La Bourse / Volume II / Nº II.09 / Workbook — the executive

A watercolour of an open book with a garden growing out of its pages: succulents, a flowering stem and new leaves. — Plate II.09 · Workbook — the executiveThe Nursery Bed.Every row in this plot is a guess. The plot is not the guess. The plot is the apparatus that lets a guess be wrong cheaply, and lets a right one be carried into next year by somebody who was not here.

WORKBOOK — THE CORPORATE EXECUTIVE

Chapter II.09 · Evolution and Economic Selection

For the person with a P&L, a board calendar and a scorecard they did not write. This workbook is about finding out what your organisation is actually selecting for, computing the rate, and changing the criterion on purpose — using the firm's own numbers, in the firm's own language.

THE PREMISE, STATED COMMERCIALLY

Your organisation has a population of units — teams, sites, product lines, subsidiaries, franchisees, suppliers. They differ. Some carry more weight into next year than others. What they do propagates, imperfectly, by transfer, promotion, imitation and acquisition.

That is a selection system whether or not anyone designed it, and it has a measurable rate. The rate is Cov(w,z)/w̄ — the covariance between a trait and who carries weight forward, divided by mean fitness — and it is computable from data your finance function already holds.

The commercial claim of this chapter is narrow and strong: the criterion in your management accounts is doing more to determine the firm's future capability mix than your strategy, and it is doing it at a rate you can put in a paper. In the chapter's worked population, switching the weighting from next period's revenue to this period's margin turns +0.6545 points a generation into −0.5688 — a swing of 1.2234 a generation, and 12.23 percentage points over thirty years, with no change in behaviour by anyone.

PART ONE — DISCOVERY

Days 1–30: find out what you are actually selecting for

Exercise 1.1 — The fitness function, as found (one week)

Do not ask what the strategy is. Establish, from documents, what determines which units carry weight into next year. Specifically:

What decides next year's budget allocation by unit? Name the metric.
What decides who gets promoted into the unit-leader population?
What decides which units are consolidated, sold or closed?
Over what period is each measured, and at what boundary?

Write the answers on one page and title it The fitness function, as found. Nobody in your organisation has this page and several people will want it.

Exercise 1.2 — The trait census (one week)

Choose one capability trait z that (a) you believe matters over five years, (b) is already recorded somewhere, and (c) nobody currently targets. Candidates that are usually already in the system: training hours per head, maintenance spend against plan, proportion of revenue from repeat customers, documented procedure coverage, internal transfers in and out.

Pull z for every unit for the last three years. Do not adjust anything yet.

Exercise 1.3 — What is already working (2 hours, out loud)

Find the unit with the highest z that is also performing well commercially, and find out how it got there. Who started it, what did it cost, and who has copied it? You are looking for an existing inheritance path, and one usually exists. It is far cheaper to widen a path than to build one.

Exercise 1.4 — The variance audit (half a day)

Compute V_between — the variance of z across units — for each of the last three years. If it is falling, your organisation is homogenising, and from the threshold b/c > 1 + V_within/V_between it is becoming progressively less able to improve by selection at all. A falling between-unit variance is a leading indicator that nobody currently reports. Put it in the pack.

PART TWO — THE ARITHMETIC

Days 31–45: compute the rate

Exercise 2.1 — The selection term on your own population (one day)

For last year: define w as weight carried forward per unit of weight held — revenue, headcount or capital employed, chosen once and defended in writing. Then, exactly as the chapter does:

  z̄ = mean of z              w̄ = mean of w              E(w·z) = mean of w·z
  Cov(w,z) = E(w·z) − w̄·z̄
  selection term = Cov(w,z) / w̄        units: trait-points per generation

Worked, in the chapter's population: E(w·z) = 7.320, w̄·z̄ = 6.600, Cov(w,z) = 0.720, term = 0.720 / 1.100 = 0.6545 points a generation.

