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

A watercolour of rolling hills in late summer, a farmhouse and a line of trees on the far ridge. — Plate IV.05 · Workbook — the Gainshare employeeThe Ten Thousandth.The ten thousandth panel costs less than the first. Nobody economised. Ten thousand people each learned something small and none of them could tell you what.

WORKBOOK — THE LUMINOUS GAINSHARE EMPLOYEE

Chapter IV.05 · Energy Systems That Compound

For the person inside the mechanism. You are on a learning curve right now, you are producing the thing the curve is made of, and this workbook is about making that visible, measured and paid.

WHY THIS CHAPTER IS ABOUT YOU

The chapter's central equation says that cost falls by a stable proportion each time cumulative production doubles. It does not say how. The equation is silent on mechanism, and that silence is where you live.

Go back to what the decomposition studies actually found when they went looking. Not one magic step — a long list of small ones, most of them discovered by people doing the work: a fixture moved closer, a step done in a different order, a tolerance that turned out not to matter, a jig somebody made on a Friday, a check that caught a fault two stations earlier. The exponent is the sum of those, and every one of them has a name and a shift pattern.

So here is the sentence to carry. The learning rate is not a property of the technology. It is the aggregate output of the people who repeat the work, and it is the only economic quantity in this chapter that is produced entirely by labour. A firm that has a 20 percent learning rate has a workforce that produces one, and the equation cannot tell the difference between a factory and a team because there is no difference.

That is not a morale argument. It is a claim about where the value is made, and in a gainshare it is a claim about who should be paid.

PART ONE — DISCOVERY

Days 1–30: find where the gain actually comes from

Exercise 1.1 — The improvement inventory (90 minutes, with your team)

Together, list every change to how the work is done in the last two years that came from inside the team rather than from a project. Include the ones that were never written down. Include the ones somebody was told off for.

Against each, write three columns: what changed, roughly what it saved per unit or per shift, and is it recorded anywhere. The third column is usually empty, and an empty third column is the entire problem this workbook solves. An improvement nobody recorded is an improvement nobody can be paid for, and it is also an improvement the next shift has to discover again.

Exercise 1.2 — Find your own curve (2 hours, with whoever holds the data)

Ask for cumulative units and time-or-cost per unit for your line, in real terms, going back as far as the records allow. Plot log against log and fit a line. Report the learning rate with its interval — a point estimate will be argued with, an interval with a standard error attached will not.

If the answer is a clean positive rate, you have just measured the value your team produces annually and nobody had it before. If the fit is poor, find the regime change; it is usually a product or tooling change and it is usually datable to a month.

Exercise 1.3 — The appreciative interview, upward (45 minutes)

Ask the longest-serving person on the line: what is different between the tenth one and the thousandth? Then ask the second question, which is the one that matters here: who taught you that, and who have you taught it to?

You are mapping how the learning propagates. In most teams it moves by apprenticeship and it dies when people leave. The value of writing it down is not documentation for its own sake — it is that a learning curve that lives only in hands can be lost, and one that is written down cannot. That distinction is worth money and it is worth saying so in those words.

PART TWO — THE ARITHMETIC

Days 31–45: compute what the improvement is worth

Exercise 2.1 — The gain, worked end to end

Here is a full worked case at the scale most teams operate at. Fourteen people. A process improvement that moved a line from $46.00 a unit to $37.40 across 18,000 units a year, under a 30 percent gainshare with a twelve-month baseline.

  unit saving                        $8.60
  annual gain                        $154,800
  gainshare pool at 30%              $46,440
  per person across 14               $3,317
  saving as a share of the old cost  18.7 %

Three things to notice before you take this into a room.

First, the pool is computed from the gain, not from a budget. That is the whole difference between a gainshare and a bonus, and it is the thing to confirm in writing before you do anything else.

Second, 18.7 percent is roughly one doubling's worth of learning at a rate in the range this chapter measures — the fitted solar figure is 22.5 percent per doubling with an interval of 19.1 to 25.7 percent. Your team produced in a year what the world's most-studied learning curve produces per doubling. Say that out loud, with both numbers, and it changes the register of the conversation.

Third, the saving persists. The pool is annual and the cost reduction is permanent until something changes it. Over five years the firm keeps $154,800 a year and the team is paid $46,440 a year, and that is a defensible arrangement — but only if somebody has said the second half out loud.

Exercise 2.2 — The baseline conversation (60 minutes, before anything else)

The baseline decides everything and it is decided once. Get these four in writing:

The period. Twelve months is standard. A shorter one imports seasonality.
The reset rule. Does the baseline reset to the improved level each year? If it does, you are being paid once for a saving the firm keeps forever, and that is a legitimate design — but it must be stated, not discovered in year two.
The volume adjustment. If volume rises, does the gain rise with it? At 18,000 units a year the answer is worth $8.60 a unit to somebody.
The exclusions. What counts as a capital-driven saving rather than a team-driven one, and who decides?

An unagreed baseline is not a baseline. It is a future dispute with a date on it.

Exercise 2.3 — The counter-argument, rehearsed

Somebody will say the saving would have happened anyway. Prepare the honest answer rather than the indignant one, because the honest answer is stronger.

The chapter gives you the evidence. Cumulative production does not cause cost reduction. France built 58 reactors across 5.86 doublings and real cost rose by a factor of about 3.50× — a learning rate of −23.8 percent. The published spread for onshore wind capital cost runs from −11 percent to 32 percent. If repetition alone produced improvement, neither of those numbers could exist.

