Home›Blog›Project KPIs: the families, and the question each answers
FundamentalsBeginner
Published September 12, 2026 · By Vincent KENNEL
Project KPIs: the families, and the question each answers
Four serious referentials cut project indicators into four different sets of families, and the vocabulary standard of the discipline does not define the word at all. What settles the value of an indicator is a second axis, and it is not the one the market uses.
In brief
Project KPIs are commonly grouped by what they measure, schedule, cost, progress, quality and risk, and no standard imposes that list. A steadier reading follows the three objects ISO 21502:2020 assigns to reporting: an indicator says where a project stands, what is drifting, or where it will land.
On a large programme under contract, the review dashboard carries some twenty lines, and every one of them is exact. Nothing in it says which line to read first, and nothing says what none of the lines shows. The usual answer to that discomfort is to add another line. This article does not judge the reflex. It gives the two questions that make an existing dashboard readable: what each line answers, and what it cannot see.
What a project KPI is, and what a metric is
The acronym arrives before the definition, and it is not the referentials' word. Key performance indicator appears twice in the PMBOK Guide, 6th edition (2017), in passing and with no glossary entry, and seven times in the PMBOK Guide, 7th edition (2021), which gives it a section of its own. Outside PMI it is far from universal. PRINCE2 7 (2023) carries no occurrence of the phrase at all and speaks of tolerances instead, and GovS 002 version 2.1 (2025), across some 153,000 characters, writes performance indicators and never once key performance indicator.
Where the PMBOK Guide, 7th edition (2021), is precise is on a distinction the market treats as a matter of vocabulary. A metric is "a description of a project or product attribute and how to measure it" (2.7, p. 93). A KPI is narrower: a quantifiable measure used to evaluate the success of the project (2.7.1.1). Metric and KPI are two objects and not two words, and the KPI is the subset qualified by that purpose. The APM Body of Knowledge, 8th edition (2025), agrees in its glossary, defining KPIs as measures of success usable throughout the project.
Three components come out of those clauses, and this article assembles them into one sentence: a project indicator is the measure of an attribute of the project, together with the way of obtaining it, retained in order to inform a decision. The purpose is carried by the same text: clause 2.7.6 of the PMBOK Guide, 7th edition (2021), asks that only what helps to learn, decide, improve, avoid a problem or prevent a degradation be measured and published, on the ground that measuring consumes time and effort taken from other productive work.
That sentence is a composition, and it is worth saying why no standard is credited with it. ISO 21506:2024, clause 3, numbers one hundred entries from 3.1 to 3.100, and not one of them is indicator, key performance indicator, measure or metric. ISO 21502:2020 uses indicator exactly once, in a peripheral sense, about confidence in estimate ranges. The vocabulary standard of the discipline does not define the word the discipline uses every day.
If no standard imposes the word, nothing imposes the list either.
The main families, and why no list is authoritative
Put the question to a practitioner and the answer comes back fast: schedule, cost, progress, quality, risk. The list deserves to be stated, because it is what circulates and what makes a conversation possible. It is a shared reading convention rather than a normative taxonomy, and no referential in the corpus prints it.
What the referentials print is four different cuts.
PMBOK Guide, 7th edition (2021) seven categories of measurement
PRINCE2 7 (2023) seven aspects of performance placed under tolerance
ISO 21502:2020, 6.6.3 fourteen items of progress and performance data to collect
NDIA, Rev 3 (2021) eight families of predictive measures
deliverable, delivery, baseline performance, resources, business value, stakeholders, forecasts.
Neither quality nor risk is a category here. Quality enters through the deliverable metrics; risk does not appear at all.
benefits, costs, time, quality, scope, sustainability, risk.
A forecast breach of the range is what triggers an exception. Sustainability is the seventh, added with version 7.
work completed, milestones achieved and costs incurred; benefits planned or realized; managing the scope; acquiring sufficient resources to complete the work; managing the schedule and costs; identifying and managing risks and issues; managing change control; quality of work; stakeholder engagement and communications; transition of outputs; reporting on progress; information and documentation; procurement activities; new lessons learned.
