Right Is Luck. Early Is a Decision.

The board is not yours; it records what happened. Everything to the right of it is a choice. Your chalk touches the board and draws your line. The work is at the far end, supplying the force, and the fulcrum between them decides how much chalk movement one unit of work buys. Reserve is force — capacity you have built and are holding, and the reason a company can meet demand that arrives earlier or larger than it planned for. Automation is ratio — slide the fulcrum toward the board and the work end travels further for less force. They are not substitutes. Force with a poor ratio barely moves the chalk; a good ratio with no force has nothing to push. Most companies have invested heavily in one and assumed it covered the other. Agility is the sweep, and it is a rate: not how high you can draw, but how fast you can change where you are drawing when the board turns. Before deciding where to put the fulcrum, it is worth being exact about the gap the chalk is drawing.

There is only one distance in the picture, between your line and the market's. Read it above and it is capacity waiting; demand that arrives early or large finds you already able to serve it. Read it below and it is catch-up already being paid for; everything built now costs more. Which side you sit on is decided years before the demand arrives, and it is not a matter of temperament. So the question is how much reserve, and that turns out to be a ratio rather than a judgement.

Capacity is planned demand plus a reserve, and one ratio sets the reserve. The ratio is what a lost sale costs you against what an idle unit of capacity costs you — two figures your business already has, the contribution lost on an order you could not serve and the carrying cost on a unit you did not use. When missing demand is the more expensive mistake you provision above the plan. When idle capacity is, you provision below it, and running slightly short is then correct rather than a failure of nerve. Planned demand is the thesis multiplied out: what the market does, times the share you expect to hold. The reserve is the only part the ratio decides, and it is the part almost nobody sets deliberately. Whether a reserve set this way earns its cost was tested against a real market, not asserted. *Basis. No general claim is made that catching up costs more than building early. A reserve carries real cost in inventory, obsolescence and the risk of building the wrong thing. The ratio is measured, not assumed.*

Perfect foresight with no reserve turns out to be worse than a bad guess with one. This is US single-family housing starts, every month from 1990 to 2026 — the boom, the collapse, the long climb back, a five-fold swing peak to trough, which is what a real market looks like. The board was used to ask which lever actually moves the cost of planning capacity: knowing what is coming, or leaving room for what you do not. Every month, capacity was set for the following three years and scored, each unit of demand that could not be served at 42 times the cost of a unit left idle. Two planners: one that knows the next 36 months exactly, and one that assumes next year looks like last month. Then the reserve each of them holds was swept, from nothing to double. Perfect foresight with no reserve cost $60.8M. The flat guess, wrong every single month, holding a 45% reserve, cost $21.6M. The same perfect foresight, once it held a 30% reserve, cost $11.3M. A bad forecast holding a reserve beat a perfect forecast holding none, by 2.8 times, and the perfect forecast beat its own no-reserve self by 5.4 times. Foresight is worth having. It is worth roughly half what the reserve is worth, and it is worth nothing at all without one. The result is not a quirk of housing; it follows from what a forecast is. *Method. Walk-forward over HOUST1F, January 1959 to July 2026. At each month a capacity is set for the following 36 months and every month in that window is scored: demand above capacity costs 42 per unit, capacity above demand costs 1; 27,036 scored months per run. The 42:1 asymmetry is illustrative — substitute your own and the curve moves, but its shape does not. A single company's order book is lumpier than a national aggregate and needs more room, not less. The textbook rule would set capacity at the 97.7th percentile of plausible demand at this asymmetry; the measured optimum sits well below that, because the rule treats every unit of shortfall as a total loss when in practice you can usually serve part of an order.*

A forecast is one number and demand is a distribution around it. Provision to the number and roughly a third of months land above it, and at 42 to 1 those months dominate everything else in the ledger. So the plan does not need to be right. It needs to be stated — a number you can be wrong about — with room sized from the asymmetry rather than from confidence, and a close each period reporting how far the market moved from the claim. Eight of those and the misses have a shape, and the shape names what the plan is not accounting for. Which is the loop, and the loop is the method.

Commit a claim. Set capacity. Order the work. Read the deviation. A strategy is a claim about a market, committed and held still so that what moves against it can be measured. The thesis is the reference frame; the market is what deviates from it, which is a different question from whether your forecast was any good and a far more useful one. Commit the market line and the share line, and multiplied they are your planned demand. Set capacity from the ratio. Order the work as one integer, no ties. Build, and close with rates, rules, formula and thesis version pinned so the numbers regenerate exactly. Then read the deviation twice — once against the market line and once against the share line — because the market moving is not the same failure as your share not arriving, and they have different owners. Adjust in sequence: the lock is never edited, refinements append as dated entries, and the record carries both what you believed and what you learned. *Basis. Advantage is not a field and not a claim. It is a share slope sustained above zero, observed across closes. One period proves nothing.*

Enter the cost of a lost sale, the cost of an idle unit and your planned demand. The tool returns the reserve the asymmetry implies and scores it against the housing-starts board. Save the result and request a copy. Open the capacity reserve calculator