Improvement is not automatic. It is produced, by named people, and where it is absent you can see the absence in the data. That is the strongest sentence in this workbook and it is supported by the chapter's own negative cases rather than by its positive ones.

PART THREE — DESIGN

Days 46–70: make the uncounted countable

Exercise 3.1 — The improvement register (one hour to build, five minutes a week)

One shared sheet. Six columns: date, what changed, who found it, estimated effect per unit, how it was measured, and whether it has been propagated to the other shifts.

Two rules keep it alive. Anybody can add a row, and no row is ever deleted — a change that did not work stays on the register with its result, because a tried-and-failed row saves the next person the attempt. That is the same discipline the rest of this edition applies to its own records, and for the same reason.

Exercise 3.2 — Propagate deliberately (ongoing)

The single largest source of unclaimed gain in most operations is an improvement that exists on one shift and not the other two. Take the top five rows of the register and check, physically, whether every shift is doing it. Where they are not, that is measurable money and the measurement is easy, because you have the before and after on one shift already.

Exercise 3.3 — Ask for the second axis (45 minutes, one conversation)

Most gainshare schemes measure cost. Very few measure the rate at which cost falls. Propose a second measured quantity: the fitted learning rate on your line, re-estimated annually with its interval.

Why this is worth asking for, in the firm's own language: a one-off saving is a step and a learning rate is a slope, and a slope is worth more to a capital committee than a step because it can be projected and committed against. If your firm is negotiating volume commitments with customers — the instrument the executive workbook builds — your measured rate is an input to a contract, and people whose work is an input to a contract are paid differently from people whose work is a cost line.

Exercise 3.4 — Read the ledger yourself (60 minutes)

Ask to see how the pool was calculated. Not the total — the calculation. You are entitled to four numbers and they should be on one page: the baseline unit cost, the achieved unit cost, the volume the saving was applied to, and the share percentage. In the worked case those are $46.00, $37.40, 18,000 and 30 percent, producing a gain of $154,800 and a pool of $46,440.

Check the volume figure first. It is the one that moves most between years and it is the one most often taken from a plan rather than from actuals. A gain computed on planned volume when actual volume was higher is a real and quiet underpayment, and it is arithmetic rather than accusation to point it out.

Exercise 3.5 — The conversation, scripted (30 minutes to write, ten to have)

Write it down before you have it. Four sentences, in this order, and no more:

Here is what the line cost before and what it costs now — $46.00 to $37.40, 18.7 percent, on 18,000 units. Here is the register showing who found each change and when. Here is the fitted rate for our line with its interval, so you can see it is a slope and not a one-off. What I am asking for is that the rate is measured annually and that the register is a named job, five minutes a week.

Notice what that script does not do. It does not complain, it does not compare the team to anybody, and it does not ask for a judgement about fairness. It hands over three artefacts and makes one specific, cheap, durable request — which is the request most likely to be granted and the one that compounds.

PART FOUR — DESTINY AND DELIGHT

Days 71–90: make it hold

Three things make a gainshare survive.

The register is somebody's named job. Five minutes a week, in the handover. Not a volunteer role, and not an additional duty for whoever is most conscientious, which is how it quietly becomes unpaid work.

The re-fit is annual and it is on a date. The learning rate is measured once a year, in the same week every year, by the same method, and the result goes on the same wall the safety figures go on.

The pool is paid on time. A gainshare paid late is a gainshare nobody will run twice, and the second year is where all the value is.

The failure modes, named honestly.

The baseline resets silently. Year one pays, year two does not, and nobody said so. This is the most common failure and it is prevented entirely by Exercise 2.2.
The improvement is captured as a headcount reduction. If the gain is realised by removing the people who produced it, the second doubling does not happen. It is worth saying this early, in a room, without heat: the exponent is made of the workforce, and a firm that spends it once has spent it.
Only the measurable improvements count. A change that reduced rework, prevented a fault or shortened a handover may not show up in a unit cost for two quarters. Record it anyway, with an estimate and the word estimate.
One person carries it. Recruit a second owner before you need one, and recruit them by giving them the credit for the first result.

Know your scheme — the checklist

What is the baseline, and when was it last reset?
What share of the gain goes to the pool, and how is the pool divided?
Are capital-driven savings excluded, and who adjudicates?
Is there a volume adjustment, and in which direction?
When is it paid, and against what verification?
Who re-fits the learning rate, and does anybody outside the team see it?

Delight — and this one is real

There is a particular satisfaction in watching a number you produced with your hands appear in a document written in a language you did not choose. The improvement register turns a Friday afternoon fixture into a line in a capital paper, and the capital paper turns it into a contract, and the contract turns it into a price the world pays.

The chapter's forty-seven-year curve is made entirely of that. Nobody on the line in 1976 knew they were producing the most reliable cost series in industrial economics. They were just doing the next one slightly better than the last one, and writing down how long it took.

APPRECIATIVE QUESTIONS FOR YOUR TEAM

What is something we do now that we worked out ourselves — and who worked it out?
When did we last teach another shift something, and what happened to the numbers on that shift?
If our improvement register had existed for five years, what would be in it that we have already forgotten?
What would it be worth to us for the rate we improve at to be measured and named, not just the cost we reached?
Where has somebody outside this team taken a number we produced and used it well — and how did that feel?
What would we want to be true about this line in ten years, and what is the first row of the register that gets us there?