A measure that has never changed a decision is not evidence. It is overhead with a chart attached, and somebody is paying to produce it every month.
Ask for the workforce report and you will be handed close to thirty numbers. Ask which of them changed a decision in the past twelve months and the room produces two, occasionally three, and a pause long enough to be informative.
The pause is not evidence of a weak people function. It is the predictable output of how measurement sets get assembled: added one at a time, each defensible in the meeting where it is proposed, almost never removed, because removal requires somebody to argue that the enterprise should know less about itself than it did last quarter. Nobody holds a view of the whole.
That would be a minor problem if measurement were free. Every measure carries a collection cost, a distortion cost and an attention cost, and only the first ever reaches a budget line.
The Strategic Context
Most workforce reporting is built to describe the organisation and is never tested against what the organisation then does differently. Hence a familiar asymmetry: the people function can give you turnover to one decimal place across nine cost centres, and cannot tell you which of the last three retention interventions was worth what it cost.
A governance problem sits inside the volume as well. An executive committee reviewing thirty workforce indicators is not exercising oversight; it is receiving a briefing. Oversight requires a threshold at which somebody must act, and a named person who must act at it. Few workforce measures carry either.
Why More Measures Feel Like Better Management
Two beliefs do most of the work here, and each contains something true.
The first is that more measures reduce risk. In a portfolio of assets, diversification does reduce risk. In a portfolio of measures the opposite holds, because executive attention is the binding constraint. Thirty indicators presented monthly do not produce thirty considered judgements; they produce a scan, and the scan settles on whichever number moved most, which is usually the noisiest rather than the most consequential.
The second is that measurement is objective and therefore fair. It is neither, automatically. A measure is a choice about what to count, and the judgement sits in the choice rather than the arithmetic.
Reframing the Issue
Treat measurement as a portfolio rather than a collection and the useful questions arrive at once: total cost, deliberate composition, a hurdle rate, a disposal discipline. A collection has none of those, which is why it only grows.
The hurdle rate is sharpest — a measure earns its place if it changes a decision that a named person owns, at a threshold stated before the number is seen. Composition matters next, because a well-composed set covers the employee, the manager, the system the work runs through, and the quality of the decisions made about people. Almost every set in circulation covers only the first.
Revenue and profit per employee deserve a further caution, because both are read as proxies for enterprise value and both are weak ones. What a board or a buyer values is whether the earnings survive the departure of particular individuals — a question about transferability, not headcount [Related article: The Business That Cannot Run Without You Is Not an Asset].
The Omission That Tells You Who Is Being Measured
Set out the standard workforce set as commonly assembled and count what is in it. Revenue per employee. Profit per employee. Cost per hire. Time to hire. Time to fill. Offer acceptance. Turnover. Retention. Early attrition among new starters. Retention of top performers. Task execution rate. The proportion rated above average. Overtime as a share of payroll. Incentive payout. Training spend per head. Absenteeism. Candidate experience.
Seventeen measures. Every one of them is measured on the employee.
Not one measures a manager. Not one measures a decision. Not one measures the system the work runs through. Turnover is attributed to the population that left rather than the conditions they left; execution rate to the person completing tasks rather than whoever specified them, sequenced them and cleared the dependencies. That is not an oversight in one dashboard. It is a theory of causation, adopted silently, in which performance is a property of individuals and the organisation the neutral background against which they succeed or fail.
W. Edwards Deming argued that the greater part of variation in performance is attributable to the system in which people work rather than to the people themselves [SOURCE DETAILS REQUIRED]. If that holds even partially, a set with no manager-side or system-side term is pointed at the wrong variable, and every intervention it prompts will be aimed at individuals for conditions they do not control.
The counterweights are not exotic — dispersion of turnover between managers running comparable work, or elapsed time from an issue raised to a decision returned. They are absent because the set was built to look downward.
The omission propagates into succession. Where the only durable record of a person is their performance in the role they currently hold, promotion will be decided on that record and the requirements of the receiving role will go unexamined [Related article: Which Anchor Does This Role Actually Require?].
What a Measure Costs Before It Tells You Anything
Collection cost is the visible one and the smallest: analyst time, system configuration, survey administration, reconciliation when two systems disagree about headcount. Distortion cost is larger and arrives later, because every published measure creates an incentive to move the number, and the cheapest route to moving a number rarely improves the underlying thing. Attention cost is largest and the only one genuinely irrecoverable: a measure that cannot change a decision spends the scarcest input the organisation has.
Philip Crosby argued that quality should be measured by the price of non-conformance rather than by indices and ratings [SOURCE DETAILS REQUIRED]. The translation is uncomfortable: an executive team is better served by the annual cost of unplanned departures in a critical function, stated in money, than by a retention percentage. The percentage is comparable; the money is decidable.
Take a deliberately hypothetical case. A mid-sized services business runs a monthly pack of twenty-eight indicators, consuming four days of analyst time and part of a day from six line managers who supply inputs — on the order of a hundred person-days a year in this illustration, of which two carry thresholds. The figures are invented, but the burden is real and invisible, because it falls on people whose time is never costed to it.
Measures Are Instructions, and People Follow Them
Charles Goodhart's observation about monetary policy — that a statistical regularity tends to break down once adopted as a target — has been generalised well beyond its original setting [SOURCE DETAILS REQUIRED]. The mechanics deserve to be concrete, because "gaming" implies bad faith and most distortion is honest.
A task execution rate is improved most cheaply by making tasks smaller, and nobody has to decide to do it; teams drift toward finer granularity because it makes reporting easier, and the measure rises while throughput does not. Time to hire measured from the first interview conceals the sourcing delay entirely. Time to fill, measured from the day the position opened, is the honest clock, and the one more often omitted.
