Haute Lumière

Commerce · IV.10 · MMXXVI · daylight

La Bourse  /  Volume IV  /  Nº IV.10  /  Workbook — the Gainshare employee

WORKBOOK — THE LUMINOUS GAINSHARE EMPLOYEE

Chapter IV.10 · Measuring Regeneration

For the person inside a gainshare, where what you are paid depends on what can be proven. This chapter is about the cost of proof — which means it is about the cost of your own claim, and about the one asset in the whole system that you personally can create.

---

WHAT A GAINSHARE ACTUALLY IS, IN THIS CHAPTER'S TERMS

A gainshare pays you a share of a measured improvement. Every word in that sentence is load-bearing, and this chapter is about the third one.

If the improvement cannot be measured at a cost less than the improvement is worth, there is no gain to share — not because the improvement is not real, but because nobody will pay for something whose proof costs more than the thing. That ratio has a name in this chapter:

  κ  =  cost of proving the gain  ÷  value of the gain

Your entire economic position inside a gainshare is a function of κ on the things you do. A κ of 163.826 means your work is genuinely valuable and entirely unclaimable. A κ of 0.079 means the same work is a line on a ledger with your name beside it.

So the highest-leverage thing you can do in a gainshare is not to work harder on the gain. It is to lower κ on the gain you are already producing. That is what this workbook is for, and it is almost never what people are told.

---

PART ONE — DISCOVERY

Days 1–30: find the gain you are already producing unpriced

Exercise 1.1 — List what you improve that nobody measures

Two columns, twenty minutes, no editing. Left: things that are measurably better because of you. Right: things that are better because of you and are not measured.

The right-hand column is longer. It always is. Every entry on it is a gain currently sitting at κ = infinity, because the numerator exists and the denominator has never been computed.

Exercise 1.2 — For the top three, ask the four questions

1. What is the effect, and what is the scatter it sits in?
2. How many observations, and how far apart in time?
3. What would proving it cost, and what is it worth?
4. If κ is above one, what structure changes it?

You are not being asked to do the science. You are being asked to write the sentence "this would take N observations over T months at roughly $C, against a gain of $V." Almost nobody in your organisation can write that sentence about anything, and the person who can is the person whose claims get paid.

Exercise 1.3 — Find the relocation assets nobody is using

Where does your organisation already return to exactly the same point to take a reading? A fixed sensor, a monthly report run on identical parameters, a recurring survey with a stable sample, a calibration standard.

Every one of those is a paired-measurement asset sitting idle. Attaching your gain to an existing repeated measurement costs you nothing and takes ρ from roughly 0.50 to roughly 0.90 — which, in the chapter's soil case, takes the observations needed from 224 to 45.

Output: your unmeasured gains list, three with the four questions answered, and every existing repeated measurement in your reach.

---

PART TWO — THE ARITHMETIC

Days 31–45: compute what you are owed

Exercise 2.1 — Learn the two formulas that decide everything

  how many:   n  =  k σ_d² / Δ²      with  σ_d² = 2σ²(1 − ρ)
  how often:  T  =  MDD ÷ rate       with  MDD = (z₀.₉₇₅ + z₀.₈₀)·√(σ_d²/n)

In the chapter's case: σ = 8.00, Δ = 1.50, ρ = 0.90 gives 45 observations; 45 observations gives MDD 1.4942 and an interval of 4.98 years.

Apply both to your own gain. If the interval your arithmetic produces is longer than your gainshare period, you have found something worth raising today, because the scheme is promising to pay on a measurement it cannot make in time.

Exercise 2.2 — Build your own cost stack

Three lines, in your own currency:

• per observation — what does one reading actually cost in labour and analysis? (The chapter's soil figure is $55.00.)
• per event — what does mobilising to take a batch cost? ($2,000.00.)
• per claim — design, verification, admin, regardless of size. ($17,000.00.)

Then: total = per claim + events × (per event + n × per observation). The chapter's paired claim is $25,950.00 and the unpaired equivalent is $70,170.00.

Exercise 2.3 — Compute your κ, honestly, including the bad answer

Divide. If κ is above one, the gain you are producing cannot be claimed by you alone, and saying so plainly is worth more to you than an optimistic number that collapses at the first review.

Then compute what would change it. The chapter's three levers are yours too:

• aggregation — pool your claim with colleagues' so the fixed cost is shared. In the chapter's case this takes κ from 163.826 on a lone two-hectare holding to 0.366 across 5,000 pooled hectares.
• tiered assurance — a model calibrated on a sampled fraction. Sampling one in five takes marginal cost from $24.75 to $4.95 per hectare and κ at 5,000 hectares to 0.131.
• outcome proxies — a cheap indicator, priced honestly with a wider deduction. κ 0.058, with saleable value falling from 3.960 to 3.080 t CO₂e/ha.

Exercise 2.4 — Know the floor, because it protects you

  κ_floor = c_ha ÷ V_ha = 24.75 ÷ 79.20 = 0.312   at $20/t

Aggregation spreads fixed cost and does not touch the floor. If somebody tells you the scheme will get to a 10 percent verification cost purely by growing, they have either not computed the marginal term or are using a proxy they have not disclosed. Either way you are entitled to ask which.

Output: your gain, its design, its cost stack, its κ, and the one structure that moves it.

---

PART THREE — DESIGN

Days 46–70: make the uncounted countable

Exercise 3.1 — Install your stake

This is the chapter's central move and it translates exactly. The single cheapest thing you can do to raise the value of your claim is to fix the point of observation before you start — the same report, the same parameters, the same population, the same cohort, the same date in the month, written down where somebody else can find it.

