Reporting
How to Choose Digital Marketing KPIs (Without Reporting 40)
A test for whether a metric is a KPI, the objective-to-KPI mapping, where each number comes from, and the metrics worth refusing to report.
A KPI is a metric that changes what someone does. If a number moves and nobody’s next action changes, it is not a key performance indicator — it is context, and context belongs in an appendix. Most reports fail because they treat those two categories as one, which is how a monthly report ends up with forty rows and no decision in it.
Here is the test, the mapping from objective to KPI, and the four metrics I would push back on being asked to report.
The four-part test for a KPI
A metric earns the label only if it passes all four. Run a candidate through these before it goes near a dashboard.
- One named person can move it. Not influence it — move it. “Brand awareness” has no owner. “Qualified enquiries” does.
- A change in it changes a decision. Write down the decision. If the sentence is “we would look into it”, the metric is context.
- It has a single agreed source. Sessions from the analytics platform and sessions from the ad platform are two different numbers and one of them will be quoted at the wrong moment.
- It cannot be gamed without something else visibly getting worse — and that something else is in the report as a guardrail.
Three of those are about accountability rather than measurement, which is the part most KPI advice skips.
The shape: one primary, two supporting, one guardrail
The site’s strategy examples use this shape and it is worth stating plainly, because it is what stops the count creeping:
- One primary metric decides whether the period worked.
- Two supporting metrics explain movement in the primary.
- One guardrail must not get worse while you chase the primary.
Four numbers. Everything else is available on request. The reason for one primary rather than three is arithmetic: with three “key” metrics there are eight possible combinations of up and down, and no agreed answer for six of them.
Mapping the objective to the KPI
The mistake is choosing a KPI from a list of marketing metrics. Start from what the business is trying to do, then work down to what proves it, then work down to where the number comes from — because a KPI with no reliable source is a monthly argument.
| Business objective | Primary KPI | Supporting | Guardrail | Source of truth |
|---|---|---|---|---|
| More revenue from existing demand | Blended ROAS | Conversion rate; average order value | Contribution margin per order | Ecommerce platform, not the ad platform |
| More qualified sales conversations | Enquiries meeting the qualification criteria | Sessions to high-intent pages; enquiry form completion rate | Enquiry-to-meeting rate | CRM |
| Lower acquisition cost | Blended CPA | Cost per click; landing page conversion rate | Lead quality score or close rate | Ad platforms plus CRM, reconciled |
| Grow an owned audience | Net new subscribers | Signup rate by source; email click rate | Unsubscribe and complaint rate | Email platform |
| Recover from a traffic decline | Impressions for priority queries | Average position; indexed page count | Conversion rate — recover the wrong traffic and this falls | Search Console |
Two things to notice. The source column is not decoration — “blended ROAS from the ecommerce platform” and “ROAS summed across ad platforms” will disagree, often by a lot, and the second one adds up to more revenue than the business actually took. And every row has a guardrail, because every primary metric in that table can be improved by doing something the business would not want.
Where the numbers come from, and where they disagree
Three reconciliation problems decide whether a KPI can be trusted at all, and each one is settled by the source column above rather than by argument:
- Attributed revenue vs banked revenue. Ad platforms count a conversion against the click that preceded it under their own attribution window. Three platforms can each legitimately claim the same sale. The blended figure — total revenue divided by total spend — is the one that matches the bank, and the gap between the two is wide enough to reverse a verdict.
- Sessions vs clicks. Reported clicks are almost always higher than the sessions that follow. Bounced redirects, blocked scripts and bot filtering all sit between the two. Pick one and label it.
- Leads vs qualified leads. The gap between them is where every “marketing generates rubbish leads” argument lives, and it is settled by writing the qualification criteria down before the period starts, not after the numbers arrive.
If the measurement itself is unreliable, no KPI choice survives it. The GA4 audit checklist is the faster fix than arguing about which number is right — and measurement is the first category worth paying for in any marketing stack, because every other purchase is judged with it.
Worked example: choosing KPIs for a lead-generation month
Assume a B2B services business with a target of 20 qualified enquiries a month, a landing page converting at 3%, and 40% of enquiries currently meeting the qualification criteria.
To get 20 qualified enquiries you need 50 raw enquiries (20 ÷ 0.40), which needs about 1,667 sessions
to the landing page (50 ÷ 0.03).
That arithmetic tells you which KPI to make primary. Sessions are the binding constraint at these rates, so the primary KPI is qualified enquiries, the supporting metrics are landing page sessions and form conversion rate, and the guardrail is the qualification rate itself — because the fastest way to hit 50 raw enquiries is to loosen the form, and that shows up as success in every metric except the one that pays.
Change one assumption and the answer changes with it. Lift qualification from 40% to 60% and you need 33 raw enquiries instead of 50 — a third less traffic for the same outcome, from a change that costs no media at all. That is the kind of thing a well-chosen guardrail surfaces and a long metric list buries. The funnel calculator does this arithmetic across all six stages if you want to test the assumptions properly.
Four metrics worth refusing
| Metric | Why it is not a KPI |
|---|---|
| Impressions, alone | Nobody’s next action changes. Useful as a supporting metric for a visibility objective, never as a primary |
| Email open rate | Since mail privacy protection began pre-fetching images, opens count image loads as much as readers. The email marketing checklist uses bounce and complaint rate for deliverability instead |
| Followers | No agreed relationship to revenue, and it can only go up, which makes it a decoration |
| Per-platform ROAS in a client summary | Double counts when platforms claim the same sale. Fine in the appendix, wrong in the headline |
Refusing a metric is easier before the reporting period than after, which is why the KPI conversation belongs in the proposal rather than in month three — “are we agreed on what success is, in a number?” is one of the questions the buyer is answering as they read it.
Putting them somewhere they get looked at
Six KPIs is the practical ceiling for a rolling view — the KPI dashboard template caps at six deliberately, with the rolling average, variance against target and a two-month miss flag calculating themselves, because a dashboard tracking twenty metrics is just a spreadsheet.
The monthly report template then carries the same rows and a KPI definitions tab, so the definitions are agreed once rather than relitigated. If you need the formulas themselves, the glossary has each one with a worked example.
Choosing four numbers is only half of it. What makes them useful in a report is the line beside each one saying why it moved and whether that was intentional — commentary is what the reader actually acts on, and a well-chosen KPI with no explanation gets skimmed like any other row.
Start with the objective. If you cannot write the decision that a metric would change, leave it out — you can always add it in month two, and you will find you never do.