Exercise 2.2 — The counterfactual criterion (half a day)

Recompute with w defined by the alternative measure your board might plausibly adopt. Present both terms side by side. This single table is the most persuasive artifact in the workbook because it holds behaviour constant and varies only the accounting, which is the one thing a board can change at will.

Exercise 2.3 — The response, and whether the mandate is achievable (2 hours)

Apply R = h²·S.

S is your selection differential: the mean z of the units you actually grew, minus the population mean. Compute it from last year. The chapter uses 4.00 points.
h² is transmission fidelity. You must estimate this from your own data, not assume it — track three practices that moved between units last year and measure how much of each survived the move. The chapter's 0.35 is an assumption and is flagged as one.

  R = 0.35 × 4.00 = 1.40 points per generation
  generations in a 10-year horizon at 3 years each = 3.33
  delivered = 4.67 points      required = 10.00 points
  shortfall = 5.33 points — the programme delivers 46.7% of its mandate

If your own numbers come out like this, you have found something worth a board paper: the transformation target cannot be reached by selection at the current intensity, and the three levers are S = 8.57 (select harder), h² = 0.750 (copy better) or a generation of 1.40 years (cycle faster).

Exercise 2.4 — The null band (1 hour)

Before you claim any historic trend in the pack, compute R·√n. At R = 1.40 over ten generations that is 4.43 points, two sigma 8.85. Any historic move smaller than that is not evidence of anything, and saying so once in a board paper will buy you more credibility than any other sentence in it.

PART THREE — DESIGN

Days 46–60: change the criterion, not the people

Exercise 3.1 — The boundary test (half a day)

Identify one behaviour that is costly to a unit and valuable to the group — lending capacity, sharing a customer, releasing a good person, publishing a method. Estimate b (benefit per point to the group) and c (cost per point to the unit) in money, roughly. Then test:

   b / c  >  1 + V_within / V_between

The chapter's case: 1 + 2.6667/9.0000 = 1.2963 against an actual b/c of 2.0000, critical b* = 0.03889, headroom 1.54 times. Below the threshold the trait is selected out — at b = 0.020 the term is −0.1789 a generation — however loudly it is encouraged in a town hall.

The management consequence is exact. If b/c is below the threshold, the only fixes are to raise the group term in the scorecard, to lower the private cost, or to raise V_between by letting units differ more. Exhortation is not on the list.

Exercise 3.2 — Write the criterion (one day)

One page, versioned, dated. It states: the trait, the boundary it is measured at, the period, the verifier, the amendment window, and the notice period for a change. Signed by you and by finance. This document is the intervention. Everything else in the programme administers it.

The notice period is not a courtesy. A criterion that can change without notice cannot be invested against, and units will rationally decline to invest in anything with a payback longer than the notice.

Exercise 3.3 — The cohort, sized (half a day)

Variants, not a pilot. A single pilot produces no variance and therefore no response — it produces an anecdote. Size it as the chapter's facility does: twelve variants at £85,000 each, £1,020,000 a cohort, three kept, an intensity of 0.250 published before results. At a response of 1.40 points a generation and a value of £310,000 per point per year, that is £434,000 a year — 42.5 percent against a 9 percent WACC, clearing by 33.5 points, payback 2.35 years, £728,571 per point of response.

Run the same table on your own numbers. If it does not clear, the honest answer is a smaller cohort or a better transmission mechanism, not a rounder estimate.

PART FOUR — DESTINY AND DELIGHT

Days 61–90: make it hold

Exercise 4.1 — The three defences (one week)

Write one paragraph on each, naming an owner:

Versioning. Where the criterion lives, who amends it, on what clock.
Variance floor. The minimum proportion of capacity held in units that are deliberately not standard, defended in the budget the way a reserve is.
Deposit. Where a routine is written down so it survives the unit. Name the repository and the condition of funding that requires it.

Exercise 4.2 — Goodhart, pre-empted (2 hours)

Every published measure becomes a target. Hold a second, unpublished measure of the same underlying thing, sample it quarterly, and amend the published one when the two diverge. Write down now what the second measure is and who holds it. Doing this before the first divergence costs an afternoon; doing it after costs the criterion's credibility.

