Skip to content

Vol. II, Ch. 12

DCT Laboratory — Volume II, Chapter 12

Multi-Objective Enterprise Optimization

This laboratory demonstrates

Seed 26212 · Companion to Chapter 12 and AXIOM Lab 2.12 (module AXIOM-12) · Mirrored in DCT_V2_Ch12_Lab.xlsx

This laboratory reproduces Worked Example 12.1, Whose Objective Runs the Automation Program?, with the book's declarations, and works the five steps of AXIOM Lab 2.12: enumerate, price the frontier, bound the reserve, choose, and add uncertainty. The program is Chapter 5's automation program; the new declarations are the three owners' objectives. The model is deterministic: for each schedule of rounds and each reserve, the best payouts solve a small linear program, solved here by SciPy's HiGHS as in the book, and no number below depends on a random draw. The lab's own seed, 26212, draws only Panel 6's illustrations of how frontiers are computed (random weights and an evolutionary search). Lab step 5 reads Chapter 9's cost–chance frontier as the book prints it (Online Appendix Example 10.B.2) instead of re-simulating its 100,000 paths. Every Meridian number below is the book's (Table 12.3, Figures 12.1–12.3, Online Appendix Example 12.B.2); the final cell checks them. Money is in $ million.

The declarations (Worked Example 12.1, Step 1; Online Appendix Remark 12.B.1)

Chapter 5's program, unchanged. The state is year-end cash qkq_k and efficiency eke_k, from q0=620q_0 = 620 and e0=71e_0 = 71. A round of automation in year kk, zk∈{0,1}z_k \in \{0, 1\}, costs $120 million and halves the remaining gap to 85 points, ek=ek−1+zk(85−ek−1)/2e_k = e_{k-1} + z_k(85 - e_{k-1})/2; the run-rate savings are Sk=10(ek−71)S_k = 10(e_k - 71), $10 million a year for each point above 71; payouts pk≥0p_k \ge 0 are free, and cash moves as qk=qk−1−120zk+Sk−pkq_k = q_{k-1} - 120z_k + S_k - p_k. The board's commitments: cash stays above $350 million at every year end, reaches $470 million before a round, and ends year 6 at $470 million. Payouts are discounted at the 10 percent hurdle rate, β=1/1.1\beta = 1/1.1; cash earns nothing.

The three owners' objectives (new declarations). The chief financial officer judges a plan by the present value of its payouts, f1=∑k=16βkpkf_1 = \sum_{k=1}^{6}\beta^{k}p_k; the chief operating officer by the run-rate savings it leaves in year 6, f2=S6f_2 = S_6, which depends only on the number of rounds nn, S6=140−140/2nS_6 = 140 - 140/2^{n}; the treasury by the reserve bb by which every cash commitment is kept: qk≥350+bq_k \ge 350 + b in every year, qk−1≥470+bq_{k-1} \ge 470 + b before a round in year kk, and q6≥470+bq_6 \ge 470 + b. Chapter 5's value is one weighted sum of the first two, f1+wf2f_1 + w f_2 with w=β6/0.10=5.645w = \beta^{6}/0.10 = 5.645: a dollar a year of year-6 savings, continued forever, priced in dollars of payouts today. It ignores the third.

The vector problem (Definition 12.1). Maximize (f1,f2,f3)(f_1, f_2, f_3) over the schedules of whole rounds, the payouts and the reserve. For a schedule and a reserve, the best payouts solve a linear program in p1,…,p6p_1, \dots, p_6 with at most twelve constraints, solved, as in the book, by SciPy's HiGHS; its multipliers price the reserve. The cell solves all 64×151=9,66464 \times 151 = 9{,}664 programs, every schedule at every reserve on a $1 million grid from 0 to $150 million, and checks them against Chapter 5's closed form: pay out, every year, everything above the war chest, the least year-end cash from which the rest of the schedule keeps every commitment, now with every commitment raised by bb: R6=470+bR_6 = 470 + b and Rk−1=max⁡{Rk+120zk−Sk, (470+b)zk, 350+b}R_{k-1} = \max\{R_k + 120z_k - S_k,\ (470 + b)z_k,\ 350 + b\}, with no floor term for R0R_0; the schedule keeps every commitment when 620≥R0620 \ge R_0.

Python Code
import itertools
import numpy as np
from scipy.optimize import linprog
import matplotlib.pyplot as plt
plt.rcParams.update({"figure.dpi": 110, "axes.spines.top": False, "axes.spines.right": False})
SEED = 26212        # the lab's seed: it draws only Panel 6's random weights and evolutionary search

K, R = 6, 0.10                                          # six years; the board's hurdle rate
BETA = 1.0 / (1.0 + R)
Q0, E0, CEIL, VAL = 620.0, 71.0, 85.0, 10.0            # today's cash and efficiency; the ceiling; $10m a year a point
COST, FLOOR, PRE, Q_END = 120.0, 350.0, 470.0, 470.0   # a round; the floor; cash before a round; cash at the end
W_BOARD = BETA ** K / R                                # Chapter 5's price of a dollar a year of year-6 savings
SCHEDULES = list(itertools.product([0, 1], repeat=K))  # the 64 schedules (z_1, ..., z_6), in binary order
NR = np.array([sum(z) for z in SCHEDULES])             # rounds in each schedule
NB = 151                                               # reserves 0, 1, ..., 150 ($1m grid)

def years(z):
    # the years of a schedule's rounds as the book prints them: three or more in a row as a range
    ys = [k + 1 for k in range(K) if z[k]]
    out, i = [], 0
    while i < len(ys):
        j = i
        while j + 1 < len(ys) and ys[j + 1] == ys[j] + 1:
            j += 1
        out += [f"{ys[i]}–{ys[j]}"] if j - i >= 2 else [str(y) for y in ys[i:j + 1]]
        i = j + 1
    return ", ".join(out) or "none"

def savings(z):
    # a round halves the gap to 85 points; $10m a year for each point above 71
    e, out = E0, []
    for zk in z:
        if zk:
            e = e + (CEIL - e) / 2
        out.append(VAL * (e - E0))
    return np.array(out)

S6 = np.array([savings(z)[-1] for z in SCHEDULES])     # the COO's objective: run-rate savings in year 6

def sv(x):
    # savings as the book prints them: up to four decimals
    return f"{x:.4f}".rstrip("0").rstrip(".")

def payouts_lp(z, b=0.0):
    # The best payouts for schedule z with every commitment kept with b to spare (SciPy/HiGHS):
    # max sum_k beta^k p_k  s.t.  q_k >= 350 + b,  q_(k-1) >= 470 + b before a round in year k,  q_6 >= 470 + b,
    # where q_k is the cash with no payouts less p_1 + ... + p_k, and p >= 0.
    # Returns (present value, payouts, year-end cash, price of the reserve), or None if no payouts keep them.
    S = savings(z)
    base = Q0 + np.cumsum(S - COST * np.array(z, float))
    if z[0] and Q0 < PRE + b - 1e-9:                    # a year-1 round needs 470 + b today
        return None
    rows, rhs = [], []
    for k in range(K):
        row = np.zeros(K)
        row[:k + 1] = 1.0
        rows.append(row); rhs.append(base[k] - (FLOOR + b))             # the floor at the end of year k + 1
        if k + 1 < K and z[k + 1]:
            rows.append(row.copy()); rhs.append(base[k] - (PRE + b))    # the cash a round in year k + 2 needs
    rows.append(np.ones(K)); rhs.append(base[-1] - (Q_END + b))          # the end condition
    res = linprog(-BETA ** np.arange(1, K + 1), A_ub=np.array(rows), b_ub=np.array(rhs),
                  bounds=[(0, None)] * K, method="highs")
    if res.status != 0:
        return None
    # a dollar more of reserve lowers every right-hand side by a dollar: the payouts lose the multipliers' sum
    return float(-res.fun), res.x, base - np.cumsum(res.x), float(-np.sum(res.ineqlin.marginals))

