The Consent Decree

What this pack runs.

A consent decree, a consent order, a deferred prosecution agreement, a settlement with conditions: the money is agreed, and so is a change in how the company runs, for a stated term, checked by someone the order names. The money is paid once. The change is paid for every year of the term, and every year a finding adds to it. But the order is negotiated by lawyers against a fine and read by the business against a headline, and no accounting line carries the cost of an order, so by the time the plan for running under it reaches an agenda the order has usually become a fine with paperwork attached, and the proposal on the table is the minimum that satisfies the letter. That is not the comparison. This pack turns it into a subtraction and the years. What running under the order costs a year in full, less what the minimum costs, across the term and discounted, is what the minimum saves. Against it: the fine once, plus every year a finding adds, at the full cost of running under the order, because a checker who has found you short will insist on the full version for every added year. What settles it is how many years a finding adds, so the whole question reduces to the break-even extension, how many years a finding would have to add to the term before the minimum stops paying, read beside what the order says a finding adds. Nine sourced undertakings hold the frame: one reached the mechanism and stood, three were broken or extended, one was never checked until a regulator did, one touched the decoration, one was never accepted, one was never agreed, and one names its checkers and its term. It stops on two halts: no one has costed running under the order in full, and no one has read what a finding triggers.

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.