How to build an influencer marketing measurement plan

Four kinds of evidence, and what each one cannot do

Campaign reporting mixes four different kinds of evidence and presents them as one story. They answer different questions, and the trouble starts when one quietly stands in for another.

Pick one primary measure that matches the objective, and write it down. Supporting metrics explain what happened. They should not replace the agreed measure later, when the agreed measure is inconvenient.

Scroll the table sideways to compare all columns.

Write the definition next to the metric

Engagement rate is the clearest example of a number that means nothing on its own. Which interactions count? Is the denominator followers, reach or impressions? Two agencies can report the same campaign and differ by a factor of ten without either of them being wrong, because they answered those questions differently.

So write the answers down. For views, keep the platform’s own definition and name the platform. For leads, define the action and say what makes one qualified.

Keep the reporting window and the included channels consistent across the campaign. When they have to differ, show the difference instead of presenting the figures side by side as though they are comparable.

Decide what you are comparing against

Every result is a comparison, whether or not anybody says so out loud. “Good” means better than something. Name the something.

A comparison with your own earlier campaign helps if the audiences, objectives, formats and measurement methods were close enough. Write down what was different, because those differences are the most likely explanation for the result.

An industry benchmark needs four things before it is worth using: a source, a time period, a sample, and a metric definition. Without them, a confident-looking average is decoration, and it is decoration you may be about to set a budget against. Apply the same suspicion to your own historical data, which was collected by people who had different tracking and different definitions.

Attribution and incrementality answer different questions

A tracked purchase tells you an order was associated with a link or a code. That is attribution, and it is useful. It does not tell you the campaign caused the order.

The buyer may have been coming anyway. They may have searched for the brand, seen a retargeting ad, and used the creator code because it was the first one they found. Attribution counts what it can see, which biases it toward the last thing that touched a customer who was already close.

How many purchases the campaign actually caused is a different question, and it needs a counterfactual: a comparable group who did not see the campaign. Google’s documentation on lift studies describes exactly this treatment and control comparison for estimating incremental outcomes.

Both numbers belong in a report. Labelling one as the other is how a channel gets a budget it did not earn, or loses one it did.

Decide when you will look, and stick to it

The date is part of the plan, and leaving it out is how measurement turns into rummaging. Somebody checks on day three, sees a slow start, and asks whether the creative is wrong. Somebody else checks in month four, when a promotion and a season change have arrived on top of the campaign.

Set the review window before launch, and set it around how the thing you are measuring behaves. Delivery data is there almost immediately. Engagement settles over days. Survey fieldwork has to run long enough to collect the sample the study needs, and the window has to sit close enough to exposure that people still remember. Purchases in a considered category can trail the campaign by weeks.

Then write down what happens at that review. Which numbers get read, by whom, and what decision they feed. A review with no decision attached is a status meeting, and status meetings are where good measurement goes to be admired and forgotten.

Agree in advance what a disappointing result means, too. If the honest answer is that a weak number would not change anything, the measurement is decoration and the budget is better spent on the campaign.

Ask for a report that helps you decide what to do next

The point of measurement is the next campaign. A report that only tells you how the last one went is a receipt.

So bring the open questions to the next brief alongside the results. “Engagement was strong but we could not tell whether it reached anyone new” is a more useful sentence for a campaign team than any single headline number.

• The objective, reporting period and delivered campaign scope.
• The source, definition and denominator for each primary metric.
• Tracking gaps, exclusions and other limitations.
• Study methods and uncertainty where a lift study is included.
• A next test tied to the evidence, with remaining unknowns stated.