Haute Lumière

Commerce · I.05 · MMXXVI · daylight

La Bourse  /  Volume I  /  Nº I.05  /  Workbook — the Gainshare employee

WORKBOOK — THE LUMINOUS GAINSHARE EMPLOYEE

Chapter I.05 · The Pilot That Pays

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 the one that decides what lands in your statement, because a gainshare pays on verified improvement, and verification is a design problem solved in week one.

---

WHY THIS CHAPTER IS YOURS

A gainshare has four parts: a baseline, a measure, a share and a verifier. Three of the four are decided by how the pilot was designed — before anybody worked a single hour differently.

So the arithmetic in Chapter I.05 is not background for you. It is the arithmetic of your own pay. Specifically:

• Power decides whether an improvement you genuinely made can be seen. An under-powered design returns "no significant change" and pays nothing, and the work was real either way.
• The realisation rate φ decides whether a measured saving becomes money at all. A saving nobody's budget releases is a saving no scheme can distribute.
• The step function decides whether a real 55 percent improvement banks zero — which is the commonest way good work goes unpaid, and it is entirely predictable in advance.
• The verifier's independence decides whether the number holds when somebody senior wishes it were smaller.

None of these is a matter of fairness or goodwill. They are design decisions, they are made early, and you can ask about every one of them before the pilot starts. That is the whole of this workbook.

---

PART ONE — DISCOVERY

Days 1–20: read the design, not the promise

Exercise 1.1 — Interrogate the design (2 hours)

Take the scheme document and the pilot design, and answer these in writing. Where you cannot answer, that is the question to ask.

1. What is the comparison? Before-and-after in this unit alone, or this unit against others? If there is no comparison group, every seasonal effect and every market movement lands in your result — in both directions.
2. How many pre-periods? If one, the detection threshold is 3.96σ and the design is nearly blind. If twelve, it is 1.81σ. This single answer moves what you can be paid for by a factor of two on the effect.
3. What is σ? The period-to-period standard deviation of the measure. If nobody can tell you, nobody knows whether your improvement is detectable.
4. Has serial correlation been accounted for? Twelve correlated months at ρ = 0.5 carry the information of 4.5. A design that treats them as twelve is overstating its precision by 63 percent, and overstated precision cuts both ways.
5. Is the analysis written down and dated? If it is chosen after the result, it can be chosen against you as easily as for you.
6. Who verifies, and to whom do they report? If the verifier reports to the person whose budget pays the gainshare, the independence is nominal.

None of these questions is adversarial. Every one of them makes the scheme better for the firm as well. A scheme that pays out on a design nobody can defend is a scheme that will be withdrawn after the first contested result.

Exercise 1.2 — Find the step (1 hour)

This is the question that most often decides whether a year of work produces a statement with a number on it.

Ask: what has to fall for this saving to be banked? A person, a shift, a vehicle, a lease, a licence band, a contracted volume, a line on an invoice.

  claimable  =  step × floor(saving / step)

   saving 12 % of a shift   ->  banked 0      φ = 0.00
   saving 55 % of a shift   ->  banked 0      φ = 0.00
   saving 110 % of a shift  ->  banked 1 shift  φ = 0.91

If your improvement is heading for 55 percent of a step, say so now, in month one, in writing. There are three good routes and all three are better raised early:

1. Aggregate. Combine with other improvements landing on the same step, and have the scheme recognise the aggregate.
2. Redefine the measure onto a continuous line — energy, consumables, freight, contracted volume — where every unit counts.
3. Have the scheme recognise capacity released, explicitly and at a stated rate, as a distinct category from cost avoided.

Raised in month one, this is a design improvement. Raised in month twelve, it is a dispute.

Exercise 1.3 — The appreciative interview (45 minutes)

Find someone who has been through a full gainshare cycle and ask:

"Tell me about a time the scheme paid out on something you were proud of. What made it possible to measure? What did somebody do early that made the payment straightforward later?"

Take notes on the design decisions, not the amount. People who have been paid properly remember exactly which decision made it possible, and it is almost always something settled before the work began.

---

PART TWO — THE ARITHMETIC

Days 21–45: the numbers that are about your pay

Exercise 2.1 — Compute your own detectability (2 hours)

Do this for the measure your gainshare actually pays on.

1. Get twelve or more periods of history for the measure.
2. Compute σ.
3. Compute the lag-1 correlation ρ and inflate σ.
4. Note m (pre-periods in the baseline) and k (post-periods in the payment window).
5. Compute detectable effect = 2.8016 · σ · √(1/m + 1/k).

Now compare it to the improvement you believe your team can make. If the improvement is smaller than the detectable effect, the scheme cannot pay you for it as designed, and that is a fact about the design rather than about your work.

Take it to the scheme owner with the three fixes, in this order: extend the baseline period (free, it is history); lengthen the measurement window; or add comparable units so there is a control arm. The first costs nothing at all.

Exercise 2.2 — Compute the realisation rate on your own line (90 minutes)

      φ  =  banked saving / measured saving

Take last cycle's verified improvement, if there was one, and trace it:

• Which budget line fell?
• By how much?
• On what date?
• Whose budget was it?

φ is the ratio. If you cannot trace it, φ is unknown, and an unknown φ is where a scheme quietly stops paying while everyone continues to work.

This is the single most valuable number you can produce this month, and it is valuable to the firm too: a measured φ is what lets the next pilot be underwritten instead of argued about.

