Haute Lumière

Commerce · VI.03 · MMXXVI · daylight

La Bourse  /  Volume VI  /  Nº VI.03  /  Workbook — the executive

WORKBOOK — THE CORPORATE EXECUTIVE

Chapter VI.03 · Holarchy Without Hierarchy

Ninety days, one business unit, one number your board has never been shown.

---

THE PREMISE, STATED COMMERCIALLY

Your organisation carries a coordination cost and does not report it. It is distributed across a managerial payroll line, an unmeasured quantity of calendar time, and a latency that shows up in the market as slowness rather than in the accounts as anything at all.

This workbook makes it a number.

The claim is narrow and it is defensible. A nested structure and a hierarchy are the same tree. Ten thousand people in teams of twelve at a span of five is 834 teams and 6 levels deep either way. What differs is one rule: whether a decision rises until it meets authority, or rises only when it cannot be contained. On the modelled parameters that single substitution is worth £85.35m a year on a 10,000-person base — £8,535 per front-line person — and the break-even containment rate a nested structure has to clear is 37.8%.

Two things follow that you should say out loud in the first meeting. This is not a culture programme. It is a coordination-cost argument with a measurable threshold, and it can be run as a pilot in one business unit inside a quarter. And it is not flatness. You are not removing a layer. You are changing what each layer is for, which is a considerably smaller organisational event and a considerably larger financial one.

---

PART ONE — DISCOVERY

Days 1–30: find what your structure already contains

Exercise 1.1 — Count your tree (half a day)

You need four numbers and your HR system has all of them.

1. Front-line headcount in the unit.
2. Average team size, h.
3. Average span of control above team level, s.
4. Coordination posts: every role whose primary output is a decision about somebody else's work.

Then compute depth: ceil(log_s(teams)) + 1.

Most executives are surprised twice. First by the depth — six levels is ordinary and nobody had counted. Second by the coordination-post ratio. A span-5 tree run to teams of twelve carries 1,043 posts per 10,000 front-line people, 10.43%. Run the same span all the way down to individuals and it is 2,500, 25.0% — which is close to what the Dutch home-care sector was actually carrying when Buurtzorg entered it with 1.00%.

Exercise 1.2 — Find the unit that already contains its own work (one day)

Somewhere in your organisation is a team that decides almost everything itself, performs well, and has never been written up. It is usually remote, usually inconvenient to supervise, and usually described as "a law unto themselves."

Find it. Interview it appreciatively — when has this worked, what made it work, what would it take to have more of that? You are looking for what it was given and by whom, and whether the giving was deliberate.

This is your pilot unit. You are not installing a structure; you are documenting one that already exists and works, which turns a proposal into a measurement.

Exercise 1.3 — The escalation sample (one day)

Pull fifty decisions from the last quarter that crossed a level. For each, classify the reason in one word:

• Span — it genuinely touched a domain the unit does not hold.
• Threshold — it exceeded an authority limit, and the unit knew the answer.
• Comfort — nobody wanted to own it.

The Threshold and Comfort counts are your addressable coordination cost. They are also, in most organisations, the majority.

---

PART TWO — THE ARITHMETIC

Days 31–45: build the business case

Exercise 2.1 — Your escalation ledger (half a day)

        L(c, d)  =  ( 1 − (1 − c)^d ) / c

At depth 6, an authority-threshold rule at c = 0.55 gives L = 1.8031 levels — 0.8031 escalations per decision. A containment rule at c = 0.85 gives 1.1765, or 0.1765. The difference is 0.6266 levels.

Price a level with your own figures: people in the room × hours × loaded rate. The chapter uses four people for ninety minutes at £45 — £270. At 240,000 coordination-requiring decisions a year that is £52.04m against £11.43m.

Exercise 2.2 — The full ledger, your unit (one day)

                              HIERARCHY        NESTED
  coordination posts              1,043            50
  posts, cost                   £73.01m        £3.50m
  inside-team coordination       £4.95m       £29.72m
  escalation                    £52.04m       £11.43m
  ------------------------------------------------------
  total                        £130.00m       £44.66m

Rebuild this table with your own headcount, your own loaded costs and your own depth. Do not omit the mesh line. A team of twelve coordinated as a mesh carries 66 ties against 11 for a hub — 6.0× — and it is £29.72m against £4.95m in the model. Omitting it is the single most common error in self-management business cases, and your CFO will find it.

