Haute Lumière
Commerce · VI.09 · MMXXVI · daylight
WORKBOOK — THE CORPORATE EXECUTIVE
Chapter VI.09 · Conflict, Repair, and the Long Table
Applied to a P&L, a board paper, a business unit.
THE PREMISE, STATED COMMERCIALLY
You are already running a conflict resolution system. It has one setting, it is called the exit interview, and nobody has ever put it on a cost line.
That is the commercial proposition in this chapter and it does not require you to agree with a single word about commons, cooperatives or restorative justice. It requires you to accept three arithmetic facts, each of which is available from your own systems inside a fortnight:
- Regretted attrition attributable to unresolved dispute is a cost you are already carrying. Replacement cost runs at about 21 per cent of annual salary on Boushey and Glynn's (2012) synthesis of thirty case studies, plus the expected cost of a claim.
- The route you have written down — tribunal, court, arbitration — is slow and expensive at a ratio nobody in your organisation has looked at. Median time from filing to trial in US civil cases runs around 30 months (913 days); employment arbitration averages 361.5 days; a mediation is one day. A properly built internal panel resolves in about 21 days — 43.5× faster than the trial track.
- A standing internal ladder costs, in the chapter's worked house of 200, £34.00 per member per year — £2.83 a month — rising to £43.50 with a binding arbitration backstop.
The board paper writes itself from those three. It is not an ethics paper. It is a substitution of a cheap, fast, measured process for an expensive, slow, unmeasured one that is already running.
PART ONE — DISCOVERY
Days 1–30: find what your own numbers already say
Task 1.1 — The exit classification.
Pull three years of leavers. Classify each as finished, moved, drifted, or unresolved dispute. HR will resist the fourth category; insist on it, and give them a written definition so the classification is reproducible rather than a judgement call.
This produces q — the share of disputes that end in an exit under your current design. It is the single most load-bearing input in the whole analysis and it is the one you can only get from your own records. Do it first.
Task 1.2 — The grievance funnel.
Count formally raised grievances for the same three years, and count their outcomes. Then compute the ratio you will not like: formally raised grievances divided by unresolved exits. If that number is below one, you do not have a low conflict rate. You have a forum people do not use, and every dispute in your organisation is being routed directly to rung six.
Task 1.3 — Find the manager who already does this.
Somewhere in your organisation is a line manager whose team has anomalously low regretted attrition and who handles disagreements early, informally and well. Find them by the number, then go and ask them what they do — in the appreciative form: when has this worked, what made it work, and what would it take to have more of that?
Write down their sequence. That sequence is rung three, already operating in your building, unpaid and undocumented. The design task is not invention. It is making one manager's practice into an institution.
Task 1.4 — Price your current route.
Ask Legal for three numbers: cases taken to external process in three years, total external legal spend on employment and member disputes, and mean elapsed time to disposition. Divide. That is your current cost per dispute resolved externally, and it will be two orders of magnitude above the £330 an internal hearing costs.
PART TWO — THE ARITHMETIC
Days 31–45: the threshold, computed on your own payroll
Task 2.1 — X, the cost of one exclusion.
X = (replacement share × loaded salary) + (P(claim) × cost of defence)The chapter's worked figure: 21% × £30,000 = £6,300, plus 15% × £8,500 = £1,275, giving X = £7,575. Yours will be larger, because your loaded salaries are larger and because your claim probability is knowable from Legal's own history rather than assumed.
Build X three ways and put all three on the page: at your median salary, at your 75th percentile, and at the salary of the roles where you actually lose people to unresolved dispute. That third figure is usually the largest and it is the one the board should see.
Task 2.2 — F and c, the ladder's two costs.
F (standing) panel training + reserved time + convening and the record
c (per dispute) hearing time + administrationWorked house: F = £3,500 a year (£1,500 training, £1,080 reserved time, £920 convening and record) and c = £330 per hearing. Scale for your own charge-out rates. Keep the two separate on the paper — a finance function that sees only a blended figure will ask the wrong question about it.
Task 2.3 — The threshold.