Four organisations, four counts, no complete overlap. The reading that holds is the one that stays with what each document says it is doing. PRINCE2 7 (2023) lists what is placed under tolerance, and therefore under authorisation. ISO 21502:2020 lists what has to be collected. The NDIA lists what a customer expects a contractor to report. Nothing here makes one of them better than another, and none of the four is out of date. The divergence is the fact, not the grievance.
One standard already ranked indicators by something other than their object. ISO 21508:2018, informative Annex A, table A.1, separated key indicators, CV, SV, CPI, SPI and VAC, from predictors, TCPI and IEAC. The split was not by what is measured. It was by the direction of the look. That standard has since been withdrawn, which is worth saying because the reading below leans on it.
What an indicator is for: the question it answers
A project report carries three things. ISO 21502:2020, 7.15.1, says so in a NOTE: "Reporting focuses on providing the status, analysis of variances, and forecasts of future performance for the project."
From there this article takes one step, and the step is ours. If a report has three objects, the indicators that feed it can be sorted by the one they serve. Three questions, then: where do we stand, what is drifting, where will we land. This is a reading proposed here, not a classification that any standard writes.
It does not stand unsupported. ISO 21508:2018 already institutionalised two of the three by separating key indicators from predictors. And the gesture has a precedent: Kaplan and Norton (1992) framed their four perspectives as questions rather than as categories, and Kaplan (2010) reports that Simon and colleagues (1954) were already sorting management information by the question asked, scorecard, attention directing and problem solving. Simon's triplet is not this one, its third category does not overlap with ours, and everything known of it here comes through Kaplan.
Figure 1: the two axes, and the cells the referentials leave empty
The grid shows what the referentials read here name, not an inventory of what exists.
The second axis also reopens a point the market treats as settled. The PMBOK Guide, 7th edition (2021), 2.7.1.1, states that "There are two types of KPIs: leading indicators and lagging indicators". The NDIA, Rev 3 (2021), section 1, says the opposite on an example: actual staffing below plan is a leading indicator that future planned work will not be accomplished, and the same indicator is a lagging indicator that the hires or transfers did not happen. Two institutions, two positions, and this article leaves them side by side. What follows from the example is only this: the type is not a property of the indicator, it depends on the direction of the look, and therefore on the question asked. Three vocabularies name the same idea, leading and lagging at PMI, predictive measures at the NDIA, key indicators and predictors in ISO 21508:2018.
On a project under contract, what a review body looks at depends on the question it is asking that day, not on the row of the table.
What each family measures, and what it cannot see
An indicator can be exact and still mislead, and three distinct mechanisms explain how.
The first is arithmetic. The NDIA, Rev 3 (2021), 2.1, documents it on the schedule performance index. By the nature of the formula, however early or late a programme finishes, the index ends at 1.00, and over the final third of a project its utility degrades. It is an average across the whole project, so it can be pulled up by work on non-critical tasks while the critical path drifts, and level of effort dampens it further. The guide is explicit that the index is to be used alongside critical path analysis and schedule risk assessment, and "never as a stand-alone indicator of the health of a program". Above 1.05 it marks the value as suspiciously good, a sign of over-cautious planning or of cherry picking. How a value is then read belongs to another article.
The second mechanism is what the indicator does not connect. Rigney (1979) put it plainly: conventional analysis indicates how much has been spent and how much is left, and does not indicate how much has been done, nor whether the remaining resources are sufficient to finish. Cost status and schedule status were both handled, and never linked to each other.
The third is what the indicator provokes. The PMBOK Guide, 7th edition (2021), 2.7.4, names the Hawthorne effect, where the act of measuring influences behaviour; the vanity metric, which shows data without providing information useful to a decision; and the gaming of the indicator, including work on out-of-sequence activities that are easy to finish. Kaplan and Norton (1992) arrive at the same place from the other side, warning that "reliance on one instrument can be fatal" and noting that a balanced set exists precisely to make sub-optimisation visible.
None of this is a verdict on any reporting system. It is a property of the instrument.
Where the families came from
The families did not come from project management. They came from management control, and the account of them is Kaplan's.
