Haute Lumière

Commerce · II.10 · MMXXVI · daylight

La Bourse  /  Volume II  /  Nº II.10  /  Workbook — the executive

WORKBOOK — THE CORPORATE EXECUTIVE

Chapter II.10 · Measurement and What It Does

For the person who owns a P&L, writes board papers, and has spent a career being told that what gets measured gets managed. It does. That is the problem and it is also the instrument.

---

THE PREMISE, STATED COMMERCIALLY

You already believe Campbell's law. You have watched a sales target produce quarter-end stuffing, a utilisation target produce work that should not have been billed, a safety metric produce under-reported incidents, and an NPS target produce a script. You did not need a citation.

What you have not had is a way to put a number on it, and the absence of a number is why the conversation always ends in a judgement call. This workbook supplies three things your reporting pack does not currently contain:

1. Boundary statements on the numbers you already publish — free, and they change the quality of the discussion in one cycle.
2. A holdout indicator beside your most consequential target, which converts "I think this is being gamed" into a percentage.
3. A step-up inequality that tells you, in basis points, what a performance-linked incentive has to be worth before it deters the cheaper route.

None of it requires a new system, a consultant, or a change to the general ledger. All of it can be done inside one reporting cycle by people you already employ.

---

PART ONE — DISCOVERY

Days 1–30: read your own pack properly

Exercise 1.1 — The pack audit (half a day, with your FP&A lead)

Take the standing management pack and, for each headline number, complete three columns:

Do not editorialise. Fill it in. Two findings appear reliably.

The first: at least one number in the pack has an exclusion nobody currently in the business decided. It was inherited from a system implementation, a previous auditor, or a definition written for a business unit that no longer exists. That number is being read as a description of the company and is actually a description of a decision made by someone who has left.

The second: at least one number has quietly expanded beyond its boundary. It was built to answer a narrow question and is now quoted in a wide one. This is the corporate form of GDP's own history — a precision instrument answering a question nobody alive chose.

Exercise 1.2 — The appreciative measurement review (90 minutes, leadership team)

Ask the room, in this order, and write the answers down:

"Which number in this pack do you genuinely trust? What is it about how it is compiled that earns that?"

"When has a measure in this business helped somebody do something they were proud of?"

You are collecting the conditions under which measurement works here: separated compilation, a stable definition, a long series, an owner who does not benefit from the result. Those four conditions are your design brief for everything below.

Exercise 1.3 — The uncounted inventory (half a day)

List what the business relies on and does not measure. Typical entries: maintenance knowledge held by three people; the supplier relationship that absorbs your demand volatility; the training that is delivered informally on the line; the reputational permission that lets you launch without explaining yourself.

Cost one of them at replacement value. Not to capitalise it — to know what it is worth if it goes. The national accounts exclude household production worth 25.7 percent of United States GDP for exactly the reason you exclude these: no transaction, therefore no entry.

---

PART TWO — THE ARITHMETIC

Days 31–50: put a figure on the gaming

Exercise 2.1 — Estimate your own Goodhart factor (2 hours)

Take your most consequential target. Find the pre-target period in which the metric and the outcome you actually want moved together, and estimate the correlation ρ.

  before the target:  a 1 s.d. rise in the metric predicted ρ² s.d. of real gain
  ρ = 0.9  ->  0.81        ρ = 0.8  ->  0.64        ρ = 0.7  ->  0.49

Now estimate what fraction of the post-target movement you believe is real. If it is a fifth, the naive reading of the number overstates the gain by 4.05× at ρ = 0.9. Write that factor on the slide beside the metric. You have just converted a private scepticism into a stated assumption, which is the only form in which it can be tested.

Exercise 2.2 — The depletion analogue in your own accounts (half a day)

The national accounts charge depreciation on a lathe and nothing on an oil field. Repetto's recomputation of Indonesia charged depletion of petroleum, forests and soils and took reported growth from 7.1 percent a year to 4.0 percent — over thirteen years the conventional account ends 46.5 percent larger than the adjusted one.

Do the same to yourself. Identify the stocks your P&L draws on and does not charge:

• Deferred maintenance on plant, property and tooling.
• Technical debt, at the cost of the remediation programme you keep postponing.
• Training debt — the gap between the capability you have and the capability the plan assumes.
• Customer goodwill drawn down by price or service actions.
• Supplier goodwill drawn down by payment terms.

Charge each at a defensible annual rate and restate three years of operating profit. You are not proposing to change the statutory accounts. You are producing a management view, once, to see the size of the gap. In most businesses the gap is not small and nobody has ever put it on one page.

Exercise 2.3 — The step-up inequality (90 minutes, with treasury)

Any incentive linked to a metric — a bonus pool, a supplier rebate, a sustainability-linked margin ratchet — deters the cheaper route only if it exceeds what the cheaper route saves:

                cost of real compliance  −  cost of gaming
  step-up  >  ---------------------------------------------
                       notional  ×  duration

Worked on a business-unit facility: notional $40,000,000, duration 3.0 years, real compliance $900,000, gaming $120,000. The required step-up is 65 basis points, against a market convention of 25 — short by 2.6×.

On the chapter's larger shape — $500,000,000 over 5.0 years, real compliance $12,000,000, gaming $1,000,000 — the required step-up is 44 basis points, the standard 25 is worth $6,250,000 against a gaming advantage of $11,000,000, and the deterrence shortfall is $4,750,000.

Run it on every metric-linked incentive you have. The conversation that follows is about basis points, and basis points get approved.

---

PART THREE — DESIGN

Days 51–70: the three instruments

Instrument one — boundary statements on the standing pack.