Exercise 2.3 — Your break-even containment (2 hours)

  L(c*, d)  =  ( hierarchy cost − nested fixed cost )
               / ( decisions × cost per escalation )  +  1

At the modelled parameters, c* = 37.8% against a modelled 85% — a margin of 47.2 points. At 3,000 people in teams of 10 at span 6 it is 28.2%. Compute yours. It is one line of arithmetic and it is the number your board paper is about.

Exercise 2.4 — The honest negative, priced (half a day)

Three, and each has a figure.

The benefit saturates. The gap between the two rules can never exceed 0.6417 levels per decision. At about 1,500 people you have 88.5% of it; at 37,500 you have 97.6%. This is not a scale play. If you are presenting it as one, the arithmetic will not support you at the second meeting.

The mesh is quadratic. Per-team crossover h* = 12.47 with a band of 10.46 to 14.16; the organisation-wide crossover at equal containment is 18.82. Teams that drift past eighteen cost more than the managed version and the cost is invisible because it sits in calendars.

The emergency. Containment goes to zero, L(0, 6) = 6, and that is 30 working days by consent against 5 under unity of command. Put the twenty-five days in the risk section with the mitigation beside it, and the mitigation is one page.

Exercise 2.5 — Where the evidence stops (2 hours)

Your board will ask for proof. Know exactly where it runs out, because saying so first is what makes the rest credible.

Buurtzorg is the strongest case: roughly 40% fewer hours of care per client on KPMG's 2015 evaluation, overhead near 8% against a sector 25% — 17 points — and a national estimate of EUR 2.0 bn a year if generalised. It is not a randomised trial; client mix, nurse qualification and self-selection all differ and none is controlled for.

Morning Star runs about 400 colleagues on some $700m — $1.75m each — with no managers, and the CLOU is documented in detail. It is one private firm in one stable commodity process, owned by one person.

Haier is testimony. No microenterprise-level accounts are published; the management literature rests on company-supplied data over a period containing four large acquisitions, including GE Appliances at about $5.6bn. The one computable figure is the headcount: about 10,000 middle-management roles removed.

Zappos is the honest counterweight and you should raise it yourself. 210 of roughly 1,500 people — 14.0% — took the 2015 severance offer; turnover that year was about 30% against a historical 20%, an excess of around 150 people; the Holacracy label was retired in 2020.

---

PART THREE — DESIGN

Days 46–60: the instrument

Exercise 3.1 — The Nested Authority Schedule (two days)

For the pilot unit, one page:

Exercise 3.2 — The microenterprise contract (two days)

Transfer-pricing basis, notional P&L, and gain participation above an agreed baseline. This is Haier's value-adjustment mechanism and it is structurally the shared-savings facility of Chapter I.01: the unit's upside is a share of a verified improvement, and the baseline is signed before anything changes.

Exercise 3.3 — The balance-sheet conversation (one meeting)

While holons are management-accounting constructs this is segment reporting and internal transfer pricing; the only documented item your auditors will want is the transfer-pricing basis. The moment a holon becomes a separate legal entity and takes outside capital, the control tests under IFRS 10 become live and the answer is not obvious. Walk into that deliberately, with advice, or not at all.

---

PART FOUR — DESTINY AND DELIGHT

Days 61–90: deploy, measure, and report once

Exercise 4.1 — Deploy and log (thirty days)

Publish the Authority Schedule. Log every decision in the unit: closed, or escalated with a reason classified Span, Threshold or Comfort. Nothing else changes in month one.

Exercise 4.2 — Rehearse the exception (one afternoon)

Declare a scenario nobody believes, hand command to the named commander, run it for two hours, revert on the stated clock without anyone requesting it back. This is the only part of the programme that looks like theatre and it is the part that makes the clause credible when it is real.

Exercise 4.3 — The three drifts, watched monthly (20 minutes)

The coach re-acquires authority — counter: no budget, evaluated on the containment rate of units served. Teams grow past the cap — counter: a written split rule and a standing report of team sizes. The exception is never rescinded — counter: reversion by date, extension by positive act.