def war_chest(z, b=0.0):
    # Chapter 5's war chest with every commitment raised by b: R_6 = 470 + b,
    # R_(k-1) = max(R_k + 120 z_k - S_k, (470 + b) z_k, 350 + b), no floor term for R_0; pay out everything above it
    S = savings(z)
    Rw = np.empty(K + 1)
    Rw[K] = Q_END + b
    for k in range(K, 0, -1):
        need = max(Rw[k] + COST * z[k - 1] - S[k - 1], (PRE + b) * z[k - 1])
        Rw[k - 1] = max(need, FLOOR + b) if k > 1 else need
    if Q0 < Rw[0] - 1e-9:
        return None
    start = np.concatenate([[Q0], Rw[1:K]])            # each later year opens with the war chest left
    p = start - COST * np.array(z, float) + S - Rw[1:]
    return float(BETA ** np.arange(1, K + 1) @ p), p, Rw

LP = {}                                                # (schedule, reserve) -> the payout program's answer
for i, z in enumerate(SCHEDULES):
    for b in range(NB):
        out = payouts_lp(z, float(b))
        if out is not None:
            LP[i, b] = out
WC_SAME, WC_GAP = True, 0.0
for i, z in enumerate(SCHEDULES):
    for b in range(NB):
        wc = war_chest(z, float(b))
        if (wc is None) != ((i, b) not in LP):
            WC_SAME = False
        elif wc is not None:
            WC_GAP = max(WC_GAP, abs(wc[0] - LP[i, b][0]), float(np.max(np.abs(wc[1] - LP[i, b][1]))))
N_PROGRAMS = len(SCHEDULES) * NB
print(f"{len(SCHEDULES)} schedules x {NB} reserves = {N_PROGRAMS:,} payout programs; {len(LP):,} keep every commitment")
print(f"the war chest against the linear programs: the same feasibility every time: {WC_SAME}; largest difference "
      f"in a payout or a value {WC_GAP:.1e}")
print(f"Chapter 5's price of savings: beta^6 / 0.10 = {W_BOARD:.4f} dollars of payouts today per dollar a year")
print("year-6 savings by number of rounds, 140 - 140 / 2^n: " + ", ".join(f"{n}: {sv(140 - 140 / 2 ** n)}" for n in range(K + 1)))
64 schedules x 151 reserves = 9,664 payout programs; 7,824 keep every commitment
the war chest against the linear programs: the same feasibility every time: True; largest difference in a payout or a value 5.7e-14
Chapter 5's price of savings: beta^6 / 0.10 = 5.6447 dollars of payouts today per dollar a year
year-6 savings by number of rounds, 140 - 140 / 2^n: 0: 0, 1: 70, 2: 105, 3: 122.5, 4: 131.25, 5: 135.625, 6: 137.8125

Panel 1 — Enumerate (Theorem 12.1; Worked Example 12.1, Step 2; Figure 12.1(a): all 64 schedules keep the commitments, 59 are dominated; five are efficient, from two rounds worth $384.17m with $105m a year to six worth $106.59m with $137.81m; the ideal point is out of reach)

AXIOM Lab 2.12, step 1. With no reserve, solve the payouts for every schedule of rounds, place each schedule by its payouts and its savings, and mark the dominated ones: a schedule is dominated when another pays at least as much and saves at least as much, and more of one (Definition 12.1). By Theorem 12.1(i), every dominated schedule is dominated by an efficient one, so the board can set it aside whoever proposes it. Savings depend only on the number of rounds, so only the best schedule with each number of rounds can be efficient. Theorem 12.1(ii) is how a board already speaks: "maximize payouts, provided savings reach $122.5 million a year and the reserve $50 million" bounds every objective but one, and its answer is a point of the frontier.

Python Code
P0 = np.array([LP[i, 0][0] for i in range(64)])        # each schedule's best payouts, present value, no reserve
N_FEAS0 = sum((i, 0) in LP for i in range(64))

def dominated(F, tol=1e-9):
    # which rows of F (objectives, all maximized) another row dominates: as good in every one, better in one
    F = np.asarray(F, float)
    ge = np.all(F[:, None, :] >= F[None, :, :] - tol, axis=2)      # ge[j, i]: row j at least as good as row i
    gt = np.any(F[:, None, :] > F[None, :, :] + tol, axis=2)
    return np.any(ge & gt, axis=0)

def dominates(a, c, tol=1e-9):
    return bool(np.all(a >= c - tol) and np.any(a > c + tol))

F2 = np.c_[P0, S6]
DOM2 = dominated(F2)
EFF2 = sorted(np.nonzero(~DOM2)[0].tolist(), key=lambda i: NR[i])
BEST_N = {n: max((i for i in range(64) if NR[i] == n), key=lambda i: (P0[i], -i)) for n in range(K + 1)}
print(f"{N_FEAS0} of 64 schedules keep every commitment with no reserve: {DOM2.sum()} dominated, {len(EFF2)} efficient")
print("\nthe best schedule with each number of rounds:")
print("rounds  years              payouts (PV)  savings   efficient  payouts, years 1-6")
for n in range(K + 1):
    i = BEST_N[n]
    print(f"{n:6d}  {years(SCHEDULES[i]):17s}  {P0[i]:12.2f}  {S6[i]:8.4f}  {'yes' if not DOM2[i] else 'no':9s}  "
          + ", ".join(f"{x:g}" for x in np.round(LP[i, 0][1], 4)))
I3 = BEST_N[3]
print("three rounds, year-end cash: " + ", ".join(f"{x:g}" for x in np.round(LP[I3, 0][2], 4))
      + f"; {LP[I3, 0][1].sum():g} paid in all")
print(f"one round ({P0[BEST_N[1]]:.2f}, {sv(S6[BEST_N[1]])} a year) is dominated by two ({P0[BEST_N[2]]:.2f}, "
      f"{sv(S6[BEST_N[2]])} a year): on that, the CFO and the COO agree")
COVERED = all(any(dominates(F2[e], F2[j]) for e in EFF2) for j in np.nonzero(DOM2)[0])
print(f"Theorem 12.1(i): each of the {DOM2.sum()} dominated schedules is dominated by an efficient one: {COVERED}")
IDEAL2 = F2.max(axis=0)
IDEAL2_HIT = bool(np.any(np.all(F2 >= IDEAL2 - 1e-9, axis=1)))
print(f"the ideal point, payouts {IDEAL2[0]:.2f} with savings {IDEAL2[1]:.4f} a year, is attained by some schedule: "
      f"{IDEAL2_HIT} (Theorem 12.1(iii): every efficient plan gives up something)")

def best_with(bound_S, b=0):
    # Theorem 12.1(ii): the most payouts with savings of at least bound_S and a reserve of b
    ok = [i for i in range(64) if S6[i] >= bound_S - 1e-9 and (i, b) in LP]
    return max(ok, key=lambda i: (LP[i, b][0], -i))