In the chapter's arithmetic, $135.00 of markers avoids $44,220.00 of measurement — a return of 327.6 times. Your equivalent costs an afternoon and is worth the same multiple, for the same reason: it buys ρ.

Do it before the gain starts accruing. A baseline established afterwards is not a baseline; it is a future dispute, and in a dispute about a gainshare you are not the party with the records.

Exercise 3.2 — Write your claim the way a verifier reads it

Six lines, no more:

1. The claim. One sentence, with the effect size and its units.
2. The baseline. Date, method, who holds the record.
3. The design. n, paired or not, and the interval — with the arithmetic.
4. The cost. Your stack, totalled.
5. κ. The number, and the band it is in.
6. The deduction you accept. State it yourself. A claimant who proposes their own uncertainty deduction is believed; one who waits to have it imposed is discounted harder.

Exercise 3.3 — Find your pool

If κ on your own claim is above 0.25, find the colleagues whose gains share your fixed cost — same method, same verifier, same measurement window — and propose one claim with an allocation rule.

The chapter's threshold in its own units: H\ = 1,303.1 hectares at $20 a tonne, which is 652 holdings of two hectares. The translation for you is: how many colleagues' claims share my fixed cost, and is that enough?*

---

PART FOUR — DESTINY AND DELIGHT

Days 71–90: make it hold

Exercise 4.1 — Ask for the four things, by name

These are reasonable, cheap, and they protect the scheme as much as they protect you.

• The verification method, in writing, before the period starts.
• The deduction rule, and whether it differentiates between directly measured and modelled claims. Uniform deductions penalise the best performers — the chapter's own failure mode, and the one that unravels a pool from the top.
• The calibration fraction as a covenanted floor, not a budget line. It is the thing most likely to be trimmed in year four, and trimming it changes no published number while destroying the instrument.
• The null-result policy. What gets published if the reading comes back inside the noise band. Agree it while nothing is at stake.

Exercise 4.2 — Keep your own register

Whatever else is true of a gainshare, the party holding the records holds the argument. This is not adversarial and it is not distrust; it is the same reason a surveyor keeps field notes and a laboratory keeps raw traces. A scheme that is being run well will be glad you have them, because your register is a second route to the same answer and a second route is how errors get found before they get published.

Keep baselines, dates, methods, raw readings, and who agreed what and when. It costs almost nothing and it is the only asset in a gainshare that is genuinely yours.

Exercise 4.3 — The delight

Come back to your fixed point at the end of the period and read it. The interval either straddles zero or it does not. Either way you built something that could have proven you wrong — which is the rarest and most valuable position anybody in a gainshare can occupy, because it is the only one that gets believed twice.

---

KNOW YOUR SCHEME — A CHECKLIST

Any unticked line is a question, not a grievance. Ask it in the same breath as the answer you would accept.

---

THE CONVERSATION, SCRIPTED

"I've costed the verification on my claim. At the current design it's about [$C] against a gain of [$V], so κ is [n]. That's above the band where it's worth anybody's money to prove.

Three things move it, and I've checked all three. Pooling my claim with [colleagues] shares the fixed cost and takes it to about [n₂]. Running it on a calibrated model with a sampled fraction takes it to about [n₃]. A proxy takes it lower still, at a wider deduction I'd accept.

I'd like the second one, I've written the design, and I need the baseline agreed before the period starts. Here's the page."

No grievance, no ask for trust, one number and three routes. That is the whole technique and it works because it hands the other person a decision rather than a problem.

---

WHERE THIS TRANSLATES OUT OF THE FIELD

Nothing in this chapter is really about soil. The arithmetic is the same wherever a gain is slow, a baseline is noisy and somebody has to pay for proof.

A retention gain. Voluntary turnover is noisy at team scale and the effect you are claiming is small against it. Same problem, same fix: fix the cohort definition in advance, use the same report parameters every quarter, and pool across teams so one analysis serves several claims.

A quality or yield gain. Here ρ is usually already high, because the production line is the fixed point — the measurement infrastructure was built for another reason and is available to you free. This is the cheapest kind of gain to claim and it is systematically under-claimed for exactly that reason: nobody thinks to look at what is already being logged.

A customer-outcome gain. The scatter is enormous and the effect is real, which is the worst combination and the one where κ is highest. Almost always this needs pooling — across cohorts, across sites, across quarters — before it can be claimed at all.

A cost-avoidance gain. The hardest, because the counterfactual is unobservable by definition and there is no fixed point to return to. The honest move is usually to claim the practice rather than the outcome, at a price reflecting the uncertainty, and to say so. That is not a lesser claim. It is a correctly labelled one, and correctly labelled claims are the ones that survive the second review.

The general rule, worth carrying out of this chapter entirely: the value of a gain to you is not what it is worth. It is what it is worth minus what it costs to prove, and the second term is the one almost nobody works on. Work on it. It is uncrowded ground and the returns are extraordinary — 327.6 times, in the one case this chapter computes all the way through.

---

APPRECIATIVE QUESTIONS FOR YOUR TEAM

1. What do we already measure repeatedly and identically, and who set that up?
2. Which of our gains is real, obvious to everybody, and has never been costed — and what would it take to cost it once?
3. If we pooled our claims into one verification, what would become provable that none of us can prove alone?
4. What would we want to be able to demonstrate in three years, and what is the one cheap thing we would have to fix in place this month for that to be possible?