Haute Lumière

Commerce · II.07 · MMXXVI · daylight

La Bourse  /  Volume II  /  Nº II.07  /  Workbook — the Gainshare employee

WORKBOOK — THE LUMINOUS GAINSHARE EMPLOYEE

Chapter II.07 · Information, Signal, and Price

For the person working inside a gainshare arrangement — where a defined share of verified improvement returns to the people who created it. This chapter is more directly about your position than any other in the volume, because a gainshare is literally a second channel, built beside the wage, to carry a dimension the wage cannot.

---

WHY THIS CHAPTER IS YOURS

A wage is a price. A price is a scalar. Everything you bring to a week of work — skill, judgement, care, the thing you noticed and fixed before anyone saw it, the person you steadied — arrives at the firm as one number a month, and that number carries about as many bits as a shelf price.

A gainshare exists because somebody accepted the kernel argument. It is a deliberate second channel: a measured attribute, a baseline, a share and a verifier, built precisely because the wage could not carry what it needed to carry. You are not a beneficiary of this chapter's theory. You are standing inside its instrument.

Which means two things follow, and they are the whole workbook.

1. You can compute what your channel is worth, in the same units the finance function uses, and that computation is the strongest thing you can bring to a scheme review.
2. You can see what your channel is still blind to — the dimensions in its own kernel — and naming those is how a gainshare gets better rather than quietly dying.

---

PART ONE — DISCOVERY

Days 1–30: read your own channel

Exercise 1.1 — The four parts, in writing (2 hours)

A gainshare has four parts. If one is missing it is a discretionary bonus wearing the word.

1. A baseline. What was true before. Agreed, signed, dated.
2. A measure. How improvement is computed, precisely enough that two people get the same answer.
3. A share. The percentage of verified improvement that returns to the people, as a number.
4. A period and a verifier. When it is computed, and by whom.

Write yours out. Then ask the question this chapter adds to the standard four:

What is the attribute vector of your work, and which single dimension is the measure projecting onto?

Count the attributes that genuinely differentiate a good week from a poor one. That is your k. A measure is a functional w · x on that vector. At k = 12 it carries 8.3 percent of the gradient and leaves eleven dimensions — 91.7 percent — indistinguishable. That is not a criticism of your scheme. It is the specification of every scheme, and knowing it is how you improve yours.

Exercise 1.2 — Where the gain actually comes from (one week)

You can see things the finance function cannot. Sweep for gains that exist and are not counted:

An uncounted gain is an unshared gain. Making one countable is the highest-leverage thing available to you inside a gainshare, and it is the move this chapter equips you to make in the firm's own language.

Exercise 1.3 — The appreciative conversation (45 minutes, with your team)

"Think of a time here when something went unusually well. Not the biggest win — the one that surprised you. What were the conditions? What did we do that we do not normally do?"

Take notes on conditions, not outcomes. Conditions are repeatable causes, and repeatable causes are what a gainshare pays for. Then ask the second question: which of those conditions does our measure currently see, and which does it not? The second list is your kernel, in your own words, from your own floor.

---

PART TWO — THE ARITHMETIC

Days 31–45: compute what the channel is worth

Exercise 2.1 — Your share, calculated yourself (90 minutes)

  verified improvement  =  current period measure  -  baseline
  pool                  =  verified improvement  x  share %
  your line             =  pool  x  your allocation basis

Three things to check that people rarely check.

• Net or gross of the cost of achieving it? It should be net. A scheme sharing gross while the firm absorbs the cost does not survive a finance review, and a scheme that dies has paid you nothing.
• What is the allocation basis? Headcount, hours, salary-weighted, or team-assessed. Each is defensible and each gives a different answer.
• Is there a cap, a floor, or a carry-forward? A floor protects a bad period; a carry-forward lets a missed period be recovered. Both are good design and their absence is worth asking about.

Exercise 2.2 — What does the baseline do when you succeed? (30 minutes)

If the baseline resets to the improved level each period, every gain raises the bar you are measured against and the same effort yields less each cycle. That is baseline ratcheting and it is how gainshares quietly die.

In this chapter's terms it is an entropy problem. A measure whose baseline chases its own result stops carrying information about your effort and starts carrying information about last period's baseline. H(p) collapses, the signal goes to zero, and the scheme is still running while transmitting nothing.

Find out which yours does. If nobody can tell you, that is the finding, and it is the most valuable thing you will produce this month.

Exercise 2.3 — The value of the bit you are proposing (90 minutes)

When you propose adding an attribute to the scheme, bring the arithmetic.

  without information        min(p L, c)
  with perfect information   p c
  EVPI                       min(p L, c) - p c
  bits supplied              H(p)
  value per bit              EVPI / H(p)

Worked at the chapter's inputs: p = 0.10, L = 1,000,000, c = 50,000. pL = 100,000; c = 50,000; pc = 5,000; EVPI = 45,000; H(0.10) = 0.4690; value per bit 95,950 — about 9.21 times the 10,421 a bit that general price discovery costs to produce.

Bring a number and you are negotiating. Bring a description and you are appealing. This is the difference the chapter buys you, and it costs ninety minutes.

---

PART THREE — DESIGN

Days 46–70: propose the dimension the scheme is missing

Exercise 3.1 — Rank your candidates (half a day, with two colleagues)

Take the kernel list from Exercise 1.3 and score each candidate attribute on p, L and c. Rank by value per bit, not by how strongly anyone feels about it.

