Every number on this page is computed by the Volume II Chapter 7 notebook's own code, on the engineering office's record; the simulation of the rule uses the lab's seed 26207. Your notes stay on this page until you leave it; copy the memo before you go.
The board has accepted Chapter 6's advice to stop defending 58. Capability does not move as planned, the transformation office has declared what a shortfall costs, and the chief human resources officer must propose a hiring rule with a margin over the floor. You support the team preparing the rule the board will approve.
Use AXIOM to solve Meridian's capability problem as a dynamic program, read its value function and its rule, compare the rule with plans, and test the answer against the declarations it rests on.
Worked Example 7.1: c′ = 0.975c + h − w, hiring h at $4 million a point and at most 2.5 a quarter. The record: in half of all quarters capability ends where planned, in four quarters in ten three-quarters of a point higher or lower, in one in ten a point and a half higher or lower. A point short for a quarter costs $1.6 million; β = 1.1−1/4 a quarter, over an open horizon.
In the Bellman equation explorer, pick a capability level and plot the value of each hiring decision against the values of the positions it leads to; find the best.
Left: −4h − 1.6(55 − c)+ + β Σ p(w) V(0.975c + h − w) for each hiring decision h. Right: the best hiring, capability by capability: a target rule (Theorem 7.2).
In recursive planning, solve a twelve-quarter horizon by backward induction and read the rule table.
Run value iteration with its error envelope and stopping bound, then policy iteration from Volume I's rule; compare the work each needs.
In the value-function view, plot the value and its slope, and value the engineering team as a lump.
Present values at 10 percent over an open horizon, $ million. Margin: planned capability above 55. Short: long-run share of quarters that start below 55.
Run fitted value iteration on a coarse grid and with a smooth fit, and check each against its error bound; then shorten the period and compare the deterministic solution with Chapter 6's in continuous time.
The same recursion on Chapter 6's deterministic problem: the build waits until capability first reaches 55 and must then stay at or above it. On a calendar the build can start only at the beginning of a period.
Values that converge from any start (Theorem 7.1); prices that fall as capability grows (Proposition 7.1); a target rule with a margin (Theorem 7.2); error bounds that hold, however loosely (Theorem 7.3); and a recursion that approaches Chapter 6's solution as the period shrinks (Section 7.5).
The rule, its margin and its cost, the comparison with Volume I's budget and with Chapter 6's answer, and the declarations the rule depends on.