Haute Lumière

Commerce · III.10 · MMXXVI · daylight

La Bourse  /  Volume III  /  Nº III.10  /  Workbook — the executive

WORKBOOK — THE CORPORATE EXECUTIVE

Chapter III.10 · Reserves and Resilience

For the executive with a P&L, a balance sheet and a board. Ninety days to move the firm's reserve position from a set of inherited numbers nobody can defend to a layered, priced policy with a published break-even. The language is the firm's and no apology is offered for it.

---

THE PREMISE, STATED COMMERCIALLY

Your firm holds reserves. You are almost certainly holding them in the most expensive available form, against events nobody has written down, and defending them with the word prudence — which is why they are cut in every working capital programme and rebuilt in a panic after every incident, at the top of the market, which is the worst possible time to buy insurance.

Three commercial facts run this workbook.

One: a buffer's cost is the spread, and the spread is large. Cash at 2 percent inside a business returning 12 percent costs ten points a year. On $50m that is $5.0m — a line most firms have never seen, because it appears nowhere.

Two: the cost collapses if you change the form. A committed undrawn revolver at 37.5 basis points is $0.188m on the same $50m — 26.7× cheaper. Consigned stock at a 6 percent supplier fee is $3.0m against $12.5m to carry it yourself at 25 percent — 76 percent lower, and it stays off working capital under IFRS 15.

Three: there is one number that ends the argument, and it is a frequency, not a value:

        loss the event causes  ×  share of it this reserve prevents
   T* = ───────────────────────────────────────────────────────────
                    all-in annual carrying cost

Hold the reserve if the event arrives more often than T*. Then put the firm's own incident log beside it. That is the whole board paper.

---

PART ONE — DISCOVERY

Days 1–30: the reserve register

Exercise 1.1 — Find every buffer the firm already holds (two weeks)

Most firms discover between eleven and twenty when they look properly, and can price fewer than half. Sweep these:

For each: the amount, the form, the annual carrying cost, and the event it is against. The last column will be empty for most rows. That is the finding.

Exercise 1.2 — Price every line (one week)

   dead reserve annual cost  =  capital held × (ROIC − return the reserve earns)
   physical reserve          =  value × carrying rate (capital + storage + obsolescence)

Use 20–30 percent for physical carrying; the chapter uses 25. Total the column. In a mid-sized industrial firm this is typically 1–3 percent of revenue and has never appeared on a single slide.

Exercise 1.3 — Pull the incident log (one week)

Not a risk register — a log of what has actually happened: every supply interruption, outage, demand shock and liquidity event in the firm's record, with its date and its cost. Ten to twenty years if the records reach.

This is the denominator of everything that follows and almost no firm has it in one place. Where the record is thin, say so and widen the frame to the sector — four correlated semiconductor shocks between 2011 and 2022 gives an interval of 3.0 years, and that is a defensible external prior when your own log is short.

---

PART TWO — THE ARITHMETIC

Days 31–45: size the buffers properly

Exercise 2.1 — Re-derive safety stock with the term that matters (three days)

   σ_DL = sqrt( L·σ_d²  +  d²·σ_L² )        safety stock = z · σ_DL

Pull actual lead-time distributions from your ERP — not the planning parameter, the realised distribution, which is usually different and always wider. In the chapter's worked case the lead-time term was 96.34 percent of the variance in 2022 against 77.37 percent in 2019, and the required buffer multiplied by 3.58× with no change in demand at all.

The commercial consequence: your safety stock is your supplier's variance, capitalised on your balance sheet. Put supplier lead-time standard deviation on the vendor scorecard this quarter, and price a reduction — you can now say exactly what a week of σ_L releases in working capital.

Exercise 2.2 — Choose the service level, do not inherit it (two days)

   1 − Φ(z*)  =  h·Q / (p·D)

At h = $12/unit/yr, Q = 26,000, p = $240 and D = 52,000 this gives z* = 1.96, a 97.5 percent service level — and the cost curve around it is a basin: half a z-unit short costs 17.35 percent more, half a z-unit long 9.10 percent.

Then run the ladder and look at the right-hand column: the price of avoiding one unit of shortage rises from $164.82 at 95 percent service to $29,834.20 at 99.99. If any product line in your firm carries a 99.9 percent internal service target that nobody derived from p, you have found money.

Exercise 2.3 — Stress the tail (two days)

Basel moved from 99 percent VaR to 97.5 percent Expected Shortfall in 2019 for a reason. Under a normal loss the two are within 0.49 percent of each other. Under a Student-t with five degrees of freedom and identical variance, expected shortfall is 16.68 percent higher.

Ask your planning team which distribution their buffers assume. The answer is normal, always, because that is what the software does. Add the fat-tail loading explicitly or accept that you are a sixth short on the events that matter.

Exercise 2.4 — Compute T* for the top five (three days)

The recovery fraction is the number to fight about. Toyota's post-2011 chip buffer bought roughly two quarters of a shortage that ran over a year — well under 1.0. Run each reserve at 15, 40 and 100 percent recovery and present the range, never a point.

Worked, on the industry case: a three-month chip buffer costs $2.41bn a year against $25.20bn of forgone operating profit. At 40 percent recovery, T* = 4.2 years against an observed interval of 3.0. It pays, and the assumptions are the argument.

---

PART THREE — DESIGN

Days 46–60: the instrument

The Standing Reserve Facility — three layers against one named scenario.

Fill from the top. Layer 4 is the most expensive reserve available and it is almost always the one built first, because it is the only one that looks like a reserve on a balance sheet.

Layer 1, priced honestly. A facility is a promise, and promises are correlated with the shock: in March 2020 corporates drew revolvers on the order of a quarter of a trillion dollars in weeks precisely because they doubted availability. At an assumed 70 percent effective availability it is still 18.7× cheaper than cash — so it is a layer, never a replacement. Read the MAC clause yourself.

