningful volume of conversions before it can train reliably — below that threshold, the tool usually falls back to a simpler rule-based model.

A concrete example of the difference: a customer sees an influencer's story (touchpoint 1), clicks a branded Google ad two days later (touchpoint 2, the last click), and buys. Last-click gives 100% of the credit to the Google ad. A linear MTA model splits it 50/50. A position-based model typically gives 40% to the influencer (first touch) and 40% to the Google ad (last touch), with 20% spread across anything in between.

Strength: it credits the influencer for the role the content actually played early in the journey, instead of that channel disappearing from the numbers entirely.

Weakness: it depends on being able to track every touchpoint back to the same user — which rarely works fully across devices, apps, and browsers with limited cookie lifetimes. A view that never involves a click (a video the customer simply watches without tapping anything) usually doesn't register as a touchpoint at all — see why in how to track influencer marketing performance.

Incrementality testing (holdout tests)

Incrementality testing asks a different question than the two methods above: not "who gets the credit", but "how many of these sales wouldn't have happened without the campaign at all?" The method needs no touchpoints or clicks to track. Instead, you randomly split your audience into two groups: an exposed group that sees the campaign, and a holdout group (typically 5-10% of the audience) that is deliberately kept out of it. After the campaign window, you compare the purchase rate between the two groups. The difference is the campaign's real, causal effect — the "lift" that neither last-click nor multi-touch attribution can answer precisely, because both of those methods can only count what left a trace, not what didn't.

Strength: the only one of the three methods that actually measures causation rather than correlation. It captures sales that never touched any tracking mechanism at all — customers who saw the content, remembered the brand weeks later, and searched for it themselves.

Weakness: it needs a large enough audience for the difference between the two groups to be statistically reliable, plus a genuine ability to exclude part of the audience from exposure — easier in paid media (where you control delivery) than in organic influencer content (where you can't stop a specific person from seeing a post).

Comparison

The attribution window shapes all three methods

Whichever method you use, you have to set an attribution window — how long after a touchpoint a sale still counts. Too short a window systematically understates all three methods, because it cuts the journey off before the customer has time to act. See how the tracking mechanics themselves work in how to track influencer marketing performance, and the trade-off between tracking mechanisms in discount codes vs tracking links. For the specific mechanics of why Google Analytics in particular under-reports influencer sales, see why influencer sales don't show up in Google Analytics. For exactly how long that window should be — and why the cookie behind a tracking link can expire even sooner than the window you've configured — see attribution windows in influencer marketing.

Worked example: what a holdout test shows that last-click misses

The numbers below are hypothetical and for illustration only. This is not a real Make Influence customer case, and none of the figures are benchmarks.

Assume a brand runs an influencer campaign against an audience of 20,000 people and sets aside 2,000 of them (10%) as a holdout group that never sees the campaign.

• The holdout group (2,000 people) buys at a baseline rate of 3.0% = 60 purchases that would have happened regardless of the campaign.
• The exposed group buys at a rate of 3.9%. In an equal-sized comparison group of 2,000 people, that's 78 purchases.
• Incremental lift = 78 − 60 = 18 purchases per 2,000 people, i.e. (3.9% − 3.0%) ÷ 3.0% = 30% lift over baseline.
• Scaled across the remaining 18,000 exposed people: 18,000 × (3.9% − 3.0%) = 18,000 × 0.9% = 162 incremental purchases the campaign actually caused.

Now assume the brand's last-click tracking (via unique links and codes) recorded only 90 "attributed" sales for the same window. 90 out of 162 is roughly 56% — the rest of the real lift happened as branded search, direct traffic, or purchases that never touched a tracked link or code at all. That isn't a sign the tracking was set up badly; it's the category of sale last-click attribution structurally cannot see, and that only a holdout test reveals.

Decision framework: which method for which situation

• IF you need to report results while the campaign is still running, and want a fast, operational number → use last-click on tracked links and codes. It isn't the most accurate figure, but it's the fastest.
• IF you want to credit the influencer for creating consideration early in the journey, not just the final click → use a multi-touch model, provided you can track most touchpoints back to the same user.
• IF you need to decide whether influencer marketing as a channel is actually worth the spend — not just how credit gets split internally → run an incrementality test. It's the only one of the three that answers "does this channel work at all". Once it gives you a reliable incremental sales figure, the natural next step is turning it into an ROI percentage — see how to calculate influencer marketing ROI.
• IF your audience is too small for a statistically reliable holdout group → stick with last-click and MTA as directional numbers, and treat them as a floor on the real effect, not the full picture.

Make Influence's operational perspective

In Make Influence's experience, the most common mistake is treating the last-click number as "the truth" and then wondering why the influencer channel looks weaker than it feels. Last-click is structurally built to undercount channels that work early in the journey — and influencer content does exactly that more often than it captures the final click itself. We recommend using last-click operationally (it's fast, and good enough for comparing creators against each other), but never as the final word on whether the channel as a whole is worth the investment. That question is better answered by an incrementality test, even though it's heavier to set up. See also is influencer marketing worth it for ecommerce brands? for the broader question of how to judge the channel's total value.

FAQ

Which attribution model do most analytics tools use by default?

Many platforms have now moved to a data-driven model as the default, but fall back to a simpler rule-based model when there isn't enough conversion data to train it reliably. Always check which model your tool is actually using before comparing numbers across channels.

Can I run a holdout test on organic influencer content?

It's harder than with paid media, because you can't stop a specific person from seeing an organic post. In practice, holdout tests on influencer content are often scoped geographically, or by staggering when different audience segments are exposed to the campaign.

Why do last-click and incrementality testing so rarely agree?

Because they answer different questions. Last-click counts traces; incrementality testing measures the difference between two groups, regardless of whether a trace exists. See the worked example above for how large that gap can get.

Should I pick one method and stick with it?

No. Use last-click for day-to-day reporting and comparing creators, and run an incrementality test periodically (quarterly, for example) to calibrate how much you can actually trust the last-click numbers.

Does the attribution window affect which method I should choose?

Not which method — how accurate the result is. Too short a window understates last-click and MTA in the same way — see attribution windows in influencer marketing for exactly how long that window should be for each tracking mechanism and ad platform.