Absenteeism defined as unexcused absence rewards attendance rather than capacity to work, and pushes cost into presenteeism where it stops being visible. It also touches genuine legal duty: pressure applied through attendance metrics can bear on psychosocial hazard obligations under Australian work health and safety law [FACT CHECK REQUIRED]. That requires professional verification — ERANORTH is not a law firm and none of this is legal advice.
Experience scores built from promoter and detractor bands warrant a specific warning. Versions in common circulation disagree about where the bands sit and even about where the scale begins, so a composite built on them is not comparable across periods, across business units, or against any external figure. No set of bands should be treated as standard.
None of this argues for measuring less honestly. It argues that the distortion route for every published measure should be named in advance and monitored, which requires someone to say out loud how the number could improve without the business improving. That conversation is only possible in an organisation that can hold disagreement about its own reporting, and many cannot [Related article: Your Organisation Has a Default Conflict Style. Who Chose It?].
Four Techniques, One Governance Question
Multi-source feedback from managers, peers, subordinates and customers. Per-employee profit attribution, setting an individual's cost against value assigned to them. Activity tracking — calls placed, visits made, handling times, resolution rates. Behavioural trait rating for attitude, loyalty and cultural fit.
They are adopted separately, by different functions, under different justifications. They should be governed as one question, because from the position of the person being measured they are one thing: the volume and intimacy of what the organisation records about an individual, and the uses to which it may be put.
That question has five parts. What is collected about a named person. Who can see it. Which decisions it may legitimately inform. How long it is retained. What recourse the person has to see it, correct it, or contest a decision made on it. Handling of personal information in Australian workplaces engages the Privacy Act and the Australian Privacy Principles, and surveillance of workers is separately regulated in some jurisdictions and not others [FACT CHECK REQUIRED]. This requires professional verification before any collection programme proceeds.
Three practices from the same tradition require a firmer position. Forced ranking, separation of a bottom quartile on that ranking, and blanket zero-tolerance policies are often presented as decisive management. In Australia they intersect with unfair dismissal and general protections provisions, with discrimination law, and with psychosocial hazard duties [FACT CHECK REQUIRED]. They require professional legal verification and are not recommended here. The commercial objection stands independently: a distribution imposed by policy rather than observed produces a number that describes the policy, not the workforce.
Decision Framework
Apply seven tests to every measure, and retire anything that fails three.
| Test | Question | Retain when | Retire when |
|---|---|---|---|
| Decision | Which decision does this change? | A specific decision with a named owner | It is reported for awareness |
| Threshold | At what value must somebody act? | A trigger is written down in advance | The answer is "it depends" |
| Controllability | Can the party measured act on the cause? | The cause sits inside their control | It aggregates causes they cannot separate |
| Cost | What does one reporting period cost? | Cost is below the value of the decision it serves | Cost has never been estimated |
| Distortion | How could this number improve without the thing improving? | The route is known and monitored | The route is unexamined |
| Comparability | Are definitions fixed across periods and units? | Written, versioned definitions exist | Definitions vary by system or source |
| Expiry | When is the decision to keep this reviewed? | A review date exists | It has run unchanged for years |
Cap the set alongside the tests — a dozen measures to the executive, five to the board. A cap converts every addition into a trade-off and forces the composition question an uncapped set lets everyone avoid.
From Strategy to Execution
The immediate work costs nothing but candour. Mark against each measure the last decision it changed and who owned it. Measures with no entry are not worthless, but they become visible as context rather than evidence, and the pack can be reorganised to say so. Publish the collection cost in person-days alongside it.
The medium-term work is compositional. Build the manager-side and system-side measures the set is missing. Then write and version definitions for everything retained, because comparability is manufactured by documentation and nothing else.
The long-term position is a change in what measurement is for. A portfolio that is designed rather than accumulated becomes an instrument of decision quality: fewer numbers, each with a threshold, an owner and a known distortion route. Honesty improves as a by-product, because there are few enough to examine how each is produced.
Signals to Monitor
Watch the count of measures reported and the direction it moves; a set that only grows is not being managed. Watch the proportion carrying a written threshold. Watch for improvement that no operational observer recognises, the clearest signature of distortion. Watch the interval between an intervention and any measured effect, and whether anybody checked. And watch external developments: changes to work health and safety regulation, privacy law and workplace surveillance rules can convert an established practice into an exposure [FACT CHECK REQUIRED].
Questions for the Leadership Team
- Of the measures in our workforce pack, which changed a decision in the past twelve months, who owned that decision, and can we name it?
- What does producing this pack cost each year in person-days, and who has ever been asked to estimate it?
- Which of our measures assesses a manager, a decision or the system rather than an employee — and if none, what theory of performance are we running on?
- For each published measure, what is the cheapest way to improve the number without improving the underlying thing, and who monitors that route?
- What do we collect about an individual across all systems combined, and what would we say if that person asked to see it?
Closing Perspective
The number of measures an organisation reports is a poor indicator of how well it understands itself and a good indicator of how it resolves internal arguments. Every measure was added because somebody wanted visibility; almost none were removed, because removal requires a judgement about relative importance that nobody wants to defend.
The choice is narrower than it first appears. You are not choosing between measuring more and measuring less, but whether measurement is an instrument of decision or a demonstration of diligence — and the two produce different sets, different costs, and different behaviour in the people who know they are being counted. An organisation that measures thirty things and decides on two is not better informed than one that measures five and decides on five. It is paying more for the same decisions, and pointing the whole apparatus at the people least able to change what it records.
About the author
Kevin Jogin is Founder & Principal Advisor at EraNorth. Meet the Founder.