Every number on this page is computed by the Chapter 15 notebook's own code, on the book's 100,000 paths (seed 8) and fresh paths for the certificate (seed 26209). Your notes stay on this page until you leave it; copy the memo before you go.
Meridian's board wants every major transformation decision brought to it as a problem it can audit: decisions with owners, objectives with declared weights, commitments with prices, and the formulation's promise stated. You prepare the first such pack, on the digital program's risk levers.
Use AXIOM to build, audit, and solve a GEOP instance, read its prices and its frontier, and compare formulations.
Worked Example 15.1: the digital program of Chapter 8 and its two levers, a committed standby line (the CFO's, at fees of 3 percent of the line over the program) and a phased build (the transformation office's, at $8 million a quarter of delay); the board's 5 percent appetite, the 4.5× covenant and the capability floor of 55; Chapter 8's 100,000 paths (seed 8).
Assemble the digital program's instance from the declared inventories (the line and the phasing as decisions, the levers' costs as the objective, the appetite, the covenant, and the capability floor as commitments) and run the well-posedness checks.
Click a component to add it to the instance; click again to take it out.
The nine components of Definition 15.1; an instance is well posed when every decision has an owner and a domain, the objective and every commitment are declared, and each package can be evaluated (Definition 15.1; Theorem 15.1(i)); audits come before optimization (Proposition 15.4).
Sliders move in steps of $5 million and a quarter of a quarter; the chips load exact packages. Below the green curve the odds exceed the board's 5 percent; the dashed line joins the packages that cost the same as yours.
Each path's critical depth is the deepest draw it survives; a package breaches on a path exactly when its depth exceeds that path's critical depth, so the odds can be read without re-simulating.
Run the consistency and viability audits. Cap the line at $100 million and the delay at two quarters, then tighten the appetite below 3 percent; read the minimal infeasible set that the audit reports.
The least-cost package meeting every active commitment, lines in $5 million steps. When none exists, the audit drops one commitment at a time and reports the smallest set that cannot hold together (Theorem 15.2(iii)). Below: the odds of a breach by quarter K at the capped corner, against the appetite: the emptying date.
The deepest draw plus the buffer for one 90th-percentile closure net of the line, against the $270 million pool above the floor.
The share of paths below the floor in each quarter (bars) against the share below it in some quarter (the commitment): a quarter-by-quarter reading certifies nothing about the path (Theorem 14.5(iv)).
In the preloaded allocation instance (one budget, three capital stocks) watch the unconstrained first-order point demand a negative investment, then find the optimum at the corner where the nonnegativity constraint binds, and read its multipliers.
Chapter 14's stability budget, h r* + 2(e² + t²) ≤ r*²: its price divides the budget among the offices as the company would (Theorem 15.6(ii); Online Appendix 15.B.3). A twitchy optimum, u* = 1/(1 − a): near a = 1 the answer swings with the calibration (Proposition 15.5; Online Appendix 15.B.1).
Trace the risk–cost frontier of the digital program by floors and by weight sweeps; compare the two, and read the exchange rate at the board's appetite.
Floors: the least cost meeting each appetite from 1 to 24.5 percent, every quarter point. Weights: the package minimizing cost + w × odds, for w from 0.05 to 10 a point. Dotted: no line, however large, brings the odds below 8.4 percent.
Run the certainty-equivalent, robust, and stochastic formulations side by side; then test a rule that slows the build when cash runs low against the committed delay.
Liquidity over the program's 12 quarters under the formulation's package: the mean path, the median and the 5th percentile of the 100,000 paths (seed 8), against the $350 million floor. Fresh: the same package on fresh paths (seed 26209).
Each rule is the cheapest found on the book's paths that still meets 5 percent: a line, a committed delay, and, at the review, a cut in the draw still ahead for the paths whose cash is lowest. Perfect foresight, delays chosen path by path knowing each path's future, bounds what any information could save (Theorem 15.4(iii)).
Chapter 14's runway after its $30 million repair: $150 million above the floor, a burn of $50 million a quarter, cut by at most $10 million a quarter; each quarter the plan is re-made over a one-quarter window.
Feasibility comes before optimality (Theorem 15.2); everything that binds has a price, and slack commitments have none (Theorem 15.3); the frontier's slope is the multiplier (Theorem 15.5(iv)); and the formulation's promise changes the answer (Proposition 15.7).
The declared instance with its owners; the binding commitments and their prices; the frontier, with the chosen point and its exchange rate; the formulation's promise; and the certificate that the numbers carry.