F / N λ
λ* = ------------------ and then ---
(q − s)·X − c λ*Worked house: numerator £17.50, denominator (0.25 × £7,575) − £330 = £1,563.75, so λ* = 0.01119 disputes per member per year — 1.12 per hundred, or 2.24 disputes a year. Observed λ of 0.05 gives 4.5×.
Task 2.4 — Check the scale invariance yourself.
Run λ* at 50, 200, 1,000 and 5,000 head. You will get 0.04476, 0.01119, 0.00224 and 0.00045 — and in every case 2.24 disputes a year. Because F is fixed, the threshold in whole disputes does not move with head count.
Put that sentence in the board paper in bold. It closes off the one objection that would otherwise carry the room: this is fine for a small cooperative and does not scale. It scales in the only direction that matters — downward, into the per-member number, which falls as you grow.
Task 2.5 — Name the boundary of your own case.
The threshold exists only while the ladder diverts more than c / X = 4.4% of disputes from exclusion. State that on the paper, unprompted, with your own figure. A board that finds the limit itself discounts the whole analysis. A board that is handed the limit reads the rest as candid.
Task 2.6 — The distributional honesty.
If you are considering arbitration as your terminal rung, the paper must carry Colvin (2011): 21.4 per cent employee win rate in AAA employment arbitration against 36.4 per cent at federal trial — 1.70× — with a median award of $36,500. Present it, then present the remedy: both-ways invocation, so any member may reach the terminal rung against the organisation. It costs nothing and it is the term that converts a management instrument into a governance one.
PART THREE — DESIGN
Days 46–60: the instrument, in a form a treasurer recognises
The structure: a Repair Reserve with a graduated ladder and an arbitration backstop.
Sizing. Twelve months of expected cost: F + (λ · N · c) + retainer + expected hearings. Worked house: £3,500 + £3,300 + £1,900 = £8,700 a year, or £43.50 per member per year. Fund annually from operating expenditure. It is too small to capitalise and arguing otherwise costs you a meeting you do not need.
Balance sheet treatment. Standing cost is opex. The arbitrator's retainer is a prepayment amortised over its term. Where the membership or employment agreement creates a present obligation to hear a matter on request, unresolved live cases are a provision under your ordinary provisions policy — disclose the count, never the names. One early conversation with the auditor settles all three; this is provisions and contingent liabilities, which they handle every year.
The counterparty. One named external arbitrator or a named panel — a CEDR- accredited mediator, an AAA or JAMS panellist, a sector scheme — on a standing retainer at £1,200 a year plus £3,500 per hearing held. At 2 per cent of ten annual disputes that is 0.20 hearings a year and £1,900 expected — £9.50 per member per year, a 27.9 per cent uplift.
What the uplift buys. Not the hearings. The 98 per cent of disputes that never reach one. Non-compliance with a purely consensual agreement runs at 29.4 per cent (McEwen and Maiman, 1984); ten disputes a year gives 2.9 broken agreements, worth £22,271 if each becomes an exclusion — £111.35 per member per year. The backstop pays for itself if it prevents a quarter of one exclusion a year.
The monitor. Internal audit or the company secretary publishes one page annually: complaints, rung, outcome, median elapsed days, resulting λ. That is Ostrom's fourth principle discharged, and it is also next year's input.
Term. Three years, reviewed against the published count at the end of year one. Review against a number, not against a mood — this is the single term that stops the programme dying with its sponsor.
PART FOUR — DESTINY AND DELIGHT
Days 61–90: deploy, and make it survive you
Task 4.1 — Elect rather than appoint.
Appointed panels produce findings the workforce does not believe, which is worse than no panel because the finding carries organisational weight without carrying legitimacy. Election, a fixed term and a term limit cost nothing and are the difference. Ostrom's formulation is exact and you should quote it in the paper: monitors are accountable to the appropriators, or are the appropriators.
Task 4.2 — Put it on the standing pack.
One line: complaints this quarter, by rung, and median days. Anything reviewed by exception disappears; anything on the monthly pack persists. This is worth more than any presentation you will give about it.
Task 4.3 — Hold the first sitting with nothing on it.
A forum that first appears the day it is needed is a tribunal, and a tribunal is convened about somebody. A forum that has met four times about nothing is furniture. Schedule four empty sittings before the first real matter, and do not let anybody cancel one for being unnecessary. The emptiness is the product.
