The Geographic Retreat

What this pack runs.

A country's result is reported after an allocation of group cost made for a different purpose: so that the center's cost adds up across the units. That allocation is usually the largest item in the loss a country reports, and it stays behind with the countries that remain when this one goes, so a country reporting a loss of twenty million may be causing two million of it and leaving stops the two, not the twenty. By the time a country reaches an agenda it has usually been reclassified in the room from a market into a mistake, and the conversation becomes how soon to leave. That is not the comparison. This pack turns it into a subtraction and an annuity. What the country earns on its own, after the costs it causes and before the costs it is charged, across the years counted and discounted, is what keeping it is worth from here. Against it: the loss it stops, plus the part of the allocation that would actually leave, less what the rest of the group loses, across the same years, less the price of leaving once, which is severance, leases, contracts, write-downs, the closing of an entity, less what a buyer would pay. Both figures are usually negative and the decision is which negative is smaller. What settles it is how much of the reported loss actually leaves, so the whole question reduces to the break-even country loss, how much more a country can lose a year on its own before leaving it is cheaper than keeping it. Nine sourced retreats hold the frame: three were paid to leave, three wrote off a stated figure, one left at no significant gain or loss, one came back, and one sold the division that paid and did not survive the decade. It stops on two halts: no one has costed the country on its own, and no one has priced what leaving would cost.

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.