What nobody tells you before defining your first KPI
The most frequent question isn’t which ones to choose, but where to start without the project dying in the second week. The short answer: choose a process that already generates data and hurts little. If your CRM is half-filled or the ERP doesn’t export well, any KPI you set up will be decorative.
How many KPIs should a company have?
Between 5 and 9 per area, and no more than 15 on the global dashboard. According to industry studies on management by objectives, teams that monitor more than 20 metrics reduce their ability to react: nobody knows which one to look at when something fails. A typical case: a company measured 34 indicators on a marketing dashboard; after cutting it to 7, weekly meetings went from 90 to 25 minutes because discussions focused on decisions, not on reading figures.
KPI, metric or objective: where is the difference?
A metric is a data point (visits, tickets, margin). A KPI is a metric linked to an objective with a threshold and an owner. An objective is the expected result. Confusing them produces phrases like “we measure engagement,” which says nothing. The test: if you can’t define an alarm value and an owner, it’s not a KPI, it’s curiosity.
A concrete example of a poorly made definition
“Reduce churn” is not a KPI. “Monthly cancellation rate of customers with more than 6 months of tenure, target ≤2.5%, owner: customer success manager, alert if it exceeds 3.5% two months in a row” is. The practical difference: the first version doesn’t allow action; the second triggers an onboarding review.
Which KPIs to choose according to company size?
- Fewer than 10 people: cash, gross margin per line, sales conversion rate and customer churn. Three are enough.
- Between 10 and 50: add customer acquisition cost (CAC), customer lifetime value (LTV), support response time and productivity per employee.
- More than 50: incorporate internal quality metrics (rework, cycle time) and employee experience metrics.
This doesn’t work the same in professional services as in ecommerce. In a consultancy, the key KPI is usually billable hours versus available hours; in an online store, margin per order after returns. Copying another company’s dashboard in the same sector is the most common mistake I see.
How do you implement it without team resistance?
Three steps that prevent failure:
- Automate the capture before asking for anything. If the employee has to fill out a spreadsheet, in three weeks it will be empty. Connect the source (CRM, analytics tools, ERP) and validate one month of historical data.
- Publish first, demand later. Share the dashboard for two or three weeks without evaluating anyone. People need to see that the numbers are fair.
- Define the ritual. A KPI without a periodic meeting dies. A 20-minute weekly review for operational indicators; monthly for strategic ones.
An implementation mistake that keeps repeating
Launching the dashboard the same month the CRM is changed or the sales team is reorganized. The data comes out strange, people distrust it and the project is abandoned. The healthy sequence is: stabilize the source → validate for 30 days → publish → evaluate.
How often are KPIs reviewed and changed?
Monthly review of values, quarterly review of relevance, annual review of the full set. A KPI that has been green for three quarters without moving probably no longer informs: either the target is poorly calibrated or the metric stopped reflecting reality. Changing it isn’t failure; keeping it out of inertia is.
What about qualitative KPIs?
They exist and are useful (customer satisfaction, internal climate), but their measurement is more fragile. A quarterly survey with 20 responses doesn’t sustain a strategic decision on its own. Use them as context, not as the sole trigger. And beware of invented scales without validation: an “internal happiness index” from 0 to 100 with no methodology behind it won’t withstand an audit.
Can they be implemented without paid tools?
Yes, up to a certain volume. A spreadsheet with exported data and a weekly update covers the needs of a small SME. The limit comes with more than 3 sources or more than 5 users: there, manual maintenance consumes more time than it saves. According to official documentation from the main analytics platforms, their native connectors cover the basic cases, but joining different systems still requires technical work.
Quick questions that always come up
Who should own each KPI? One person, not a department. “Marketing” doesn’t answer for it; a person with a first and last name does.
Is it worth measuring everything measurable? No. Measurement has a cost of capture, maintenance and discussion. If the cost exceeds the improvement it enables you to decide on, out it goes.
What if the KPI reveals something uncomfortable? That is exactly its purpose. A dashboard that only confirms what was already known isn’t serving its purpose.