Task 4.4 — Name the second chair.
One person who can run a hearing is a single point of failure with a name and a notice period. Train three.
THE FAILURE MODES, NAMED
So you can see them coming
Capture. A panel elected by a dominant function becomes that function's instrument and carries legitimacy while doing it. Defences: term limits, rotation, a recusal rule with a named decider for contested recusals, and a composition rule that no two panellists come from the same reporting line. These are partial and you should say so.
Drift to the top rung. Under pressure the organisation stops using rungs that take three weeks and reaches for the one that takes an afternoon. The tell is in your own count: rung-three outcomes falling while rung-six outcomes rise, with total complaints flat. Watch that ratio quarterly. It is the earliest available signal that the system is failing, and it arrives about two quarters before attrition does.
The refuser who is correct. Some refusals to engage are the most important signal the organisation will receive that year — a minority member, a whistleblower, a complaint against the panel's own constituency. A system that reads every refusal as bad faith will crush exactly that one. The defence is both-ways invocation, and it is the reason that clause is not optional.
Nothing behind it. A repair process with no terminal rung is exploitable by anybody patient enough to keep refusing, at a measured rate of 29.4 per cent. Design the last rung so you never use it. Do not design it away.
Legal capture of the cheap process. The moment one party brings representation the other must, and you have rebuilt the expensive mechanism inside the cheap one. Write no-representation into rungs one through four, with a support-person right instead — accompanied, not represented. This is the distinction that keeps a twenty-minute hearing possible.
THE NINETY DAYS ON ONE PAGE
| Day | Action | Artifact |
|---|---|---|
| 1–15 | Classify three years of exits; count the grievance funnel | The exit classification — produces q |
| 16–30 | Price the current external route; find the manager who already does rung three | Current cost per dispute; the documented sequence |
| 31–45 | Compute X three ways, F, c, λ and λ* | The threshold page |
| 46–55 | Draft the six rungs, the movement rule and the no-representation rule | Draft procedure |
| 56–60 | Auditor conversation on provisions; size the reserve | Facility memo |
| 61–70 | Elect the panel; train three chairs | Panel roster |
| 71–80 | Name the arbitrator; sign the retainer with both-ways invocation | Signed retainer |
| 81–90 | Publish the day, the rungs and the count template; hold the first empty sitting | The first standing sitting |
BOARD PAPER TEMPLATE
One page. Six blocks. The number on the front.
1. The decision. Approve a Repair Reserve of £[X] a year — £[Y] per head — establishing a six-rung graduated procedure with a named external arbitration backstop invocable by either party.
2. What we spend now. [N] unresolved-dispute exits over three years at £[X] each = £[total]. Plus £[legal spend] external. Current mean time to disposition: [D] days.
3. The threshold. λ* = (F/N) ÷ [(q − s)·X − c] = [figure] disputes per member per year. Observed λ = [figure]. Ratio: [n]×. The threshold in whole disputes is [figure] a year and does not change with head count.
4. What it costs. Standing £[F]. Per hearing £[c]. Backstop £[retainer] + £[per hearing] × [expected]. Total £[Y] per head per year.
5. Where this fails. The ladder must divert more than [c/X]% of disputes from exclusion or the arithmetic reverses. q is estimated from [source] and is the weakest input; at q below [figure] the case does not hold.
6. What we will publish. One page annually: complaints, rung, outcome, median days, resulting λ. Reviewed at twelve months against that number.
The number that decides it is λ ÷ λ\*. If it exceeds 1, the ladder is already cheaper than what you are doing, and the paper is a cost-substitution paper. Present it as one.
APPRECIATIVE QUESTIONS FOR YOUR LEADERSHIP TEAM
- Think of a serious disagreement in this business in the last two years that ended well. Who handled it, how long did it take, and what did they do?
- Which manager here loses the fewest people to things that could have been heard — and what is their actual sequence?
- If we published one page a year about how we handle disagreement, what would we want on it?
- What are the real rungs between a quiet word and a dismissal, and which of them do we already use without having named them?
- What would have to be true for someone to raise a complaint against one of us and be confident of a fair hearing?
- What number would tell us this had quietly collapsed back to its top rung, and who would see it first?