EPS_A, EPS_B = best_with(122.5), best_with(122.5, 50)
EPS_UNIQUE = sum(abs(LP[i, 0][0] - LP[EPS_A, 0][0]) < 1e-9 for i in range(64) if S6[i] >= 122.5 - 1e-9) == 1
print(f"maximize payouts with savings of at least 122.5 a year: rounds in years {years(SCHEDULES[EPS_A])}, "
      f"{LP[EPS_A, 0][0]:.2f}, the only answer: {EPS_UNIQUE}, hence efficient")
print(f"... and a reserve of $50 million: rounds in years {years(SCHEDULES[EPS_B])}, {LP[EPS_B, 50][0]:.2f}")

fig, ax = plt.subplots(figsize=(7.6, 3.6))
dom = np.nonzero(DOM2)[0]
ax.scatter(S6[dom], P0[dom], s=14, color="#9E9E9E", zorder=2, label=f"dominated ({len(dom)})")
ax.plot(S6[EFF2], P0[EFF2], "o-", color="#1F4E79", lw=1.8, ms=5, zorder=3, label=f"efficient ({len(EFF2)})")
for i in EFF2:
    ax.text(S6[i] + 2.0, P0[i] + 5.0, f"{NR[i]} rounds" if i == EFF2[0] else f"{NR[i]}", fontsize=8, color="#1F4E79")
ax.plot([IDEAL2[1]], [IDEAL2[0]], marker="*", ms=13, color="#E65100", ls="none", zorder=4)
ax.text(IDEAL2[1] + 3.0, IDEAL2[0], "ideal:\nout of reach", fontsize=8, color="#E65100", va="center")
ax.annotate("one round: dominated by two", xy=(S6[BEST_N[1]], P0[BEST_N[1]]), xytext=(8, 310), fontsize=8,
            arrowprops=dict(arrowstyle="-", lw=0.6, color="#333333"))
ax.set(xlim=(-5, 165), ylim=(0, 430), xlabel="run-rate savings in year 6 ($ million a year)",
       ylabel="payouts, present value ($ million)", title="Five of the 64 schedules are efficient (seed 26212)")
ax.legend(frameon=False, fontsize=8, loc="lower left"); plt.tight_layout(); plt.show()
64 of 64 schedules keep every commitment with no reserve: 59 dominated, 5 efficient

the best schedule with each number of rounds:
rounds  years              payouts (PV)  savings   efficient  payouts, years 1-6
     0  none                     136.36    0.0000  no         150, 0, 0, 0, 0, 0
     1  1                        370.67   70.0000  no         220, 70, 70, 70, 20, 0
     2  1, 2                     384.17  105.0000  yes        100, 105, 105, 105, 90, 0
     3  1–3                      330.45  122.5000  yes        85, 0, 122.5, 122.5, 122.5, 2.5
     4  1–3, 6                   260.88  131.2500  yes        85, 0, 122.5, 122.5, 2.5, 11.25
     5  1–3, 5, 6                186.82  135.6250  yes        85, 0, 122.5, 2.5, 11.25, 15.625
     6  1–6                      106.59  137.8125  yes        85, 0, 2.5, 11.25, 15.625, 17.8125
three rounds, year-end cash: 485, 470, 350, 350, 350, 470; 455 paid in all
one round (370.67, 70 a year) is dominated by two (384.17, 105 a year): on that, the CFO and the COO agree
Theorem 12.1(i): each of the 59 dominated schedules is dominated by an efficient one: True
the ideal point, payouts 384.17 with savings 137.8125 a year, is attained by some schedule: False (Theorem 12.1(iii): every efficient plan gives up something)
maximize payouts with savings of at least 122.5 a year: rounds in years 1–3, 330.45, the only answer: True, hence efficient
... and a reserve of $50 million: rounds in years 1–3, 285.00
Five of the 64 schedules are efficient (seed 26212)

Panel 2 — Price the frontier (Theorem 12.2(i), (iv); Worked Example 12.1, Step 3; Figure 12.1(b): between neighbouring efficient plans a dollar a year of savings costs $3.07, $7.95, $16.93 and $36.68 of payouts; Chapter 5's price, $5.645, picks three rounds, $1,021.93m)

AXIOM Lab 2.12, step 2. Between two neighbouring efficient plans, the price of savings is the payouts given up per dollar a year of savings gained. A weighted sum f1+wf2f_1 + w f_2 prices a dollar a year of savings at ww dollars of payouts, and picks the plan whose neighbouring prices bracket ww; with positive weights its answer is efficient (Theorem 12.2(i)). Chapter 5's board value is the weighted sum with w=β6/0.10w = \beta^{6}/0.10, the year-6 savings continued forever. Its answer, three rounds, is a choice of price, not a fact about the program: a board that valued savings more would run more rounds.

Python Code
RATES = []
for a, c in zip(EFF2[:-1], EFF2[1:]):
    RATES.append(dict(frm=int(NR[a]), to=int(NR[c]), dP=float(P0[a] - P0[c]), dS=float(S6[c] - S6[a]),
                      rate=float((P0[a] - P0[c]) / (S6[c] - S6[a]))))
print("between neighbouring efficient plans, payouts given up for savings gained:")
for r in RATES:
    print(f"  {r['frm']} -> {r['to']} rounds: {r['dP']:6.2f} for {r['dS']:7.4f} a year = {r['rate']:6.2f} per dollar a year")
edges = [0.0] + [r["rate"] for r in RATES] + [np.inf]
print("a weighted sum, payouts + w x savings, picks")
for k, i in enumerate(EFF2):
    print(f"  {NR[i]} rounds for w from {edges[k]:.2f} to {edges[k + 1]:.2f}")
INCREASING = all(a["rate"] < c["rate"] for a, c in zip(RATES[:-1], RATES[1:]))
print(f"the prices rise along the frontier: {INCREASING}; with no reserve, every efficient plan wins some weighted sum")
BOARD_V = {int(NR[i]): float(P0[i] + W_BOARD * S6[i]) for i in EFF2}
BOARD_I = max(range(64), key=lambda i: P0[i] + W_BOARD * S6[i])
print(f"\nChapter 5's price, w = {W_BOARD:.4f}: board values " + ", ".join(f"{n} rounds {v:,.2f}" for n, v in BOARD_V.items()))
print(f"the board's weighted sum picks rounds in years {years(SCHEDULES[BOARD_I])}: {P0[BOARD_I]:.2f} of payouts + "
      f"{W_BOARD:.4f} x {S6[BOARD_I]:g} of savings = {P0[BOARD_I] + W_BOARD * S6[BOARD_I]:,.2f}")
print(f"a price below {RATES[0]['rate']:.2f} would choose two rounds, and one above {RATES[1]['rate']:.2f} four")

ws_ = np.exp(np.linspace(np.log(1.0), np.log(100.0), 801))
pick = [NR[max(range(64), key=lambda i: P0[i] + w * S6[i])] for w in ws_]
fig, ax = plt.subplots(figsize=(7.6, 3.6))
ax.step(ws_, pick, where="post", color="#1F4E79", lw=2)
ax.set_xscale("log")
for r in RATES:
    ax.axvline(r["rate"], color="#999999", lw=0.8, ls=":")
    ax.text(r["rate"] * 1.06, r["frm"] + 0.5, f"{r['rate']:.2f}", fontsize=8, color="#616161", va="center")
ax.plot([W_BOARD], [3], marker="o", ms=7, mfc="white", mec="#E65100", mew=1.6, ls="none", zorder=5)
ax.annotate(f"Chapter 5's board: {W_BOARD:.3f}\nthree rounds, {BOARD_V[3]:,.2f}", xy=(W_BOARD, 3), xytext=(1.25, 4.6),
            fontsize=8, color="#E65100", arrowprops=dict(arrowstyle="-", lw=0.7, color="#E65100"))
ax.set_xticks([1, 2, 5, 10, 20, 50, 100], ["1", "2", "5", "10", "20", "50", "100"]); ax.minorticks_off()
ax.set(xlim=(1, 100), ylim=(1.5, 6.5), yticks=[2, 3, 4, 5, 6],
       xlabel="price of a dollar a year of savings ($ of payouts, log scale)", ylabel="rounds in the plan it picks",
       title="The board's price of savings, 5.645, picks three rounds (seed 26212)")
plt.tight_layout(); plt.show()
between neighbouring efficient plans, payouts given up for savings gained:
  2 -> 3 rounds:  53.72 for 17.5000 a year =   3.07 per dollar a year
  3 -> 4 rounds:  69.57 for  8.7500 a year =   7.95 per dollar a year
  4 -> 5 rounds:  74.06 for  4.3750 a year =  16.93 per dollar a year
  5 -> 6 rounds:  80.23 for  2.1875 a year =  36.68 per dollar a year
a weighted sum, payouts + w x savings, picks
  2 rounds for w from 0.00 to 3.07
  3 rounds for w from 3.07 to 7.95
  4 rounds for w from 7.95 to 16.93
  5 rounds for w from 16.93 to 36.68
  6 rounds for w from 36.68 to inf
the prices rise along the frontier: True; with no reserve, every efficient plan wins some weighted sum