Layer 2, and the objection that is really a covenant objection. Under IFRS 15 title has not passed on consigned stock, so it does not consolidate into working capital or touch leverage. Get the finance director's carve-out for strategic buffer stock from the incentive metric in writing, before deployment. A buffer that scores against the holder's own bonus has a half-life of two quarters.

Layer 3, where the operating leverage is. Retained skill, maintenance practice, repairable design, a supplier paid above the cheapest bid. Its carry is at or below zero because it produces in ordinary periods. Germany held six million employment relationships through 2020 and its unemployment rose 0.90 points against 11.20 in the United States — a factor of 12.4 — at a replacement cost avoided of roughly $168.8bn across that cohort.

Balance sheet treatment. Layer 1 is disclosure until drawn. Layer 2 sits with the supplier and is priced into unit cost. Layer 4 is IAS 2 at the lower of cost and net realisable value, and it will hit the working capital metric — so pre-agree its treatment. Where the buffer improves a long-lived asset rather than sitting as stock, capitalise and depreciate over its regenerated life, and open that conversation with the auditors early; useful economic life is a discussion they have annually.

Who pays — and why this must be settled in the design. In 2021 General Motors' revenue fell 7.43 percent while net income rose 48.83 percent. US used-vehicle prices rose 45.2 percent in the year to June 2021, an illustrative $452bn across that market. The buffer's cost sits on a producer's balance sheet; its benefit sits in a buyer's driveway. If your proposed reserve is not mandated, mutualised or regenerative, it will be cut, and cut by someone behaving rationally. Design it into one of the three on day one: a regulatory requirement, an industry pool or captive, or a living asset whose ordinary-year return funds its own cover.

---

PART FOUR — DESTINY AND DELIGHT

Days 61–90: publish it, and draw it once

Exercise 4.1 — Publish T* into the standing pack (day 61)

One line per reserve: the event, the annual carry, T*, and the observed interval from the incident log. Anything reviewed monthly persists; anything reviewed by exception erodes. This single line converts the annual attack on the reserve from a values argument you can lose into an arithmetic check you usually win.

Exercise 4.2 — Run one drawdown drill (days 70–85)

Small, deliberate, documented. Draw part of the buffer, use it, replenish it. What you are testing is not the quantity — it is the authority chain. The failure mode of an untested reserve is not emptiness. It is that nobody has the standing to open it in the eight hours when opening it mattered. Svalbard's credibility rests on the 2015 ICARDA withdrawal; the SPR's on 2022.

Exercise 4.3 — Set the review trigger (day 90)

Write the condition under which the reserve is resized, in both directions. In both directions is the part that makes it credible: an executive who will only ever argue for more reserve is discounted, correctly, and stops being listened to on the cases that matter.

---

THE FAILURE MODES, NAMED

So you can see them coming

Over-reserving, which loses by compounding. A 20 percent cash buffer drags growth from 12 to 10 percent. Over twenty years that is 9.646× against 6.727× — the lean competitor ends 43.39 percent larger. Making ruin absorbing and solving for the probability that equalises them: p* = 1.79 percent a year, or once in 56 years. Below that frequency, the unbuffered competitor wins.

Note what moves the threshold: at a 6 percent buffer return it is once in 93 years; at 9 percent, once in 187. The trade-off is not efficiency against resilience. It is efficiency against dead reserves.

The normal-tail buffer. A sixth short before anything else goes wrong.

The correlated reserve. The living layer is embedded in the system it insures. The skill buffer is in the same firm as the downturn. Keep a dead layer sized precisely for the correlated case.

The reserve nobody may open. Untested authority, every time.

The good decade. Erosion, not cancellation. One quarter it is trimmed, the next it is not replenished, and four years later it is a number in a policy nobody has read.

---

BOARD PAPER TEMPLATE

Reserve policy — [business unit]

The event. [One sentence with a duration.] Our own record shows it [n] times in [m] years, an interval of [m/n] years.

Current position. We hold $[X] against it, in [form], carrying at $[Y] a year — [Z] percent of EBIT, currently unstated in any pack.

Break-even. T = (loss $[L] × recovery [r]) / carry $[Y] = [T] years, against an observed interval of [m/n] years. Range at 15–100 percent recovery: [low]–[high] years.

The proposal. Re-layer: $[a] to a committed facility at [bp], $[b] to consignment at [%], $[c] retained in own stock. Annual carry falls from $[Y] to $[Y'], releasing $[ΔWC] of working capital, with cover unchanged against the named scenario.

What we are deliberately not covering. [The event whose return period exceeds T*.] Stated so the board owns the choice.

Review trigger. Resize if [condition], in either direction.

---

APPRECIATIVE QUESTIONS FOR YOUR LEADERSHIP TEAM

1. When has this firm absorbed a shock without the market noticing? What absorbed it, and had anyone costed that thing beforehand?
2. Which of our reserves has actually been drawn in the last five years, and what did we learn about our own authority chain that day?
3. Where are we already holding a living reserve — skill, maintenance, a supplier relationship — and who has been funding it out of their own line?
4. If supplier lead-time variance sat beside price on our scorecard, what would we start paying for that we do not pay for now?
5. Which three firms would we trust enough to hold one pooled buffer with, and what would each of us release?
6. What is the event our largest reserve is actually against — said aloud, with a duration in it?

---

THE NINETY DAYS ON ONE PAGE

The one number on the front page is T against the observed interval. If your own log shows the event twice in eighteen years and T is thirteen, the argument is finished — and it was finished numerically, which is the only way it stays finished through a change of chief executive.