Then apply the discipline that will make your proposal credible: strike out everything you already know the answer to. An attribute on which the team is certain carries no bits and can change nothing, however much it matters. H(p) is zero at both ends and so is EVPI, which peaks at p = c/L — 5 percent in the worked case, where EVPI reaches 47,500.

A proposal that arrives having already removed its own weakest items is read differently from one that has not.

Exercise 3.2 — Cost the channel and find its floor (half a day)

Every measurement costs something to run. Compute what yours costs a year and what scale it needs:

  MRV cost 20,000/yr against output:
  500,000 units -> 0.04 per unit      5,000 units -> 4.00 per unit
   25,000 units -> 0.80 per unit      1,000 units -> 20.00 per unit

If your unit is below the floor, do not abandon the proposal — pool it.

  members  >=  channel cost / EVPI per member
  20,000 / 1,500  ->  14 members, at 1,429 each

Fourteen teams, fourteen sites, fourteen small suppliers. Regulation (EU) 2018/848 already recognises exactly this structure for organic certification: a group of operators sharing one channel through an internal control system. A channel too expensive for one holder is affordable to fourteen, and that sentence has carried more schemes over the line than any argument about fairness.

Exercise 3.3 — Compress it to four bits (60 minutes)

Whatever you propose to measure, design what it looks like on the day. Four bits — sixteen possible messages. Then one.

Chile's front-of-pack octagon is a few bits and changed what a country bought; a forty-number nutrition panel on the back of the same package had been there for years and changed almost nothing. The constraint is the reader's capacity, not yours. Your scheme statement is read by a tired person in four seconds. Design for that person.

---

PART FOUR — DESTINY AND DELIGHT

Days 71–90: make it hold after you

Exercise 4.1 — Find who is paid for accuracy (45 minutes)

Grossman and Stiglitz, generalised: a channel whose operators earn nothing from its accuracy will not stay accurate. Ask, in writing, who in the scheme earns something from the measure being right. If the honest answer is nobody, propose someone — and propose it as a design improvement, because it is one.

Exercise 4.2 — Write the retirement clause (30 minutes)

Propose that any attribute whose measured result has not moved for four consecutive periods comes off the scheme and its budget is released. This is counter-intuitive and it is the clause that protects you. A scheme that only ever adds measures eventually costs more to run than it pays out, and then it is cut in one line by someone who never read it.

Exercise 4.3 — Ask for the three things (one conversation)

At your next scheme review, ask for exactly these, in this order.

1. The baseline's behaviour, in writing. Fixed for a stated term, or ratcheting on a published schedule everyone can see coming.
2. One added dimension, with its EVPI, its ceiling, its cost and its value per bit on a single page.
3. A retirement clause, which is the thing that makes the second ask affordable and which almost nobody offers unprompted.

Asking for a retirement clause alongside an addition is what marks the difference between someone asking for more and someone improving an instrument. It is also true, which is why it works.

---

WHAT TO CLAIM, AND HOW

Claim the measurement, not the outcome. Outcomes have many parents and the argument is unwinnable. A measurement you built, ran and published has one author, and it keeps producing after the argument is over.

Claim in the firm's units. EVPI, ceiling, cost per bit, break-even scale, pooling threshold. Every one of those is a number your finance function already knows how to read, and none of them requires anybody to agree with you about anything first.

Claim the kernel honestly. Say which dimensions your proposal still does not see. It costs you nothing, it is the only part nobody else will say, and it is what makes the rest of the page believable.

---

THE LEDGER, AS IT READS FROM INSIDE

A gainshare with the last four lines is a different instrument from one without them, and the four lines are the whole contribution of this chapter to your position.

---

SELF-ASSESSMENT

---

WHEN THE HONEST ANSWER IS "DO NOT MEASURE THIS"

There are attributes of your work that should not go into the scheme, and being the person who says so is worth more to your standing than any addition you propose.

When everybody already knows. If the team is certain about something, H(p) is near zero, EVPI is near zero, and a measure of it costs money and attention and changes nothing. It also invites the worst outcome available: a number that looks like a result and is actually a foregone conclusion, which is how schemes lose their credibility.

When the measure would be a proxy. If you cannot observe the attribute and would have to observe something correlated with it, expect the proxy to be optimised and the attribute not to be. That is Goodhart's mechanism and it is the most reliable failure in this whole field. A proxy is sometimes still worth it — but only with Ostrom's conditions attached: monitors accountable to the people measured, graduated rather than binary consequences, and a dispute step that costs less than the thing in dispute.

When it is below the floor and nobody will pool. If the channel costs more than the decisions it changes are worth, and you cannot find the members to share it, the honest answer is to say so and keep the budget. The European Union reached the same conclusion about its own carbon market and wrote it into Article 27.

Say it out loud, with the arithmetic. A person who brings both the additions and the subtractions is read as an operator of the instrument rather than a claimant on it, and that difference compounds over every review you will ever sit in.

---

APPRECIATIVE QUESTIONS FOR YOUR TEAM

1. When has something we did well been visible in the numbers? What made it visible that time?
2. What do we know about the quality of our work that nothing here records — and which one of those would be worth the cost of recording?
3. If the scheme could see one more thing, what would change about how we work on a Tuesday?
4. What would we want still measured here in ten years, and who would have to care about it for that to be true?