Chapter 5's price, w = 5.6447: board values 2 rounds 976.87, 3 rounds 1,021.93, 4 rounds 1,001.75, 5 rounds 952.39, 6 rounds 884.51
the board's weighted sum picks rounds in years 1–3: 330.45 of payouts + 5.6447 x 122.5 of savings = 1,021.93
a price below 3.07 would choose two rounds, and one above 7.95 four
The board's price of savings, 5.645, picks three rounds (seed 26212)

Panel 3 — Bound the reserve (Theorem 12.2(iv); Worked Example 12.1, Step 4; Figure 12.2: $0.909 of payouts per dollar of reserve, for every schedule and reserve; above $85m and $100m the rounds move later, at extra costs of $4.02m and $24.38m)

AXIOM Lab 2.12, step 3. Raise the reserve, read its multiplier, and find where the schedule of rounds must change. The multiplier of the bound on f3f_3 is the sum of the payout program's multipliers, because a dollar more of reserve takes a dollar from the slack of every commitment. The war chest shows why it is the same everywhere: every term of the recursion rises by bb, so the war chest rises by bb in every year; the payouts after year 1 do not change and the year-1 payout falls by bb. The reserve is a year-1 payout held back until the end, worth β=$0.909\beta = \$0.909 of payouts per dollar, whatever the schedule. A schedule keeps the reserve only while today's cash covers its war chest, 620≥R0+b620 \ge R_0 + b. After a round in year 1, the cash before a year-2 round is at most 620−120+70=570620 - 120 + 70 = 570, so rounds in years 1 and 2 need b≤100b \le 100; after rounds in years 1 and 2, the cash before a year-3 round is at most 570−120+105=555570 - 120 + 105 = 555, so rounds in years 1–3 need b≤85b \le 85. Beyond these, the best plan with the same number of rounds moves a round later, and the payouts fall by more than the price: the trade-off surface of Theorem 12.2(iv) jumps.

Python Code
PRICES = np.array([v[3] for v in LP.values()])
print(f"the reserve's price from the multipliers of all {len(LP):,} feasible programs: {PRICES.min():.4f} to "
      f"{PRICES.max():.4f} of payouts per dollar (beta = {BETA:.4f})")
Z3 = (1, 1, 1, 0, 0, 0)
for b in (0.0, 50.0):
    v, p, Rw = war_chest(Z3, b)
    print(f"rounds in years 1-3, reserve {b:3.0f}: war chest R_0..R_6 " + ", ".join(f"{x:g}" for x in Rw)
          + "; payouts " + ", ".join(f"{x:g}" for x in np.round(p, 4)) + f"; value {v:.2f}")
BMAX = np.array([max(b for b in range(NB) if (i, b) in LP) for i in range(64)])
print("the largest reserve a schedule keeps: " + "; ".join(f"{v}: {int(np.sum(BMAX == v))} schedules" for v in sorted(set(BMAX))))
S_ONE, S_TWO = savings((1, 0, 0, 0, 0, 0))[0], savings((1, 1, 0, 0, 0, 0))[1]
CASH2 = Q0 - COST + S_ONE
CASH3 = CASH2 - COST + S_TWO
print(f"after a year-1 round, the cash before a year-2 round is at most {Q0:.0f} - {COST:.0f} + {S_ONE:.0f} = {CASH2:.0f}: "
      f"rounds in years 1 and 2 need b <= {CASH2 - PRE:.0f}")
print(f"after rounds in years 1 and 2, the cash before a year-3 round is at most {CASH2:.0f} - {COST:.0f} + {S_TWO:.0f} = "
      f"{CASH3:.0f}: rounds in years 1-3 need b <= {CASH3 - PRE:.0f}")
print(f"no plan keeps more than {BMAX.max()}: before the first round, or at the end without one, the cash is at most "
      f"{Q0:.0f} against {PRE:.0f} plus the reserve")

def best_at(n, b):
    # the best schedule with n rounds at reserve b, the first in binary order on ties (as the book's engine)
    best = None
    for i in range(64):
        if NR[i] == n and (i, b) in LP and (best is None or LP[i, b][0] > LP[best, b][0] + 1e-12):
            best = i
    return best

ENV3 = [best_at(3, b) for b in range(NB)]
PHI = [max(LP[i, b][0] for i in range(64) if S6[i] >= 122.5 - 1e-9 and (i, b) in LP) for b in range(NB)]
SW3 = [b for b in range(1, NB) if ENV3[b] != ENV3[b - 1]]
PHI_IS_3 = all(abs(PHI[b] - LP[ENV3[b], b][0]) < 1e-9 for b in range(NB))
print("\nthe trade-off surface: the most payouts with savings of at least 122.5 a year and a reserve of at least b")
print(f"(always the best three-round plan, as more rounds pay less: {PHI_IS_3})")
runs = [0] + SW3 + [NB]
for lo, hi in zip(runs[:-1], runs[1:]):
    print(f"  reserves {lo:3d} to {hi - 1:3d}: rounds in years {years(SCHEDULES[ENV3[lo]])}")
B_SHOW = (0, 25, 50, 85, 86, 90, 100, 101, 105, 120, 150)
print("  reserve  " + "".join(f"{b:8d}" for b in B_SHOW))
print("  payouts  " + "".join(f"{PHI[b]:8.2f}" for b in B_SHOW))
SLOPES = [PHI[b - 1] - PHI[b] for b in range(1, NB) if b not in SW3]
DROPS = {b: PHI[b - 1] - PHI[b] - BETA for b in SW3}
print(f"where it is smooth, the surface falls by {min(SLOPES):.4f} to {max(SLOPES):.4f} a dollar, the multiplier; where the "
      f"rounds move it falls by more: " + "; ".join(f"at {b}, {PHI[b - 1] - PHI[b]:.2f} = {BETA:.2f} + {d:.2f}"
                                                     for b, d in DROPS.items()))
the reserve's price from the multipliers of all 7,824 feasible programs: 0.9091 to 0.9091 of payouts per dollar (beta = 0.9091)
rounds in years 1-3, reserve   0: war chest R_0..R_6 535, 485, 470, 350, 350, 350, 470; payouts 85, 0, 122.5, 122.5, 122.5, 2.5; value 330.45
rounds in years 1-3, reserve  50: war chest R_0..R_6 585, 535, 520, 400, 400, 400, 520; payouts 35, 0, 122.5, 122.5, 122.5, 2.5; value 285.00
the largest reserve a schedule keeps: 85: 16 schedules; 100: 16 schedules; 150: 32 schedules
after a year-1 round, the cash before a year-2 round is at most 620 - 120 + 70 = 570: rounds in years 1 and 2 need b <= 100
after rounds in years 1 and 2, the cash before a year-3 round is at most 570 - 120 + 105 = 555: rounds in years 1-3 need b <= 85
no plan keeps more than 150: before the first round, or at the end without one, the cash is at most 620 against 470 plus the reserve

the trade-off surface: the most payouts with savings of at least 122.5 a year and a reserve of at least b
(always the best three-round plan, as more rounds pay less: True)
  reserves   0 to  85: rounds in years 1–3
  reserves  86 to 100: rounds in years 1, 2, 4
  reserves 101 to 150: rounds in years 1, 3, 4
  reserve         0      25      50      85      86      90     100     101     105     120     150
  payouts    330.45  307.72  285.00  253.18  248.25  244.61  235.52  210.23  206.60  192.96  165.69
where it is smooth, the surface falls by 0.9091 to 0.9091 a dollar, the multiplier; where the rounds move it falls by more: at 86, 4.93 = 0.91 + 4.02; at 101, 25.29 = 0.91 + 24.38