Exercise 4.3 — The sealed prediction (1 hour)

Publish, before the period: predicted Δz̄ and an interval. The chapter's example is 0.6545 ± 0.25, so the claim dies outside [0.4045, 0.9045] — 0.71 survives, 0.40 and 1.10 falsify it.

A predicted rate with an interval, stated in advance, is the single thing that separates this from every transformation programme your board has previously approved. It is also the reason they will approve the second cohort.

THE FAILURE MODES, NAMED

So you can see them coming

The strategy loses to the scorecard, silently. This is the default outcome and it produces no incident, no escalation and no red line anywhere. The capability trait declines at −0.5688 a generation while the strategy document sits unamended in the board pack. The only early warning is the selection term itself, computed annually, which is why Exercise 2.1 belongs in the standing pack rather than in a project.

The cohort becomes a pilot. Twelve variants are approved and nine are cancelled at the first review, on the grounds that they are not working — which was the plan. Without the variance there is no response, and the surviving three then carry a claim they cannot support. The defence is the published kept-to-funded ratio: three of twelve, an intensity of 0.250, stated before the results and cited at the review.

The criterion is amended mid-period. A new chief executive, a new segment structure, a new reporting line, and the population is scored against a rule that did not exist when the decisions were taken. Every subsequent investment with a payback longer than the notice period is rationally declined. The defence is the amendment window, written into the criterion on day one.

Between-unit variance is tidied away. Consolidation, standardisation and best-practice rollout each look like good management in isolation, and together they drive V_between toward zero. From b/c > 1 + V_within/V_between, the required ratio then rises without bound and group-level improvement becomes arithmetically impossible. No single decision looks wrong. The defence is a variance floor with a named owner and a budget line.

Transmission is assumed rather than measured. Every figure downstream of h² is linear in it. Halve 0.35 to 0.175 and the response halves to 0.70 a generation, delivering 2.33 points instead of 4.67. Spend the month measuring it before the facility is approved, and put the measured number, with its method, in the paper.

THE NINETY DAYS ON ONE PAGE

Day	Action	Artifact
1–10	Establish what actually determines next year's allocation	The fitness function, as found
11–20	Pull the trait for every unit, three years	The trait census
21–30	Compute `V_between` by year	The variance audit
31–38	Compute `Cov(w,z)/w̄` for last year	The selection term, with units
39–45	Recompute under the alternative criterion	The two-number table
46–52	Measure `h²` from three practices that moved	A transmission estimate with its method
53–60	Draft and sign the criterion with its amendment window	The signed criterion
61–75	Size the cohort; take it through the decision inequality	The facility memo
76–85	Name the three defences and their owners	The versioning, floor and deposit note
86–90	Publish the predicted `Δz̄` and its interval	The prediction, dated

THE BOARD PAPER — one page

Title. What we are selecting for, and at what rate.

The fitness function, as found. What actually determines which units carry weight forward. One paragraph, sourced from documents.
The measured rate. Cov(w,z)/w̄ for the trait, with units, for the last three years. One table.
The counterfactual. The same population under the alternative criterion. Two numbers, side by side.
The null band. R·√n, so the board knows which historic moves are real.
The ask. The criterion, one page, with its amendment window; and the cohort, sized, with its intensity published.
The number that decides it. h²·S·(horizon/generation)·value per point > cohort cost × WACC — £1,446,667 against £91,800 in the worked case.
The prediction. Δz̄ and its interval, dated, for the coming period.

APPRECIATIVE QUESTIONS FOR YOUR LEADERSHIP TEAM

Which of our units is most unlike the others, and what have we learned from it that uniformity would never have shown us?
When has a practice travelled between our units without anyone mandating it? Who carried it, and what made it easy to carry?
What number would we be willing to publish before the period and be scored against afterwards?
Where does one of our teams win its own targets in a way that costs the group more than it gains — and what would the group term have to be worth for that to reverse?