Key Performance Indicator (KPI)

Definition

A Key Performance Indicator (KPI) is a quantitative measure tied to a specific objective, baseline, and threshold, used to signal whether a project, programme, or portfolio is performing as planned. KPIs are not metrics in general — they are the small, deliberate subset of metrics chosen because they drive decisions. A dashboard with thirty KPIs has zero KPIs.

Principles

• Tie every KPI to an objective. An orphan KPI is a vanity metric.
• Baseline first, then measure. A KPI without a baseline is a number without meaning.
• Set thresholds before measurement starts. Defining "good" after seeing results corrupts the entire system.
• One owner per KPI. The owner has the authority to act on the signal.
• Five to seven KPIs at any level. More than that, attention collapses.
• Leading and lagging together. Lagging KPIs tell you what happened; leading KPIs tell you what is about to happen.

Leading vs Lagging

The distinction is operational, not academic. Lagging KPIs measure outcomes — CPI, SPI, cost variance, defect count at handover. They are accurate but late. Leading KPIs measure inputs and behaviours — design drawings issued per week, look-ahead schedule adherence, near-miss reports filed, sprint commitment hit rate. They are noisier but earlier. A balanced dashboard mixes both; a lagging-only dashboard tells you when it is already too late to act.

The KPI Hierarchy

1. Portfolio KPIs — strategic, monthly, owned by the steering committee. Examples: portfolio NPV, capital allocation efficiency, strategic objective coverage.
2. Programme KPIs — quarterly, owned by programme director. Examples: cross-project resource utilisation, benefits realisation rate.
3. Project KPIs — weekly, owned by project manager. Examples: SPI, CPI, scope-change rate, RFI ageing.
4. Operational KPIs — daily, owned by discipline lead. Examples: concrete cubic-metres placed, story points completed, NCR closure rate.

Real-World Construction Example

On a USD 480 million LNG storage tank package the project director's dashboard carried seven KPIs: SPI, CPI, drawing release rate against plan, vendor document submission rate, weld repair rate, near-miss reporting frequency, and contractor manpower against ramp curve. Three of these were leading. When the drawing release rate dipped from 96% to 78% in two consecutive weeks, the director re-allocated two senior engineers from a sister project — six weeks before the SPI would have shown the slippage. The intervention avoided an estimated 11-week delay on tank inner-shell welding. The seven KPIs were chosen deliberately from a starting list of 34; the discipline was the cut.

Real-World IT / Agile Example

A platform team's dashboard once displayed 22 metrics, all green, while customer satisfaction quietly cratered. The team reset to four KPIs: deployment frequency, change-failure rate, mean time to recovery (MTTR), and net promoter score (NPS) — the DORA four plus customer signal. Within two quarters the platform reliability incidents had dropped 60% because attention was concentrated. Most dashboards fail by addition; the best fail in the opposite direction and add the right thing.

Project Controls Perspective

From a controls standpoint, the integrity of KPIs depends on three things: the integrity of the baseline they reference, the integrity of the data feeding them, and the integrity of the threshold logic. A pristine SPI calculated against an unrealistic baseline is misleading; an honest baseline measured with corrupt actuals is worse. Controls teams therefore audit not only the metric value but the upstream data flow on a quarterly cycle.

Common Mistakes

• Too many KPIs. Twenty KPIs equals zero KPIs; attention is finite.
• Lagging-only dashboards. By the time CPI shows red, the recovery cost is multiplied.
• Vanity KPIs — measures that look good but drive no decision (e.g. "training hours delivered" without competency outcomes).
• Thresholds set after measurement begins, allowing the team to grade its own paper.
• KPI gaming — selecting the metric so that the easiest path to a green light is also the path of least real progress.
• No owner, or an owner without authority to act.

Expert Tips

• Start with the decision, not the data. Ask "what decision will this KPI trigger?" If the answer is "none," delete it.
• Pair every lagging KPI with at least one leading partner. CPI travels with productivity rate; SPI travels with look-ahead adherence.
• Refresh thresholds annually, not monthly. Frequent adjustment is gaming in slow motion.
• Publish KPI definitions, owners, and data sources alongside the dashboard. A KPI without a written definition is open to interpretation.
• Run a quarterly KPI audit. Which KPIs drove a decision in the last 90 days? Which did not? Retire the dead.

Key Takeaways

• A KPI is a metric that drives a decision; if it doesn't, it isn't.
• Five to seven KPIs per level — discipline of subtraction beats discipline of addition.
• Mix leading and lagging; lagging alone is a post-mortem dashboard.
• Baselines, thresholds, owners, and data sources are non-negotiable.
• Audit the dashboard quarterly; retire dead metrics ruthlessly.

Related Concepts

KPIs interlock with Earned Value, OKRs, S-Curves, Risk Management, and Lessons Learned. KPI libraries and dashboard templates are available at PMMilestone.org.