Three lines above every headline number: counts, excludes, decision rule. Owner: the FP&A lead. Cost: one afternoon. Effect within two cycles: questions in the management meeting move from is this right to is this the right boundary for what we are deciding, which is a materially better meeting.

It also has a defensive property worth naming to your audit committee: a number whose exclusions are written down cannot silently expand, because expansion now requires somebody to edit the header, and somebody will object.

Instrument two — the holdout indicator.

For your single most consequential target, name a second indicator that tracks the same underlying reality and that is contractually excluded from every incentive plan for the life of the target. Publish both in the pack.

Then compute, every quarter:

  targeted gain     18.0 % of the base
  holdout gain       4.2 % of the base
  divergence        13.8 percentage points
  real share        23.3 %

That real-share figure is the output of this entire workbook. It is the number you have never had. And note what it does not say — it does not say the programme failed, and it does not accuse anybody. It says how much survived contact with an untargeted measure.

Instrument three — the parameter register.

Every composite you use — a supplier scorecard, an internal ESG rating, a customer health score, a risk heat map — gets one page: the parameters, their values, who set them, when, and the result at two other plausible values.

The chapter's evidence for why this matters: the Human Development Index's weights imply a year of life expectancy is worth $68 a head in a poor country and $5,272 in a rich one, a ratio of 77.5×, and nobody wrote that price down. Bhutan's happiness headcount moves from 43.0 percent to 90.9 percent — 47.9 points — on the sufficiency cutoff alone. Your supplier scorecard is not more robust than these. It is less.

---

PART FOUR — DESTINY AND DELIGHT

Days 71–90: make it survive you

Compilation is separated from the result. The team computing the holdout cannot be the team judged on the target. In a company this usually means internal audit or FP&A, not the operating unit. It is the single structural feature most often dropped on import, and dropping it voids the instrument.

The revision policy is written before the first revision. How a definition changes, who signs, how far back the series is restated. Without it the first inconvenient restatement becomes a negotiation, and a series that can be negotiated is not evidence.

One named person can refuse. The pressure to move the holdout into the bonus plan is constant, reasonable-sounding, and irresistible unless somebody's job is to say no. Name them in the terms of reference.

The failure modes, so you can see them coming.

• The boundary statement becomes boilerplate: copied forward, unread, no longer a constraint. Countermeasure: the pack owner re-approves them annually, by signature.
• The holdout diverges and the response is to question the holdout. This is the most natural response available and it is wrong every time. Countermeasure: write the response protocol before the first divergence.
• Somebody senior proposes a composite index because four series "do not communicate". It communicates by hiding the weights, and a hidden weight will not be argued with. Countermeasure: the parameter register, produced on the spot.
• The satellite view is built once for a board paper and never updated, becoming an artefact rather than an instrument.

---

THE BOARD PAPER TEMPLATE

One page. Six blocks, in this order.

1. What we measure and what it excludes. Three lines per headline number.
2. The target, and its holdout. Both named, both defined, holdout contractually outside incentives.
3. The divergence. Targeted gain, holdout gain, divergence, real share.
4. What the restatement shows. Operating profit with deferred maintenance, technical debt and training debt charged — three years, one chart.
5. The step-up inequality on our metric-linked incentives. Required versus actual, in basis points, per instrument.
6. The decision requested. One. Usually: approve the holdout, approve its exclusion from incentive plans, and name the person who can refuse to move it.

What not to put in it. Any suggestion that anyone has been dishonest. The whole force of Campbell's law is that this happens to conscientious people under a single number, and a paper that implies otherwise will be resisted for reasons that have nothing to do with its arithmetic.

---

THE NINETY DAYS ON ONE PAGE

The only two artifacts that must exist at day ninety are the holdout term sheet and the first divergence reading. Everything else is preparation, and everything else will be produced again next quarter anyway. If the ninety days produce only those two, the quarter was a success.

What it costs. One afternoon of the FP&A lead's time a week, one half-day of the leadership team, and a treasury review that was due anyway. No system, no licence, no consultant, no change to the ledger. The expensive part of measurement has never been the measuring; it has been the arguing that follows a number whose boundary nobody wrote down.

---

DELIGHT, IN A COMMERCIAL REGISTER

There is a specific pleasure available here that executives rarely get, and it is worth saying plainly because it is the reason this survives past the first cycle.

It is the meeting in which the holdout confirms the target. The two series move together, the divergence is close to nothing, and what you have in your hands is no longer a claim but a measurement that survived an attempt to break it. You can take that to a board, a lender, an acquirer or a regulator, and none of them can do to it what they would have done to the unaudited version.

That is worth money, and the amount is not abstract: it is the discount a counterparty applies to a number they cannot verify, and it is usually larger than the cost of verification.

---

APPRECIATIVE QUESTIONS FOR YOUR LEADERSHIP TEAM

1. Which number in our pack do we trust most, and what about its compilation earns that trust?
2. When has a measure in this business helped somebody do something they were proud of — and what made it work?
3. Where are we already measuring something well that nobody outside the function knows about?
4. If every headline number carried what it counts and excludes, what question would this meeting start asking?
5. What is the largest thing we depend on and do not count, and what would a first estimate of it cost us to produce?
6. Which of our targets most deserves a holdout, and what would the honest holdout be?
7. Who in this room would be willing to be the person who refuses to move the holdout into the bonus plan?
8. When our holdout first diverges, how do we want to have agreed to respond?
9. If we restated three years of operating profit with our stocks charged, what would we want to do differently in the next plan?
10. What would have to be true for these instruments to still be running when everyone currently in this room has moved on?
11. Which of our metric-linked incentives is currently under-deterring, and by how many basis points?
12. What is the first sign we would see if this practice were quietly dying, and who would notice it first?