Kaplan (2010), Harvard Business School Working Paper 10-074, opens his own genealogy by writing that the Balanced Scorecard "was not original" in advocating that non-financial measures be used to motivate, measure and evaluate company performance. What he then describes, at 1.1, pages 4 and 5, is a set built at General Electric in the 1950s for a precise problem: steering decentralised divisions without relying on the financial result alone. Eight measures, one financial and seven not. Profitability as residual income, market share, productivity, product leadership, public responsibility, personnel development, employee attitudes, and the balance between short-range and long-range objectives. Kaplan maps the eight onto his own four perspectives and writes that the eighth carries the essence of the Balanced Scorecard.
Kaplan also tells what became of that set, and he tells it without sourcing it: the goals of the corporate project never became part of the management system and the incentive structure of the business units. A set balanced on paper, undone by what surrounded it.
Thirty-seven years later, Kaplan and Norton (1992) published in the Harvard Business Review under a sentence that has outlived the article, "What you measure is what you get", and asked their reader to "think of the balanced scorecard as the dials and indicators in an airplane cockpit".
Between those two dates, nothing in the corpus attributes the families of indicators to project management. They were borrowed, and they were borrowed from a discipline that had a different problem to solve.
Even where performance measures are instituted purely for purposes of information, Ridgway (1956), p. 247, holds that they are probably interpreted as definitions of the important aspects of the job, and therefore carry consequences for the motivation of behaviour.
In short
The reusable move is two questions asked of every line of a dashboard. Which of the three does the line answer: where do we stand, what is drifting, where will we land. And what is it blind to by construction.
The second question is the one only the end of this article can supply. An indicator can be exact and useless, and it can mislead without being false, because what it cannot see is a property of the way it is built rather than a defect in the data. In the reading proposed here, a line that answers none of the three serves none of the three objects of the report.
Read that way, the disagreement between referentials stops being an obstacle. It matters little that one of them counts seven categories of measurement and another eight families, because the first axis was never the one that decides what an indicator is worth. Which line to remove from an existing dashboard is a different question, and this article does not settle it.
Stop guessing. See the real impact.
Frequently asked questions
Q.How many KPIs should a project track?
Mosaic Projects recommends a maximum of six to eight metrics, and three to five as better (Weaver, Gateways and Score Cards, consulted on 28 August 2026). The number matters less than the target: a raw measure says little until it is compared with a target set in advance.
Q.Is a percentage of completion a KPI?
It is, and it is the family a review body looks at first. ISO 21502:2020 lists work completed among the items on which the project manager collects and analyses data. The rule that turns work in progress into a percentage is a subject of its own, and it belongs to another article.
References
Administrative Science Quarterly - V. F. Ridgway - Dysfunctional Consequences of Performance Measurements - 1956
Controllership Foundation - Herbert A. Simon, Harold Guetzkow, George Kozmetsky, Gordon Tyndall - Centralization vs. Decentralization in Organizing the Controller's Department - 1954
Harvard Business Review - Robert S. Kaplan, David P. Norton - The Balanced Scorecard - Measures that Drive Performance - Reprint HBR 92105, 1992
Harvard Business School - Robert S. Kaplan - Conceptual Foundations of the Balanced Scorecard - Working Paper 10-074, 2010
ISO - ISO 21502:2020 - Project, programme and portfolio management, Guidance on project management - 1st edition, 2020
ISO - ISO 21506:2024 - Project, programme and portfolio management, Vocabulary - 1st edition, 2024
ISO - ISO 21508:2018 - Earned value management in project and programme management - 2018
Mosaic Project Services - Patrick Weaver - Gateways and Score Cards - Measuring Project Performance - Consulted on 2026-08-28
NDIA - A Guide to Managing Programs Using Predictive Measures - Revision 3 - 2021
PMI - A Guide to the Project Management Body of Knowledge (PMBOK Guide) - 7th Edition - 2021
PMI - A Guide to the Project Management Body of Knowledge (PMBOK Guide) - 6th Edition - 2017
R. A. Rigney - Performance Measurement System: Recent Systems Development and Applications - 1979
UK Government - Government Functional Standard GovS 002 - Project delivery - Version 2.1, 2025