Panel 4 — Three objectives and the dents (Theorem 12.2(iii); Worked Example 12.1, Step 4; Figure 12.2; Online Appendix Example 12.B.2: 992 plans, 740 efficient, 453 supported and 287 in dents, all with reserves from $86m to $149m; ideal point (384.17, 137.81, 150), nadir (22.06, 70, 0))

The three-objective frontier on the $1 million grid of reserves (Online Appendix Remark 12.B.1). For each number of rounds and each reserve only the best schedule can be efficient, since the savings depend only on the number of rounds: 992 plans. A plan is efficient when no other is at least as good for all three owners and better for one. An efficient plan is supported when some weights, each at least 0.001 of their sum, with the objectives scaled by their ranges over the efficient plans, make it the best weighted sum: one small linear program in the weights for each plan. The cell also keeps a certificate of each answer that can be checked without a solver, and the workbook checks all 740: the weights, for a supported plan; for an unsupported one, an average of efficient plans that is better in every objective, which by Theorem 12.2(iii) rules out every choice of nonnegative weights, not all zero. The ideal point holds each objective's best value; the nadir, each one's worst over the efficient plans.

Python Code
PTS = [(n, b, best_at(n, b)) for n in range(K + 1) for b in range(NB)]
PTS = [p for p in PTS if p[2] is not None]                   # (rounds, reserve, best schedule)
F3 = np.array([[LP[i, b][0], S6[i], b] for n, b, i in PTS])
EFF3 = ~dominated(F3)
E3 = [p for p, e in zip(PTS, EFF3) if e]                     # the efficient plans, by rounds, then reserve
FE = F3[EFF3]
IDEAL, NADIR = FE.max(axis=0), FE.min(axis=0)
SPAN = IDEAL - NADIR
GE = FE / SPAN                                                # the objectives scaled by their ranges
M = len(E3)
SUP = np.zeros(M, bool)
WEIGHTS, AVERAGES, MARGINS = {}, {}, {}
for k in range(M):
    # supported: weights w >= 0.001, summing to 1, with w . (G_j - G_k) <= 0 for every efficient plan j
    res = linprog(np.zeros(3), A_ub=GE - GE[k], b_ub=np.zeros(M), A_eq=np.ones((1, 3)), b_eq=[1.0],
                  bounds=[(1e-3, None)] * 3, method="highs")
    if res.status == 0:
        SUP[k], WEIGHTS[k] = True, res.x
        continue
    # unsupported: the average of efficient plans that beats it in every objective by the widest margin t
    res = linprog(np.r_[np.zeros(M), -1.0], A_ub=np.c_[-(GE - GE[k]).T, np.ones(3)], b_ub=np.zeros(3),
                  A_eq=np.r_[np.ones(M), 0.0][None, :], b_eq=[1.0], bounds=[(0, None)] * M + [(None, 1.0)],
                  method="highs-ds")
    lam = res.x[:M]
    AVERAGES[k] = [(int(j), float(lam[j])) for j in np.nonzero(lam > 1e-12)[0]]
    MARGINS[k] = float(res.x[-1])
CHECK_W = max(float(np.max((GE - GE[k]) @ w)) for k, w in WEIGHTS.items())
CHECK_A = min(float(np.min(sum(l * GE[j] for j, l in AVERAGES[k]) - GE[k])) for k in AVERAGES)
print(f"{len(PTS)} plans are the best for their number of rounds and reserve; {M} are efficient: {SUP.sum()} supported, "
      f"{M - SUP.sum()} in dents")
print(f"certificates: at its weights, no efficient plan beats a supported one (by at most {max(CHECK_W, 0.0):.0e}); each "
      f"plan in a dent is beaten in every objective by an average of at most {max(len(a) for a in AVERAGES.values())} "
      f"efficient plans, by at least {CHECK_A:.1e} of each range")
print("so no weights reach the dents even with zero components allowed: the answer is the same with weights merely "
      "nonnegative (Online Appendix Remark 12.B.1)")
print("\nrounds  efficient  supported  in dents  reserves of the plans in dents")
DENTS = {}
for n in range(K + 1):
    ks = [k for k in range(M) if E3[k][0] == n]
    if ks:
        DENTS[n] = [E3[k][1] for k in ks if not SUP[k]]
        print(f"{n:6d}  {len(ks):9d}  {sum(SUP[k] for k in ks):9d}  {len(DENTS[n]):8d}  "
              + (f"{min(DENTS[n])} to {max(DENTS[n])}" if DENTS[n] else "none"))
ALL_DENT_B = [E3[k][1] for k in range(M) if not SUP[k]]
print(f"every plan in a dent keeps a reserve from {min(ALL_DENT_B)} to {max(ALL_DENT_B)}")
KN = int(np.argmin(FE[:, 0]))
print(f"\nideal point: payouts {IDEAL[0]:.2f}, savings {IDEAL[1]:.4f}, reserve {IDEAL[2]:.0f}")
print(f"nadir point: payouts {NADIR[0]:.2f} (rounds in years {years(SCHEDULES[E3[KN][2]])} with a reserve of "
      f"{E3[KN][1]}), savings {NADIR[1]:g}, reserve {NADIR[2]:g}")
KX = next(k for k in range(M) if E3[k][:2] == (3, 92))
avg = sum(l * FE[j] for j, l in AVERAGES[KX])
print(f"\na plan in a dent: rounds in years {years(SCHEDULES[E3[KX][2]])} with a reserve of 92: payouts {FE[KX][0]:.2f}, "
      f"savings {FE[KX][1]:g}")
print("  is beaten in every objective by " + " + ".join(f"{l:.3f} x (years {years(SCHEDULES[E3[j][2]])}, reserve {E3[j][1]})"
                                                   for j, l in AVERAGES[KX])
      + f": payouts {avg[0]:.2f}, savings {avg[1]:.2f}, reserve {avg[2]:.2f}")

