Every number on this page is computed by the Volume II Chapter 8 notebook's own code, with the program office's and human resources' declarations; no step is random. Your notes stay on this page until you leave it; copy the memo before you go.
Human resources must propose a standing rule for approving hires against the digital program's engineering gap. The program office has declared what a gap costs, HR what speed costs, and approvals take a quarter to land. The board wants a rule it can approve and a certificate that no other rule does better. You support the team preparing the rule.
Use AXIOM to solve and certify Meridian's hiring rule, test it against Volume I's rules and the delay, and solve Chapter 6's capability problem on a grid.
Worked Example 8.1: the gap d (bench minus target, in engineers) is −100 today with nothing in the pipeline; it closes on its own at a = 4 ln(10/9) = 0.42 a year; approvals v, in engineers a year, add 60 percent of themselves to the bench a quarter later: d′(t) = −a d(t) + b v(t − ¼). A gap costs 0.0032d² a year, approving at a pace a premium of 0.000125v² a year; the board discounts at 10 percent a year, ρ = ln 1.1. Chapter 6's capability instance (Worked Example 6.1): $4 million a point up to 10.13 points a year, decay δ = 0.101 a year, and the build waits at $32 million a year until capability first reaches the floor of 55.
In the HJB equation explorer, try quadratic values −pd² for the hiring problem and map the residual and its error bound; find the p that makes it vanish, and the second root.
The residual is ρW − maxv H(d, v, W′(d)) for W = −pd², at 100 short ($ million a year). It vanishes where (b²/r)p² + (2a + ρ)p − q = 0 (Theorem 8.3).
$ million, from 100 short. ε: the largest residual on the gaps from 100 short to 100 over. The value is within ε/ρ of the best, and the greedy rule within 2ε/ρ.
Plot it against the approval rate at several gaps; find where the marginal premium equals the value of an approval.
The part of the Hamiltonian that depends on the approval rate: −rv² + m·b·v, with m = −2pd the price of an engineer of gap at the value's slope ($ million a year).
From Volume I's cautious gain, 2 a year: evaluate the rule's cost, then make the rule greedy for it. Costs from 100 short, $ million.
Generate the rule from the value and simulate it with and without the delay, beside Volume I's rules and the rule on the predicted gap; compare overshoot and cost.
Present values at 10 percent over an open horizon, $ million. Gain: approvals a quarter as a multiple of the gap. Peak: the largest surplus (engineers). At a year: bench minus target after a year. At once: the same rule's cost if approvals landed at once. Rules on today's gap are proportional and run continuously.
For the capability instance, check, comparing the maximum principle with the HJB equation, that the value's slope along the best path is Chapter 6's costate.
Line: Chapter 6's costate, m(t) = m(0)e(ρ + δ)t. Dots: the slope of the value at the plan's capability c(t). $ million a point.
For Chapter 6's capability problem, run the finite-difference solver, with its monotonicity audit, and policy iteration on grids from one point to a sixteenth; find the corners and measure the error.
A quadratic value whose Riccati coefficient makes the residual vanish, and a second root that fails the limit condition (Theorems 8.2 and 8.3); approvals where the marginal premium equals the value of an approval (Proposition 8.1); a rule on the predicted gap that never overshoots (Proposition 8.3); corners in the capability value at 43.45 and 55 (Proposition 8.2); and errors that halve with the grid (Proposition 8.4).
The approval rule and its cost, the comparison with Volume I's rules, the value of faster onboarding, the certificate, and the declarations the rule rests on.