Skills on AI 459 skills

Key Performance Indicators

Use when asked to define, choose, or improve a Key Performance Indicator (KPI) — leading vs lagging indicators, measure vs metric, or how to check whether a proposed KPI is actually good — grounded in joelparkerhenderson/key-performance-indicator, as distinct from OKRs (see objectives-and-key-results).

A Key Performance Indicator (KPI) is a type of performance measurement that evaluates the success of an organization or activity. Choosing the right KPIs requires genuinely understanding what matters to the organization — a KPI list built without that understanding measures whatever's easy to measure instead of what's actually important.

How to define a KPI

A well-specified KPI covers:

  • Title — an exact, unambiguous name.
  • Objective — its clear relationship to an organizational objective (see Objectives and Key Results).
  • Scope — which parts of the business/organization it includes or excludes.
  • Target — the benchmark used to judge progress.
  • Formula — the exact calculation.
  • Units — the unit(s) of measurement.
  • Frequency — when it's recorded and reported.
  • Data source — exactly where the underlying data comes from.
  • Owner — the accountable person.
  • Comments — outstanding issues with the indicator itself.

A KPI missing several of these (especially Formula, Data source, and Owner) tends to be reported inconsistently across teams, or to quietly drift in meaning over time.

How to judge whether a KPI is good

Ask: does it clearly define success? Does it clearly relate to a strategic OKR? Does it support setting Smart Criteria-compliant goals? Does it accurately reflect progress toward both long-term objectives and near-term milestones? Does it identify root causes of barriers? Does it focus the organization on priority improvement needs? Does it drive the behavior actually needed to achieve the objective? Does it align work with value? A KPI that fails most of these is a number being tracked, not a genuine performance indicator.

Indicator types

  • Quantitative (a number) vs. Qualitative (not numeric).
  • Leading (predicts an outcome, actionable ahead of time) vs. Lagging (reports success or failure after the fact).
  • Input (resources consumed) vs. Process (efficiency/productivity of the activity) vs. Output (the result of the process).
  • Practical (interfaces with existing processes), Directional (shows whether something is getting better or worse), Actionable (sufficiently within the organization's control to actually change), Financial.

A Key Leading Indicator (KLI) is a KPI that tends to show up earliest — worth identifying specifically, since it's the one that gives advance warning before a lagging indicator confirms a problem has already happened.

Measure vs. metric

  • A measure is concrete, usually captures one thing, quantitative ("I have five apples").
  • A metric describes a quality and requires a measurement baseline to be meaningful ("I have five more apples than yesterday").

Measures are useful for demonstrating workload/activity; metrics are useful for evaluating compliance, process effectiveness, and progress against objectives — conflating the two (reporting a raw measure as if it were a comparative metric, or vice versa) muddles what a number actually tells you.

The 7 habits of highly effective KPI users

KPIs help lead, not just manage — they should help anticipate the future, not just report the past. KPIs align the organization around shared priorities. KPIs should provide an integrated view of the customer. Real-time KPI analysis enables better prioritization than retrospective-only analysis. KPI data should be shared across business units, not siloed. KPIs shouldn't be allowed to proliferate indiscriminately — there's no magic number, but fewer, better KPIs beat many mediocre ones (a rough rule of thumb: a small handful of enterprise KPIs plus a small handful of functional ones, not dozens of each). And KPI data can itself be a training set for improving decision models over time.

Common pitfalls

  • A KPI with no named owner or exact formula — leads to inconsistent reporting and disputes about what the number actually means.
  • Confusing a measure for a metric — reporting "5 deployments this week" (a measure) as if it demonstrated improvement without a baseline to compare against.
  • Tracking only lagging indicators — gives no early warning; pair every important lagging indicator with at least one leading indicator that predicts it.
  • Too many KPIs — a long dashboard of indicators dilutes focus and makes it hard to tell which few actually matter for the next decision.
  • A KPI the organization can't actually act on — an indicator outside anyone's real control (e.g. a macroeconomic figure) can be worth watching as context, but shouldn't be treated as an actionable performance indicator for a team.

Learn more

View key-performance-indicators/SKILL.md on GitHub