Choose influencer marketing KPIs by writing the decision first: continue a creator partnership, change the content, repair the buying journey, or stop spending. Then select one outcome that can inform that decision, supporting measures that explain it, and a condition that prevents a misleading conclusion. Awareness, consideration and sales need different evidence.
A campaign can collect dozens of numbers and still leave its owner unsure what to do. Modash's guide to influencer marketing metrics groups measures by program type and raises useful questions about what goes into ROI. For your brief, take the next step: specify which result would change a planned action.
Start with a decision you can take
"Improve awareness" leaves too much open. "Choose which creator to rebook for another product introduction" gives the report a job.
Before selecting a metric, write down:
- The action you will take after the report.
- The audience and content the decision concerns.
- The date you need the answer.
- The evidence you can obtain before that date.
- The condition that would make you postpone the decision.
The last item matters. A missing creator report, broken destination link or unresolved order count can make a comparison unusable. Record missing evidence as unavailable. A blank cell must not become zero performance.
Match the goal to observable evidence
This map uses Instagram, Google Analytics 4 and Shopify as concrete examples. The selection rules are recommendations. The metric definitions come from the linked platform documentation.
| Goal and decision | Lead measure | Supporting evidence | What it cannot establish |
|---|---|---|---|
| Awareness: rebook for distribution | Instagram Viewers for the agreed content and window | Views, available viewer demographics, campaign cost | Brand recall, attention or unique people across multiple creators |
| Consideration: keep or revise the product explanation | Shopify sessions attributed to the campaign | GA4 web engagement time; a separately defined product-interest action | Whether a visitor intends to buy or whether every visit was captured |
| Sales: fund another paid collaboration | Paid, non-canceled orders assigned under a declared attribution rule | Cost per qualifying order; order value and refund status | Sales caused by the campaign that would otherwise not have happened |
Choose one row as the campaign's main purpose. Other rows can explain results without becoming competing success criteria. An awareness campaign may record orders, but deciding afterward that orders were its only goal changes the evaluation.
Awareness needs a boundary around the audience count
Instagram defines Views as plays or displays and Viewers as unique accounts seeing content on screen. For Reels, a screen appearance can count toward Viewers even without playback. Neither measure establishes that someone remembered the product.
Public accounts can access insights for their own content. That does not give a brand access to a creator's private analytics. Agree on which reports the creator will supply and when. Viewer demographics require a preset timeframe and more than 100 reached accounts. Missing demographics are a reporting limitation, not evidence of an unsuitable audience.
Keep the label supplied in the report. Meta notes that some metrics are estimated or rolling out. Do not rename an older export to fit today's overview. For a multi-creator total, summing individual Viewers counts may count the same account more than once. Call it a sum of creator-level counts unless you have a defensible way to remove overlap.
Consideration needs a defined action
Campaign-attributed store sessions can show recorded visits after distribution. Shopify's documentation explains that sessions rely on cookies and end after 30 minutes of inactivity. One visitor can produce several sessions. Sessions therefore cannot stand in for a count of interested people.
For a more specific signal, define an action suited to the product: submitting a sample request or opening a sizing guide, for example. These are proposed measurements that your team must implement and verify. They are not automatically available platform metrics.
GA4 user engagement measures time with a webpage in focus or an app in the foreground. Use web engagement time to investigate what visitors do after arrival. Longer time alone could reflect interest or difficulty finding an answer. Google also states that the user_engagement event cannot be a key event.
Sales needs an order definition and a credit rule
Write out what counts as an order before calculating cost per order. The recommended definition here is paid, non-canceled orders, with refunds reviewed separately. It is an analyst-defined measure.
That distinction matters because Shopify marketing reports include canceled, pending and unpaid orders. Their sales figures also differ from money received. Reconcile your selected order records before using a default report total as a renewal target.
Google describes attribution as assigning credit. Record the model and lookback window alongside the number. Shopify's Any click model gives full credit to each clicked channel, so adding channel totals can exceed the number of orders. Pick one stated rule for the decision and keep other attribution views separate.
If the decision requires evidence of additional sales caused by the campaign, an attributed-order report cannot answer it alone. Use the separate guide to measuring incremental impact when attribution is incomplete before designing that evaluation.
A hypothetical renewal decision
Suppose a team needs to choose a creator for another sales-focused booking. The hypothetical records below use the same reporting window, attribution rule and order filter. Campaign costs include the same cost categories for both creators.
| Measure | Creator A | Creator B |
|---|---|---|
| Campaign cost | $1,200 | $1,200 |
| Attributed store sessions | 600 | 200 |
| Paid, non-canceled attributed orders | 24 | 20 |
| Orders divided by sessions | 4% | 10% |
| Cost per qualifying order | $50 | $60 |
The calculations are reproducible: A's cost per order is $1,200 divided by 24; B's is $1,200 divided by 20. Their order-to-session ratios are 24 divided by 600 and 20 divided by 200.
B has the higher order-to-session ratio. A has more qualifying orders and the lower cost per order. If the team's pre-agreed renewal ceiling were a hypothetical $55 per qualifying order, A would meet it and B would miss it.
This arithmetic does not prove A will perform better next time. The creators may reach different audiences, and order value or refunds may change the commercial decision. The ratio also uses sessions, not individual shoppers. Treat B's result as a reason to investigate traffic volume and audience differences, rather than proof that B persuades people better.
Before setting a real ceiling, build a complete campaign cost ledger. A target based only on the creator fee can leave out costs your next booking will incur.
A one-page measurement brief
The following is a filled hypothetical brief for the renewal decision. Replace its choices with your campaign's terms before work starts.
| Field | Agreed measurement plan |
|---|---|
| Decision | Choose a creator for the next sales-focused booking |
| Lead measure | Cost per paid, non-canceled attributed order |
| Cost basis | Creator fee, product cost, shipping and allocated production labor |
| Credit rule | Shopify Last non-direct click, applied consistently to both creators |
| Window | First 14 days after each publication; identical collection cutoff rules |
| Supporting evidence | Attributed sessions, order value and refund status |
| Data owners | Campaign manager collects costs; store analyst reconciles orders |
| Renewal condition | Cost per qualifying order at or below the hypothetical $55 ceiling |
| Pause condition | Missing cost records, unresolved payment status or failed tracking check |
| Later review | Revisit refunds 30 days after publication before expanding the budget |
| Stated limitation | Recorded attribution supports this booking decision; additional sales caused remain unknown |
Those dates and thresholds are planning choices, not platform requirements or industry benchmarks. Give the report owner the brief before publication. If the buying cycle needs more time, choose the reporting window before looking at results.
For your next campaign, write the decision and pause condition first. Then ask the data owners to confirm that they can produce the evidence needed to act.