COLR = {1: "#9E9E9E", 2: "#1F4E79", 3: "#E65100", 4: "#2E7D32", 5: "#6A1B9A", 6: "#B71C1C"}
fig, ax = plt.subplots(figsize=(7.6, 3.6))
for n in range(1, K + 1):
    ks = sorted((k for k in range(M) if E3[k][0] == n), key=lambda k: E3[k][1])
    runs, cur = [], [ks[0]]
    for a, c in zip(ks[:-1], ks[1:]):              # runs of one schedule over consecutive reserves
        if E3[c][2] == E3[a][2] and E3[c][1] == E3[a][1] + 1:
            cur.append(c)
        else:
            runs.append(cur); cur = [c]
    runs.append(cur)
    for run in runs:
        bs, ps = np.array([E3[k][1] for k in run]), FE[run, 0]
        solid = [bool(SUP[run[a]] and SUP[run[a + 1]]) for a in range(len(run) - 1)]
        a = 0
        while a < len(solid):                      # one line for each stretch of the same style
            c = a
            while c + 1 < len(solid) and solid[c + 1] == solid[a]:
                c += 1
            ax.plot(bs[a:c + 2], ps[a:c + 2], color=COLR[n], lw=1.6, ls="-" if solid[a] else (0, (3, 2)))
            a = c + 1
    ax.plot([], [], color=COLR[n], lw=1.6, label=f"{n} round{'s' if n > 1 else ''}")
ax.plot([], [], color="#616161", lw=1.2, ls=(0, (3, 2)), label="dashed: no weights pick it")
for x in (85, 100):
    ax.axvline(x, color="#999999", lw=0.8, ls=":")
ax.text(84, 395, "rounds in years 1–3\nout of reach", fontsize=7.5, color="#616161", ha="right", va="top")
ax.text(101, 395, "rounds in years 1, 2\nout of reach", fontsize=7.5, color="#616161", ha="left", va="top")
ax.plot([92], [FE[KX][0]], marker="o", ms=6, color="black", ls="none", zorder=5)
ax.annotate("years 1, 2, 4, a \\$92m\nreserve: in a dent", xy=(92, FE[KX][0]), xytext=(106, 318), fontsize=8,
            arrowprops=dict(arrowstyle="-", lw=0.6, color="#333333"))
ax.set(xlim=(0, 150), ylim=(0, 400), xlabel="reserve on every cash commitment ($ million)",
       ylabel="payouts, present value ($ million)", title="Above $85m the rounds move later, into dents (seed 26212)")
ax.legend(frameon=False, fontsize=7.5, loc="upper left", bbox_to_anchor=(1.01, 1.0))
plt.tight_layout(); plt.show()
992 plans are the best for their number of rounds and reserve; 740 are efficient: 453 supported, 287 in dents
certificates: at its weights, no efficient plan beats a supported one (by at most 1e-14); each plan in a dent is beaten in every objective by an average of at most 3 efficient plans, by at least 8.9e-05 of each range
so no weights reach the dents even with zero components allowed: the answer is the same with weights merely nonnegative (Online Appendix Remark 12.B.1)

rounds  efficient  supported  in dents  reserves of the plans in dents
     1         50          1        49  101 to 149
     2        151        102        49  101 to 149
     3        151         88        63  86 to 149
     4        151         88        63  86 to 149
     5        151         88        63  86 to 149
     6         86         86         0  none
every plan in a dent keeps a reserve from 86 to 149

ideal point: payouts 384.17, savings 137.8125, reserve 150
nadir point: payouts 22.06 (rounds in years 1, 3–6 with a reserve of 150), savings 70, reserve 0

a plan in a dent: rounds in years 1, 2, 4 with a reserve of 92: payouts 242.80, savings 122.5
  is beaten in every objective by 0.498 x (years 1–3, reserve 85) + 0.487 x (years 1, 2, 4, reserve 100) + 0.015 x (years 1–3, 6, reserve 85): payouts 243.52, savings 122.63, reserve 92.30
Above $85m the rounds move later, into dents (seed 26212)

Panel 5 — Choose (Theorem 12.3; Proposition 12.1; Worked Example 12.1, Step 5; Table 12.3; Figure 12.3: equal weights choose rounds in years 1, 2 and 4 with a $92m reserve, $242.80m, a plan in a dent; the CFO weighted twice, two rounds with a $96m reserve, $296.90m; goals that can all be met: $300.00m first, $307.72m after a second stage)

AXIOM Lab 2.12, step 4. The board states aspirations instead of prices: the ideal point as the reference, each owner's shortfall measured as a share of the range from the nadir to the ideal, and weights for their urgency. The augmented Chebyshev program picks the efficient plan with the least largest weighted shortfall, plus ρ=10−4\rho = 10^{-4} times their sum; every solution is efficient (Definition 12.2; Theorem 12.3(ii)). By Theorem 12.3(iii), with a reference point above the ideal in every objective, every efficient plan is the choice for some positive weights, the plans in dents included: the cell checks all 740. Each owner's own best optimizes that owner's objective first and the others after (Proposition 12.1(iii)). Goal programming minimizes the weighted shortfalls below the board's goals, as shares of the goals: payouts of at least $300 million, savings of $122.5 million a year and a $25 million reserve. These goals can all be met, so every plan that meets them is a solution, dominated ones included (Proposition 12.1(ii)). The book's engine returns the first such schedule it meets, in binary order of the schedules, and pays just the goal, since the shortfalls do not reward paying more; the second stage maximizes payouts among the plans that meet every goal.

Python Code
def chebyshev(w, rho=1e-4, ref=None):
    # the augmented Chebyshev choice: least largest weighted shortfall from the reference point, shortfalls as shares of
    # the ideal-nadir range, plus rho times their sum; the first on ties (by rounds, then reserve)
    ref = IDEAL if ref is None else ref
    d = np.asarray(w, float) * (ref - FE) / SPAN
    k = int(np.argmin(d.max(axis=1) + rho * d.sum(axis=1)))
    return k, d[k]

def plan(k):
    n, b, i = E3[k]
    return dict(k=k, n=n, b=b, i=i, z=SCHEDULES[i], P=float(FE[k][0]), S=float(FE[k][1]), p=LP[i, b][1],
                J=float(FE[k][0] + W_BOARD * FE[k][1]), supported=bool(SUP[k]))

def paid(p):
    return ", ".join(f"{x:g}" for x in np.round(p, 4))

CHOICE = {}
for key, w in (("equal", (1, 1, 1)), ("cfo2", (2, 1, 1))):
    k, d = chebyshev(w)
    CHOICE[key] = c = dict(plan(k), d=d, w=w)
    print(f"weights {w}: rounds in years {years(c['z'])}, reserve {c['b']}, payouts {c['P']:.2f} ({paid(c['p'])}), "
          f"savings {sv(c['S'])}; weighted shortfalls " + ", ".join(f"{x:.3f}" for x in d)
          + ("; supported" if c["supported"] else "; in a dent: no weights reach it"))
REF = IDEAL + 0.01 * SPAN
REACHED = 0
for k in range(M):
    w = SPAN / (REF - FE[k])                       # w_j = 1 / (r_j - f_j(x*)), in units of the ranges
    s = np.max(w * (REF - FE) / SPAN, axis=1)
    win = np.nonzero(s <= s.min() + 1e-12)[0]
    REACHED += bool(s[k] <= s.min() + 1e-12 and np.all(np.abs(FE[win] - FE[k]) < 1e-9))
print(f"Theorem 12.3(iii), the reference 1% of each range above the ideal: {REACHED} of {M} efficient plans are the only "
      f"choice for their own weights, the {M - SUP.sum()} in dents among them")

OWN = {"CFO": plan(max(range(M), key=lambda k: (round(FE[k][0], 9), FE[k][1], FE[k][2]))),
       "COO": plan(max(range(M), key=lambda k: (FE[k][1], round(FE[k][0], 9), FE[k][2]))),
       "treasury": plan(max(range(M), key=lambda k: (FE[k][2], round(FE[k][0], 9), FE[k][1])))}
print()
for nm, c in OWN.items():
    print(f"the {nm}'s own best: rounds in years {years(c['z'])}, reserve {c['b']}, payouts {c['P']:.2f} ({paid(c['p'])}), "
          f"savings {sv(c['S'])}; board value {c['J']:,.2f}")
BOARD3 = plan(max(range(M), key=lambda k: FE[k][0] + W_BOARD * FE[k][1]))
print(f"the board's weighted sum on the three-objective frontier: rounds in years {years(BOARD3['z'])}, reserve "
      f"{BOARD3['b']}, {BOARD3['J']:,.2f}")

