Short answer
Marketing KPIs are chosen by working down from the business goal: one result measure per goal becomes the main KPI, with two or three measures beside it that explain it. Each measure gets a written definition, one source and one owner, and the target comes from the firm’s own past; industry averages are not used.
Picture this: an imaginary accountancy practice wants to win clients from its website. Every Monday a table of twenty rows is opened: impressions, clicks, followers, form submissions, email open rates. The meeting ends and nobody can say what to do next, because none of the rows is tied to the question “which goal does this number serve?”. That is where the question of how to choose marketing KPIs starts. The answer is not to add or remove rows; it is to pick a few numbers working down from the goal of the business and to write down what each one means.
In the steps below, every recommendation that is not marked as resting on a Google help page is our own working advice, and some are also marked “Our advice”. The Google pages are listed as sources. All examples belong to an imaginary service firm; the numbers are round and invented and do not show any real result.
Step 1: Reduce the business goal to one sentence
Our advice: before choosing any measure, write in one sentence what the business wants this period, for example “to increase the number of new customers coming from the website”. If the goal contains no result, no measure can be tied to it. Then write the path backwards: before someone becomes a customer, an enquiry arrives, some of the enquiries fit the service, and some of those that fit become customers. Each link is a candidate measure. The choice comes after that, and every number you pick has a ready answer to “why this one?”.
Step 2: Separate result measures from activity measures
A result measure counts the trace the work leaves outside: a customer, a qualified enquiry, an appointment. An activity measure counts the work the team does: articles published, emails sent, tests run. Our advice: the main measure should always be a result measure, and activity measures should stay in the team’s own work plan. The table below puts the two side by side.
| Type | What it counts | Example | What it tells you for a decision |
|---|---|---|---|
| Result measure | The trace the work leaves outside | New customer, qualified enquiry | It is the number closest to the goal; to explain why, you need other measures beside it |
| Activity measure | The work the team does | Number of articles published, number of emails sent | It shows that the work was done; it does not show whether it worked |
Step 3: One main measure per goal, with two or three that explain it
Our advice: every goal gets one main measure. It is the measure that counts the goal itself, in the example the number of new customers. Beside it, add two or three measures that explain why the main measure moves. The example has three: the number of enquiries, the number of qualified enquiries and the ad spend per qualified enquiry. The main measure gives a single answer to “is it going well?” in the meeting, and the explaining measures answer “why?”. If someone wants to add a fourth explaining measure, first ask which one it would replace.
Step 4: Write a definition card for each measure
Two people can count different things with the same word. Our advice is to write a short definition card for each measure: its name, what it counts, what it does not count, which date it is counted by, its source, its owner and its target. An example card for our imaginary firm: “Qualified enquiry: an enquiry that arrives by form or by phone and that the salesperson judges to fit our service and marks as qualified in the CRM. It does not count the same person’s second enquiry in the same month. The month is assigned by the day the enquiry arrived.” In our data and CRO service too, the measurement plan defines in writing what each event means.
If you use GA4, the card has to be made concrete in three places. The first is which event is counted. According to Google, a key event is an event that measures an action particularly important to the success of your business, and any event you collect can become a key event; which one to mark is your decision. For a short definition of the term, see key event in the glossary. The second is the counting method. On Google’s page, “Once per event (Recommended)” counts a key event every time it is triggered, and “Once per session (Legacy)” counts it only once within a session. In the page’s example, if someone triggers a key event five times in a session, the first method counts 5 and the second counts 1. The default method depends on how the key event was created. If you do not write the counting method down, the same form submission can appear as a different number in two reports. The third is the denominator of a rate. In GA4, “Session key event rate” is the number of sessions in which a key event happened divided by the total number of sessions. If the denominator were people, the same rate would come out differently, so write the denominator on the card too. To put event names, parameters and the key event decision on one page, see how to write a GA4 measurement plan.
The same distinction applies in the sales process. Google’s recommended events page recommends a set of events for anyone who wants to measure the whole lead generation funnel. Three of them are generate_lead (a form submitted online, or information submitted offline), qualify_lead (marked as fitting the criteria to become a qualified lead) and close_convert_lead (became a converted lead, that is, a customer). Those three follow the same path as our enquiry, qualified enquiry and new customer. For a short definition of the term, see lead in the glossary.
Step 5: Write down each number’s source and owner
Our advice: every measure gets one source and one owner. The source is where the number is read: the sales result from the CRM, spend from the ad platform, the key event from GA4. According to Google’s page, many reports, such as Landing pages and User acquisition, have a “Key events” column; if you will read the number there, write the report’s name on the card. Google also states that if you create a Google Ads conversion from the same key event, you can report the same conversion counts in Google Ads and Analytics, and it reduces the risk of data discrepancies. Even so, the two systems differing is often expected; we explained which differences are normal in why GA4 and ad platforms disagree.
Our advice: the owner should be the one person responsible for explaining why the number moved. The authority to change the definition sits with that person too. To gather the numbers on one screen, you can build a report dashboard in Looker Studio, and to see where people drop out between steps, see GA4 funnel analysis.
