A dashboard can contain hundreds of metrics and still leave a team unsure whether the product is succeeding. Page views describe activity, response times describe service speed and renewals describe a commercial outcome. Calling all three “key” does not establish which decision they should guide or what improvement would mean for customers.
A key performance indicator, or KPI, is a measure selected to assess progress towards a particular objective. The selection gives it a role: people agree why it matters, how it is calculated and what they will investigate or change when it moves.
Begin with the outcome you want to recognise
Imagine a fictional support team aiming to resolve customers’ problems more quickly. Average ticket-closing time is easy to measure, but it can improve if staff close difficult cases prematurely. Looking at repeat contact or reopened cases alongside resolution time gives the team a better chance of noticing that failure.
The example illustrates why a KPI needs an explanation connecting the measure to the objective. The GOV.UK Service Manual’s guidance on measuring success provides a useful starting point for relating performance measures to the purpose of a service. The particular indicators should follow from that purpose rather than from the default charts in an analytics tool.
Some measures describe outcomes that have already happened, such as renewals. Others are candidates for earlier signals, such as successful completion of an initial task. An early signal becomes useful when there is evidence that it helps anticipate the later outcome; it is not inherently predictive because a team has labelled it “leading”.
Write a definition another team could reproduce
For a rate, specify the numerator, denominator and time window. Decide how to handle duplicate accounts, missing events and people who have not yet had enough time to qualify. Identify an owner for the definition and record changes so that a revised calculation does not masquerade as improved performance.
A target also needs context. Aiming for 80% completion means something different from a baseline of 40% than from one of 79%. Explain whether the target reflects customer need, a practical improvement goal or a relevant comparison. Avoid presenting a convenient round number as a natural boundary between success and failure.
Check what the pursuit of the number changes
Once a metric carries incentives, people may find ways to raise it without improving the intended outcome. A product can increase tutorial completion by making the tutorial compulsory, while making the first experience more frustrating. The measure has moved; the underlying benefit remains in question.
Use companion measures to detect obvious trade-offs, and periodically examine the experience directly through user research. A qualitative account can reveal that “successful” customers are relying on an exhausting workaround that the dashboard never records.
Keep the set of KPIs small enough to influence decisions, while retaining supporting metrics for diagnosis. When a KPI stops reflecting the objective, revise or retire it openly. Its usefulness depends on the connection to the outcome, not on how long the organisation has reported it.