Exercise 4.4 — The one page (one hour)

One page, one person, one comparison.

Measured containment c — __._% · Break-even c* — __._% Coordination posts at current shape — ____ · at nested shape — ____ Annual coordination cost, both ways — £___m and £___m The one risk, priced: __ working days of emergency latency, mitigated by the exception clause attached.

If c clears c*, say the sentence plainly: this is the cheaper structure, and the difference is £____ per front-line person per year.

---

WHAT TO SAY WHEN THE ROOM PUSHES BACK

Four objections you will get, and the answer to each

"This is the flat-organisation fad, and we have seen it fail." Agree, and name the failures first — Zappos and Medium, with the figures. Then draw the distinction that matters: both of those removed structure. Nesting keeps the tree at the same depth and changes only the escalation rule. The organisations that kept their trees — Buurtzorg with its coaching cadre, Morning Star with its peer contracts, Haier with its four thousand contracting units — are the ones still running the model.

"Our work is too interdependent." This may be true, and if it is, the instrument will say so. Simon's near-decomposability is the load-bearing assumption: where almost every decision genuinely touches almost every unit, c sits below c* and a hierarchy with a wide span is the right answer. Offer to measure it rather than to argue it. A four-week log settles a two-year argument.

"Our managers will not accept it." Notice what the arithmetic actually says about the alternative. Widening the span from five to seven takes the hierarchy's own cost from £130.00m to £123.91m — widening is the hierarchy's best move, and a wide span is only safe when units are self-contained. Your managers' own best case and this proposal are converging on the same requirement from opposite directions, and that is a far easier conversation than a loyalty test.

"What happens in a crisis?" The strongest question, and you should have answered it before it is asked. Containment goes to zero, L(0, 6) = 6, and that is 30 working days by consent against 5 under unity of command. The answer is the exception clause, priced, attached and rehearsed — and it is a better answer than most hierarchies have, because most hierarchies have never written down who commands either.

---

THE FAILURE MODES, NAMED

So you can see them coming

You present it as flatness. The room hears "remove managers," the managers hear "remove me," and the argument becomes political within one meeting. The correct framing is that the tree is unchanged — 6 levels, same span — and what changes is what each level is for.

You skip the exclusion list. A domain statement with only "decide alone" and "consult" is incomplete, and people fill the gap by asking whoever has standing. That is Freeman's mechanism arriving by the back door: a team of twelve has 66 possible ties, about 30 are actually sustained — 45.5% — and the other 54.5% of the information routes through the people who already had the relationships.

You let teams grow. Past 18.82 the mesh costs more than the managed version and nobody can see it.

You never rehearse the exception. Then the first real emergency is the rehearsal, and the clause is invoked by someone improvising, and it is not rescinded.

You measure engagement instead of containment. Engagement will rise, it will be attacked as a soft number, and you will lose a true argument with the wrong evidence. Containment is countable from a log and survives a CFO.

---

THE NINETY DAYS ON ONE PAGE

---

BOARD PAPER TEMPLATE

Recommendation. One sentence, with the structure named and the unit named.

The number. c against c*, and the annual coordination cost both ways.

What is being changed. The escalation rule, not the number of levels. State the depth before and after and note that it is the same.

Evidence. Buurtzorg with its limits stated. Morning Star with its limits stated. Haier as testimony, labelled. Zappos raised by you, first.

Risk, priced. Emergency latency, 30 working days against 5, mitigated by the attached clause. Mesh cost at team sizes above 18.82, mitigated by the split rule.

The decision requested. One pilot unit, one quarter, one signature.

---

APPRECIATIVE QUESTIONS FOR YOUR LEADERSHIP TEAM

1. Which unit here already decides almost everything itself and performs well? What was it given, and by whom?
2. What is the last decision that came to this table that should have closed two levels down — and what would have had to be true for it to?
3. If every unit knew precisely what it may not decide, what would change about how this team spends a Thursday?
4. Where do two of our domains touch with nobody owning the seam?
5. When has someone here taken command in a real emergency and handed it back cleanly? What made the handing-back possible?
6. What would we need to see in ninety days to make this permanent — and is it a number we could already be counting?