GOALS = (300.0, 122.5, 25)
FIRST = None
for i in range(64):                                # the book's engine meets the schedules in binary order
    if (i, GOALS[2]) in LP:
        dev = max(0.0, GOALS[0] - LP[i, GOALS[2]][0]) / GOALS[0] + max(0.0, GOALS[1] - S6[i]) / GOALS[1]
        if FIRST is None or dev < FIRST[1] - 1e-12:
            FIRST = (i, dev)
G1 = FIRST[0]
G1_AVAIL = LP[G1, GOALS[2]][0]
G1_PAID = min(G1_AVAIL, GOALS[0])                  # the shortfalls do not reward paying more than the goal
MEET = [i for i in range(64) if (i, GOALS[2]) in LP and S6[i] >= GOALS[1] - 1e-9 and LP[i, GOALS[2]][0] >= GOALS[0] - 1e-9]
G2 = max(MEET, key=lambda i: (LP[i, GOALS[2]][0], -i))
G2_P = LP[G2, GOALS[2]][0]
G1_DOMINATED = dominates(np.array([G2_P, S6[G2], GOALS[2]]), np.array([G1_PAID, S6[G1], GOALS[2]]))
print(f"\ngoals {GOALS}: {len(MEET)} schedules meet them all: " + "; ".join(f"years {years(SCHEDULES[i])}, "
                                                                       f"{LP[i, GOALS[2]][0]:.2f}" for i in MEET))
print(f"the solver's first answer: rounds in years {years(SCHEDULES[G1])}, paying {G1_PAID:.2f} of the {G1_AVAIL:.2f} "
      f"available; dominated: {G1_DOMINATED}")
print(f"the second stage, the most payouts among the plans meeting every goal: rounds in years {years(SCHEDULES[G2])}, "
      f"{G2_P:.2f}; the first answer leaves {G2_P - G1_PAID:.2f} unpaid")

T123 = [("CFO's best", OWN["CFO"]["z"], OWN["CFO"]["b"], OWN["CFO"]["P"], OWN["CFO"]["S"]),
        ("COO's best", OWN["COO"]["z"], OWN["COO"]["b"], OWN["COO"]["P"], OWN["COO"]["S"]),
        ("Treasury's best", OWN["treasury"]["z"], OWN["treasury"]["b"], OWN["treasury"]["P"], OWN["treasury"]["S"]),
        ("Board value (Chapter 5)", SCHEDULES[BOARD_I], 0, P0[BOARD_I], S6[BOARD_I]),
        ("Chebyshev, equal weights", CHOICE["equal"]["z"], CHOICE["equal"]["b"], CHOICE["equal"]["P"], CHOICE["equal"]["S"]),
        ("Chebyshev, CFO twice", CHOICE["cfo2"]["z"], CHOICE["cfo2"]["b"], CHOICE["cfo2"]["P"], CHOICE["cfo2"]["S"]),
        ("Goals, first answer", SCHEDULES[G1], GOALS[2], G1_PAID, S6[G1]),
        ("Goals, second stage", SCHEDULES[G2], GOALS[2], G2_P, S6[G2])]
TABLE = [dict(name=nm, years=years(z), reserve=int(b), P=float(P), S=float(S), J=float(P + W_BOARD * S))
         for nm, z, b, P, S in T123]
print("\nTable 12.3: plans by owner and by method of choice (board value: payouts + 5.645 x savings)")
print(f"{'plan':26s} {'rounds (years)':15s} {'reserve':>7s} {'payouts':>8s} {'savings':>8s} {'board value':>12s}")
for r in TABLE:
    print(f"{r['name']:26s} {r['years']:15s} {r['reserve']:7d} {r['P']:8.2f} {r['S']:8.2f} {r['J']:12,.2f}")
J_RANGE = (min(r["J"] for r in TABLE), max(r["J"] for r in TABLE))
print(f"Figure 12.3's board-value axis runs from {J_RANGE[0]:,.2f} to {J_RANGE[1]:,.2f}")
weights (1, 1, 1): rounds in years 1, 2, 4, reserve 92, payouts 242.80 (8, 90, 0, 122.5, 122.5, 2.5), savings 122.5; weighted shortfalls 0.390, 0.226, 0.387; in a dent: no weights reach it
weights (2, 1, 1): rounds in years 1, 2, reserve 96, payouts 296.90 (4, 105, 105, 105, 90, 0), savings 105; weighted shortfalls 0.482, 0.484, 0.360; supported
Theorem 12.3(iii), the reference 1% of each range above the ideal: 740 of 740 efficient plans are the only choice for their own weights, the 287 in dents among them

the CFO's own best: rounds in years 1, 2, reserve 0, payouts 384.17 (100, 105, 105, 105, 90, 0), savings 105; board value 976.87
the COO's own best: rounds in years 1–6, reserve 0, payouts 106.59 (85, 0, 2.5, 11.25, 15.625, 17.8125), savings 137.8125; board value 884.51
the treasury's own best: rounds in years 1, reserve 150, payouts 234.31 (70, 70, 70, 70, 20, 0), savings 70; board value 629.44
the board's weighted sum on the three-objective frontier: rounds in years 1–3, reserve 0, 1,021.93

goals (300.0, 122.5, 25): 3 schedules meet them all: years 1, 2, 5, 300.05; years 1, 2, 4, 303.71; years 1–3, 307.72
the solver's first answer: rounds in years 1, 2, 5, paying 300.00 of the 300.05 available; dominated: True
the second stage, the most payouts among the plans meeting every goal: rounds in years 1–3, 307.72; the first answer leaves 7.72 unpaid

Table 12.3: plans by owner and by method of choice (board value: payouts + 5.645 x savings)
plan                       rounds (years)  reserve  payouts  savings  board value
CFO's best                 1, 2                  0   384.17   105.00       976.87
COO's best                 1–6                   0   106.59   137.81       884.51
Treasury's best            1                   150   234.31    70.00       629.44
Board value (Chapter 5)    1–3                   0   330.45   122.50     1,021.93
Chebyshev, equal weights   1, 2, 4              92   242.80   122.50       934.28
Chebyshev, CFO twice       1, 2                 96   296.90   105.00       889.60
Goals, first answer        1, 2, 5              25   300.00   122.50       991.48
Goals, second stage        1–3                  25   307.72   122.50       999.21
Figure 12.3's board-value axis runs from 629.44 to 1,021.93

Panel 6 — Compute frontiers (Proposition 12.2; Table 12.2; Section 12.4: 9,664 payout programs and one support program for each of the 740 efficient plans; weights find only supported plans; an archive that keeps every undominated plan ends with the efficient set)

Section 12.4's methods on Meridian's instance. (i) Bracketing: with two objectives, bounds on savings box the frontier, every efficient plan between two consecutive bounds lying between their two answers. (ii) Hypervolume: the area a set of schedules dominates above a reference point below the nadir, here (0, 0); the sets of one, two and three schedules that maximize it hold only efficient schedules, checked by trying every such set of the 64. (iii) Weights are not enough: 10,000 random weights, each at least 0.001 of their sum, on the three scaled objectives, drawn with the lab's seed 26212; every plan they return is supported, and none is in a dent. (iv) Evolutionary search: in each generation each of the 992 plans is proposed with chance 0.02 (the same stream, seed 26212), and an unbounded archive keeps every plan that no plan of the archive and the proposals dominates; by Proposition 12.2(iv) it equals the efficient set after finitely many generations. An archive bounded below 740 plans never could, the correction the book makes to the previous edition. These draws are the lab's own illustration; none of the book's numbers depends on them.

