What makes a KPI usable
A usable KPI measures something the employee can influence, is calculated from data you actually have, and is agreed before the period starts. Most failed KPIs fail one of those three: they measure the whole company's result, depend on data nobody collects, or are written in December.
Three to six KPIs per person is usually enough. More than that and each one carries too little weight to matter.
Define it so anyone can calculate it
Write each KPI with the same parts:
- Name: short and plain, such as On-time delivery rate.
- Definition: what it measures, in one sentence.
- Unit: %, a count, a score from 1 to 5, an amount.
- Formula: exactly how the number is produced, such as deliverables on time ÷ deliverables due × 100.
- Source and evidence: where the data comes from and what proves the result.
If a colleague would have to ask you how to calculate it, the definition is not finished.
Targets and thresholds
The target is the result that means the KPI was achieved. Thresholds bound the score: below the minimum the KPI scores nothing, at or above the maximum it scores in full. Between them, the score rises with the result.
Set the target from last period's actual result and the team's plan, not from a round number. A target nobody believes in stops being managed by March.
Weights that add up
Weights say how much each KPI counts towards the overall score, and together they must total 100%. Give the most weight to the results that matter most to the role, and avoid weights below 10%: a KPI worth 5% rarely changes anything.
When a KPI is added or removed during the year, rebalance the weights straight away and record why. A weight total that is not 100% makes every overall score wrong.
Cascading from team to person
Cascading turns team goals into individual KPIs. A department KPI such as On-time delivery rate can apply unchanged to each team member who delivers, or be split into the part each person controls.
Keep the definition identical as a KPI cascades, so results can be added up and compared. When two cascaded KPIs conflict, for example two managers assigning the same KPI with different weights, settle it by a rule everyone knows in advance, and record the decision.
A worked example
An employee has three KPIs: On-time delivery rate (weight 40%, target 95%, minimum 85%), Customer satisfaction score (weight 35%, target 4.5, minimum 4.0) and Process improvements delivered (weight 25%, target 2).
Results: 92% on time, 4.6 satisfaction, 1 improvement. Scoring each against its target and capping at 100%: delivery 92 ÷ 95 = 97%, satisfaction 100%, improvements 1 ÷ 2 = 50%.
Weighted score: 97% × 0.40 + 100% × 0.35 + 50% × 0.25 = 38.8 + 35 + 12.5 = 86.3%. Anyone with the same sheet gets the same number, which is the point.
Your own scoring rule may differ, for example scoring from the minimum rather than from zero. What matters is that the rule is written down and the same for everyone.
Common mistakes
- KPIs the employee cannot influence.
- Definitions that need the author to explain them.
- Targets set after the results are known.
- Weights that do not total 100%.
- Changing a KPI mid-year without recording when and why.
- Too many KPIs, each too small to matter.