The Single Source

What this pack runs.

Buying a critical input from one supplier is chosen for price, quality, or simplicity, and it is usually right on all three. What makes it a decision later is the day the first source stops: a fire, a flood, an insolvency, an allocation, a refusal to sell. The line runs on stock for some days and then goes dark until a replacement is qualified from a standing start. By the time a second source reaches an agenda, the meeting is arguing about the discount it would forfeit and the surcharge it would pay, which is the known half of the decision and the smaller half. A second source is insurance, and insurance is priced against what it insures against. This pack turns that into a subtraction and one division. What a second source costs a year is the forfeited discount, the surcharge, the running cost of keeping two sources, and the qualification spread across the years. What a day without the input costs is the margin on what cannot ship, the idle line, the expediting, and the customers who would not wait, and in most companies it has never been priced because the margin sits in an account by product rather than by input. The first divided by the second is the break-even outage, how many days a year the first source can stop before the second one pays for itself, and read against the supplier's own log rather than its reputation it becomes a number of years between stoppages that a buyer can check in an afternoon. The model sweeps how likely the first source is to stop, from never to one year in two, because it is the one figure no meeting will settle. It stops on six conditions: no spend on the input, no price for a day without it, no count of how long a replacement takes, no length for a stoppage, no ramp for a second source, and no figure for how often the first source stops.

The casebooks these cases come from.

The pack reads nine sourced decisions through the framework. They are drawn from these casebooks, each one a set of verified records on a single decision type.

A verified record has passed a mechanical claim-versus-source check, a second model's read, and human approval. The count is what exists today, not a target.