Exercise 2.3 — Check the baseline for ratcheting (1 hour)

Two things to establish, both in writing:

1. Does the baseline reset to the improved level each cycle? If it does, you are on a treadmill: every gain raises the bar, the same effort yields less each cycle, and eventually nothing. A well-designed scheme holds the baseline for a stated term — three to five years is typical — or ratchets on a published, gradual schedule everybody can see coming.
2. Is the baseline protected against regression to the mean? If the pilot was placed at the worst-performing unit, that unit improves by roughly (1 − ρ) × distance from the mean on its own — about 0.51σ at the 10th percentile with reliability 0.6. Paying out on that is generous this cycle and indefensible next cycle, and schemes that pay on artefacts get closed.

Argue for the control group. It is the one structure that protects the scheme's credibility and your payment at the same time, because it cancels the regression identically in both arms.

Exercise 2.4 — The honest negative, applied to you (30 minutes)

Write the strongest case against your own claim for this cycle. Not a straw version — the version that troubles you. Seasonality? A market movement? A co-intervention somebody else ran at the same time? A definitional change in the system?

Then bring it to the verification meeting yourself, before anyone else raises it.

The person who names the confounder first is the person whose remaining claim is believed. This is not a concession. It is the highest-return move available to you in the entire cycle.

---

PART THREE — DREAM AND DESIGN

Days 46–70: what to ask for, and how to ask

Exercise 3.1 — The six asks (2 hours)

Six requests. Every one of them improves the scheme for the firm as well, which is why they are askable.

1. A pre-registered analysis. One page, dated before deployment, naming the primary measure, the period, the exclusions and the failure condition. Ask for a copy.
2. A baseline of at least twelve periods, or a stated reason why fewer.
3. A control group or a staggered order, drawn at random, from comparable units in the same estate.
4. A named, independent verifier who reports to finance rather than to the sponsor, named before the result is known.
5. A claim map showing which budget line each saving falls out of, the step it must cross, and whose budget falls — shared with the people in the scheme, not held by finance alone.
6. A stated baseline term, with any ratchet published in advance.

Ask for them as design questions, not as demands. Every one of the six is something a scheme owner will want on the day the first result is contested.

Exercise 3.2 — Contribute where the design says the gain is (1 hour)

Look at the claim map and find where a marginal hour of your effort moves a banked pound rather than a measured one. They are often different places.

Write three specific actions you can take this cycle that:

• affect the primary pre-registered measure, not a secondary one;
• push a saving across a whole step rather than further into a fraction;
• are recorded in a system the verifier will read.

The third is not administration. An improvement the verifier cannot see did not happen, as far as the ledger is concerned, and the ledger is what pays.

Exercise 3.3 — Write the one page (45 minutes)

At the end of the cycle you want one page in the verification meeting that somebody who does not know you can read in ninety seconds:

1. The baseline, with its dates and its signature.
2. The intervention, with the date it started.
3. The measured result, against the pre-registered analysis.
4. The confounders, named by you, with what you did about each.
5. The banked amount, with the budget line and the date it falls.
6. The share due under the scheme, computed.

Item four is the one that makes the rest believed.

---

PART FOUR — DESTINY AND DELIGHT

Days 71–90: make the scheme outlast the enthusiasm

Exercise 4.1 — Read the fund's arithmetic (30 minutes)

The gainshare and the pilot fund are the same instrument seen from two sides. The firm's version:

      r  =  φ · S / C            doubling  =  ln 2 / ln(1 + r)
   φ 0.65  S 96,000  C 120,000   ->   r = 0.52/yr   doubles in 1.66 years

Yours is the same equation with your share in the numerator. Both halves grow from the same thing: a verified saving that lands in a named budget. That is why pressing for a clean design is not adversarial. It is the only version in which there is anything to share.

Learn to say the portfolio number too: p·φ·S / (C + V + A) = 27.7 percent against a 9 percent WACC, break-even at a 19.5 percent success rate. It is the sentence that keeps the fund funded, and the fund is what keeps the scheme alive.

Exercise 4.2 — The three failure modes, from the inside (1 hour)

• The baseline ratchets quietly. Watch for it every cycle. Ask for the term in writing.
• The verifier drifts into advocacy. It looks like success and it ends in a withdrawn scheme after the first contested payment. Ask for rotation.
• The fund is raided. A gap appears elsewhere and the replenishment is swept into it. Ask for the reversion clause to be written and for the fund's regeneration rate to sit in the standing pack, where removing it requires an explanation.

Raise each once, concretely, with the fix attached, and then let it go. You are improving an instrument, not conducting an audit.

Exercise 4.3 — Delight, and why it belongs here (30 minutes)

The best moment in a gainshare cycle is not the payment. It is the verification meeting where the sceptical question has already been answered on page one, in writing, dated before anybody knew the result — and the room moves straight to what to do next.

The second best is the cycle after, when somebody in another team asks you for your one-page template. The method has stopped being yours. That is when it becomes how things are done here, and it happens without a mandate.

Write one sentence: the part of this I look forward to is ___. If you cannot complete it, the scheme has become a report rather than a practice, and that is worth raising.

---

THE CYCLE'S WORK — what you hold at day ninety

1. A written answer to all six design questions in Exercise 1.1.
2. Your own σ and detectable effect for the measure that pays you.
3. A measured φ from last cycle, traced to a budget line.
4. The step identified, named, and the route chosen if the saving falls short of it.
5. A copy of the pre-registration, dated before deployment.
6. The one page, with the confounders named by you.

---

SELF-ASSESSMENT

The two that matter most are the third and the sixth. The third decides whether there is money. The sixth decides whether you are believed about it.

---

APPRECIATIVE QUESTIONS FOR YOUR TEAM

1. When has this scheme paid out on something we were genuinely proud of, and what was decided early that made it measurable?
2. Which of our measures has the longest clean history, and what could we now ask of it?
3. What is the smallest whole step one of our improvements could cross this cycle, and what would it take to cross it?
4. If we named every confounder against our own claim before the verification meeting, what would change about how that meeting goes?