Step 6: Set the target from your own past
Our advice: do not aim at an industry average. An average does not know your product, your price or your channels. Take the target from your own recent periods and calculate backwards. In our imaginary firm, last quarter brought 100 enquiries, 40 of them were qualified, 10 became customers and ad spend was 20,000 units. Two rates come out of that: 40 per cent of enquiries were qualified, and 25 per cent of qualified enquiries became customers. Spend per qualified enquiry is 20,000 divided by 40, which is 500 units.
Let the target for this quarter be 12 new customers. If 25 per cent of qualified enquiries become customers, 12 customers need 48 qualified enquiries (12 divided by 0.25). If 40 per cent of enquiries are qualified, 48 qualified enquiries need 120 enquiries (48 divided by 0.40). If spend per qualified enquiry stays at 500 units, spend comes to 48 times 500, which is 24,000 units. This is not a forecast; it is arithmetic done on the assumption that the past rates stay the same, and if the rates change, the calculation changes. You can also do the division in the cost per order (CPA) calculator. For how to cap spend, see how to set an ad budget, and for the questions to answer before adding budget, see measure before you raise the budget.
If you have no history, for example because you are opening a new channel, set the first period aside for measuring only. That period’s numbers become the starting point for the next target.
Step 7: Tell leading measures from lagging ones
A lagging measure is the goal itself and it shows late: a new customer is recorded weeks after the enquiry. A leading measure moves before the result: a qualified enquiry comes before the customer. Our advice: look at the leading measures week by week, and at the lagging measure at the end of the period. A leading measure is useful only if it moves together with the lagging one. To see this, write the numbers of a few periods side by side. If enquiries rise and customers do not, that measure has lost its leading quality, and its definition or its place in the order needs another look.
Step 8: Decide what to drop
Our advice: for each measure, ask “if this number changes, which decision do I change?”. If there is no answer, take the measure off the scorecard. What usually goes is this: numbers tied to no goal, those whose definition cannot be written, those without an owner, and those that give two different values in two sources. Numbers such as followers can stay if they explain a business result; if they do not, they go to the archive. A dropped number can sit in the archive; it just does not enter the scorecard. We covered the same choice for social media in how to prepare a social media report.
Step 9: Put it all on a one-page scorecard
The table below is an example of a quarterly scorecard for our imaginary firm. The numbers are round and invented; spend is written in plain units. The target column comes from the calculation in Step 6.
| Measure | Last quarter | Target | Source and owner |
|---|---|---|---|
| New customers (main measure, lagging) | 10 | 12 | CRM; salesperson |
| Qualified enquiries (leading) | 40 | 48 | CRM; salesperson |
| Number of enquiries (leading) | 100 | 120 | GA4 key event and CRM; web owner |
| Spend per qualified enquiry (units) | 500 | 500 | Ad platform and CRM; ad manager |
Read the first row first. If there are not 12 new customers at the end of the period, turn to the explaining measures: were there too few enquiries, did the qualified share drop, or did the cost rise? In the weekly report these four rows go into the numbers section of the Monday report; the decisions part of the report is written with an owner and a review date.
Common mistakes
- Changing the definition when you miss the target. If the qualifying criterion is loosened, qualified enquiries rise, but the number now counts something else. Our advice: change a definition only with a written, dated reason, and recalculate earlier periods with the new definition.
- Tying a reward or pressure to one number. If the number of enquiries is rewarded, unqualified forms may also be counted as enquiries. Our advice: read the main measure and the explaining ones together.
- Deciding on one week’s movement. Our advice: read the number side by side with last period and with your own recent periods; do not turn a single movement in a leading measure into a change of target.
Common questions
What is the difference between a KPI and a metric?
In our usage, a metric is any number that can be measured, and a KPI is the few numbers tied to a goal that steer decisions. In GA4 menus “metric” has its technical meaning; Google’s dimensions and metrics page explains the available dimensions and metrics and how they are populated, and the “Session key event rate” formula is written there. A metric becomes a KPI when it is tied to a goal and has an owner and a target written down.
How many KPIs should I track?
Our advice: one main measure per goal and two or three that explain it. There is no industry number for this; the limit is that each measure can change a decision. A measure that has no answer to the question in Step 8 is surplus.
How do I set a target if I have no history?
Set the first period aside for measuring and make that period’s numbers your starting point. Then, as in Step 6, calculate backwards from the new customers you need, using your own rates. Instead of taking an average from outside, write the target with the real rates you hold at the end of the first period.
A small job you can do on your own today: pick one goal, write its main measure, and fill in the seven fields of the definition card above for that measure. If one field stays empty, that measure is not ready yet; close that gap first.
Sources
This article is general information, not legal, financial or technical advice. Platforms, their rules and the law change over time, so some details may have changed since 14 November 2026 or contain small errors. Before you make a decision for your business, check the current official sources or write to us.