How to Measure Influencer Marketing Without Lying to Yourself

Open up two brands' influencer programs side by side and you can usually spot the same two mistakes within a minute, just pointing in opposite directions.

The first brand has decided influencer is a pure performance channel. They're reading click-attributed ROAS straight off a dashboard, calling it the truth, celebrating or panicking on a number that captures maybe a third of what's happening. The second brand has given up. Influencer is "brand", it gets a vague halo and a shrug, and nobody can tell you whether the last A$40k did anything.

Both are lying to themselves. One's over-claiming on last click, the other's stopped counting because counting got hard. And the honest answer sits in the uncomfortable middle, where you accept you'll never get a perfect number and you triangulate towards a believable one instead.

Here's how I'd actually build that. Not a perfect measurement, an honest one.

Start with the click number, but treat it as a floor, not the answer

The click-attributed ROAS is where everyone starts, and that's fine. It's just that you have to understand what it is.

For most influencer programs, the directly trackable return, the sales that come through a link or a code inside the attribution window, sits low. It's common to see a click ROAS around 0.4 to 0.5 in the early months. If you stop reading there, you conclude the channel loses money and you kill it.

But that number is a floor, not a verdict. It's the portion of impact that happened to fall through a trackable door inside a short window. A huge amount of real, caused revenue walks through doors you're not measuring: someone sees the post, doesn't click, Googles you three days later and buys. Someone watches on their phone and purchases on desktop. Someone screenshots the code and uses it next payday.

So the click number is the start of the maths, not the end. The whole game is working out the honest multiple to apply to it. And you earn that multiple, you don't guess it.

Find your real multiple, don't borrow someone else's

This is the bit most brands get lazy with. They hear "apply a multiplier" and they slap on a flat 2x or 3x because someone on a podcast said so.

Don't. Every brand's true multiple is different because every AOV, product and audience is different. If an agency tells you it's a 20% or 30% uplift across the board, that's a tell they're guessing, because there is no across-the-board number.

You build your own from data you already have. Two honest sources:

Run hold-outs where you can. A clean geo or audience hold-out, or even a structured on/off test, gives you an incremental read. It's slow, it might take two or three months, but it tells you what genuinely wouldn't have happened without the spend.

Then translate that read into a multiple you can use day to day. Say a proper test tells you the channel drove A$1.40 of real revenue for every A$1.00 the click dashboard credited. That's a 1.4x multiplier. Now you can stop running expensive tests every week and just apply 1.4 to the live click number, sense-checking it every quarter so it doesn't drift away from reality. That's exactly how good operators translate a slow incrementality test into a fast daily decision tool.

A 0.45 click ROAS at a real 1.4x multiple is 0.63. Still not break-even on its own, but now layer the other value in and the picture changes completely.

The leakage problem makes you look worse than you are

Here's the one almost nobody accounts for, and it cuts the opposite way to how people assume.

Discount-code leakage. You give an influencer a code, it works site-wide, and within about a week it leaks onto the coupon sites and into the group chats. Now strangers who never saw the influencer are using the code. Most people assume that flatters the influencer's numbers. For attribution, the bigger problem is usually the reverse: the leaked sales drown the signal and you lose the ability to read true performance.

The fix is a checks-and-balances pass before you trust any influencer's number. Compare their code redemptions to their link clicks. If someone drove 27 link clicks but shows 300 code purchases, something doesn't add up, and you can see it plainly. Pull the influencer's own post insights, their real reach and clicks from the back end, and compare. Then build a click-to-purchase ratio for your programme, with leakage stripped out and without, so you know your genuine conversion rate.

A worked version. One brand sees a code show 300 purchases. The influencer's actual reach and a sane click-to-purchase ratio say they could only plausibly have driven about 40 of those directly. The other ~260 are leakage. Whether you credit those 260 to the influencer or to "general discount" changes your read of the channel enormously, and the only way to split it honestly is to do the ratio work rather than take the raw code count at face value.

The point isn't to be precious about it. A sale is a sale, and a leaked code still brought you a customer. The point is that without the leakage pass you can't tell a genuinely brilliant creator from a code that escaped, and you'll renew with the wrong people.

Post-purchase surveys: read the direction, not the decimal

The single highest-value tool here, and the cheapest, is the "how did you hear about us" survey at checkout.

Now, a warning, because people misuse this. The raw percentages are noisy and they undercount badly, plenty of buyers won't name the exact creator who tipped them. So don't treat the survey as a precise revenue meter. Treat it as a direction-finder.

What you're hunting for is over-indexing. If three creators keep showing up in the write-ins out of proportion to what you paid them, that's gold, even if it's not the net read of every dollar they drove. It tells you who to triple down on and who to drop. And you'll get genuine surprises: someone you assumed was a dud over-indexes hard, someone you were sure about barely registers. The crowd in your checkout knows things your spreadsheet doesn't.

There's a second trick hiding in here. The survey surfaces people you never paid. If you're seeding product to a hundred creators a month, you can't put them all in a formal tracking setup, but their names show up in the write-ins anyway. That's a free shortlist of who's actually moving people, drawn from your own buyers.

So: surveys for direction and discovery, never for the precise decimal. Anyone who quotes you a survey-derived ROAS to two decimal places is overselling what the tool can do.

Match the measurement window to the channel

One more honest adjustment, because a single attribution window across every platform quietly lies to you.

Different formats decay completely differently, so the window you measure in has to match the channel. An Instagram story is gone in 24 hours, so its sales arrive in a tight little spike and then stop, and a short window captures most of it. A YouTube integration is the opposite. The video keeps getting found for weeks, so you want to be reading it at 14, 30, 60, even 90 days, because the sales that matter haven't all landed yet.

If you judge a YouTube placement on a 7-day window, you'll write off something that was quietly compounding. If you give an Instagram story a 30-day window, you'll credit it with sales it had nothing to do with. The number isn't wrong because the channel underperformed, it's wrong because you measured it on the wrong clock.

Stop apologising for the halo, but keep it honest

Then there's the part you genuinely can't click-track: the lapsed customer who comes back because a creator reminded them you exist, the Google searches that tick up, the general lift across the account when influencer is loud.

This is real, and there's decent evidence it's the strongest long-run driver of consideration and future growth. The mistake is letting "halo" become an excuse for never measuring anything. Halo is the bit you estimate conservatively after you've done the trackable work, not a fog you hide a wasteful budget inside.

So put a sane, modest number on it and label it as an estimate. One way I like: if you're getting paid usage rights as part of every deal, footage you'd otherwise pay a UGC creator for, you can fairly move a slice of the cost, say 20%, out of "influencer" and into "creative production", because that's genuine value you received. Now your influencer ROAS reflects the cost the influencer activity should actually carry. That's not fiddling the books, it's allocating cost to where the value landed.

Putting it together

None of these layers is exact. That's the whole point. You take the click floor, apply a multiple you earned from a hold-out, correct for leakage so you can read true performance, let the survey tell you who's over-indexing, measure each channel on its own clock, and add a conservative halo you can defend. Stack those and you get a believable picture, which beats a precise-looking false one every time.

Here's what I'd actually do this week if you run influencer and you're not sure it's working. Don't commission a big measurement project. Just pull your last few months of influencer codes and run the leakage pass, redemptions against clicks, and add a "which creator" line to your post-purchase survey. Those two moves alone will show you more truth than any dashboard. Then build your model from what you find, and re-check it every quarter so it stays honest. A rough model you keep honest beats no model, and it absolutely beats a precise one that's quietly wrong.