Python Code
def phi2(eps):
    # the most payouts with savings of at least eps, no reserve
    return max(P0[i] for i in range(64) if S6[i] >= eps - 1e-9)

BOUNDS = [0.0, 60.0, 100.0, 110.0, 120.0, 125.0, 130.0, 134.0, 136.0, 137.8125]
BOXED = all(phi2(c) - 1e-9 <= P0[i] <= phi2(a) + 1e-9 for a, c in zip(BOUNDS[:-1], BOUNDS[1:]) for i in EFF2
            if a <= S6[i] <= c)
print("(i) bounds on savings and the most payouts they allow: " + ", ".join(f"{sv(e)}: {phi2(e):.2f}" for e in BOUNDS))
print(f"    every efficient plan between two consecutive bounds lies between their answers: {BOXED}")

def hv2(F, ref=(0.0, 0.0)):
    # the area that points (payouts, savings) dominate above the reference point
    F = np.asarray(F, float).reshape(-1, 2) - np.asarray(ref)
    F = F[(F[:, 0] > 0) & (F[:, 1] > 0)]
    if len(F) == 0:
        return 0.0
    F = F[np.argsort(-F[:, 1])]
    top = np.maximum.accumulate(F[:, 0])
    return float(np.sum((F[:, 1] - np.r_[F[1:, 1], 0.0]) * top))

HV_E = hv2(F2[EFF2])
HV_BEST, HV_OK = {}, True
for m_ in (1, 2, 3):
    best = max(itertools.combinations(range(64), m_), key=lambda s: hv2(F2[list(s)]))
    HV_BEST[m_] = (best, hv2(F2[list(best)]))
    HV_OK &= set(best) <= set(EFF2)
    print(f"(ii) the {m_} schedule{'s' if m_ > 1 else ''} with the largest hypervolume: "
          + "; ".join(f"years {years(SCHEDULES[i])}" for i in best) + f": {HV_BEST[m_][1]:,.2f}")
HV_ADD = all(hv2(F2[[e for e in EFF2 if e != x]]) < HV_E - 1e-9 for x in EFF2) and \
    all(abs(hv2(F2[EFF2 + [j]]) - HV_E) < 1e-9 for j in np.nonzero(DOM2)[0])
print(f"    the five efficient schedules: {HV_E:,.2f}; dropping any one lowers it, adding a dominated one does not raise "
      f"it: {HV_ADD}")

rng = np.random.default_rng(SEED)
WTS = 1e-3 + (1.0 - 3e-3) * rng.dirichlet(np.ones(3), 10_000)
G3 = F3 / SPAN
FOUND = sorted(set(int(np.argmax(G3 @ w)) for w in WTS))
POS = {p: k for k, p in enumerate(E3)}
FOUND_K = [POS.get(PTS[j]) for j in FOUND]
SWEEP_OK = all(k is not None and SUP[k] for k in FOUND_K)
print(f"(iii) {len(WTS):,} random weights (seed {SEED}) return {len(FOUND)} different plans, all efficient and supported: "
      f"{SWEEP_OK}; none of the {M - SUP.sum()} in dents")
print("     most supported plans lie inside stretches where the payouts fall at the reserve's price: a weighted sum is "
      "indifferent along a stretch only at one exact ratio of its weights, so random weights pick its ends")

def beats(Fa, Fb, tol=1e-9):
    # [a, b]: row a of Fa dominates row b of Fb
    ge = np.all(Fa[:, None, :] >= Fb[None, :, :] - tol, axis=2)
    gt = np.any(Fa[:, None, :] > Fb[None, :, :] + tol, axis=2)
    return ge & gt

EFFSET = set(np.nonzero(EFF3)[0].tolist())
archive, T_EVO, HIST = np.array([], int), None, {}
for t in range(1, 5001):
    prop = np.nonzero(rng.random(len(PTS)) < 0.02)[0]
    new = np.setdiff1d(prop, archive)
    pool = np.r_[archive, new]
    keep_old = ~beats(F3[new], F3[archive]).any(axis=0)
    keep_new = ~beats(F3[pool], F3[new]).any(axis=0)
    archive = np.r_[archive[keep_old], new[keep_new]]
    HIST[t] = (len(archive), len(EFFSET & set(archive.tolist())))
    if set(archive.tolist()) == EFFSET:
        T_EVO = t
        break
print(f"(iv) the unbounded archive (seed {SEED}), plans kept and how many of them efficient: "
      + "; ".join(f"generation {t}: {HIST[t][0]}, {HIST[t][1]}" for t in (1, 10, 50, 100, 200) if t in HIST))
print(f"     it equals the efficient set, all {len(EFFSET)} plans and nothing else, from generation {T_EVO}")
print(f"the book's computation: {N_PROGRAMS:,} payout programs ({len(LP):,} feasible) and {M} support programs")
(i) bounds on savings and the most payouts they allow: 0: 384.17, 60: 384.17, 100: 384.17, 110: 330.45, 120: 330.45, 125: 260.88, 130: 260.88, 134: 186.82, 136: 106.59, 137.8125: 106.59
    every efficient plan between two consecutive bounds lies between their answers: True
(ii) the 1 schedule with the largest hypervolume: years 1–3: 40,480.37
(ii) the 2 schedules with the largest hypervolume: years 1, 2; years 1–3, 6: 47,186.32
(ii) the 3 schedules with the largest hypervolume: years 1, 2; years 1–3; years 1–3, 5, 6: 48,573.14
    the five efficient schedules: 49,454.33; dropping any one lowers it, adding a dominated one does not raise it: True
(iii) 10,000 random weights (seed 26212) return 18 different plans, all efficient and supported: True; none of the 287 in dents
     most supported plans lie inside stretches where the payouts fall at the reserve's price: a weighted sum is indifferent along a stretch only at one exact ratio of its weights, so random weights pick its ends
(iv) the unbounded archive (seed 26212), plans kept and how many of them efficient: generation 1: 12, 11; generation 10: 115, 115; generation 50: 472, 471; generation 100: 653, 653; generation 200: 731, 731
     it equals the efficient set, all 740 plans and nothing else, from generation 307
the book's computation: 9,664 payout programs (7,824 feasible) and 740 support programs

Panel 7 — Add uncertainty (Proposition 12.3; Section 12.5; Figure 10.3; Online Appendix Example 10.B.2: Chapter 9's cheapest committed packages cost $12.35m at 4%, $10.44m at 5% and $8.82m at 6%, a point of appetite worth about $1.6–1.9m; Chapter 11's records' family moves the 5% point to $24.00m)

AXIOM Lab 2.12, step 5. Chapter 9's appetite, at most a 5 percent chance of a breach of the $350 million floor in some quarter, bounds one objective of a vector problem: the cost of a package and its chance of a breach. The least cost C(α)C(\alpha) at each appetite α\alpha is the trade-off surface; it never rises with the appetite, and its slope is the appetite's price (Proposition 12.3(i)). Chapter 9's packages are entered here as the book prints them (Online Appendix Example 10.B.2: a standby line and a committed delay for each appetite, on Chapter 9's 100,000 paths, seed 8); they are declared values, not re-simulated. Declaring a family of laws never lowers C(α)C(\alpha) (part (ii)): Chapter 11's records' family keeps 5 percent at $24.00 million (Worked Example 11.1). Part (iii): with conditional value at risk the frontier is convex and weighted sums reach all of it, but the chance itself can have dents, as the small example of Online Appendix 12.B shows.

Python Code
# Chapter 9's cheapest committed packages as Online Appendix Example 10.B.2 prints them (declared, not re-simulated):
# appetite (%), standby line ($m), committed delay (quarters), cost ($m)
CH9 = [(1.0, 190, 2.174, 23.09), (2.0, 170, 1.595, 17.86), (3.0, 180, 1.161, 14.68), (4.0, 165, 0.925, 12.35),
       (5.0, 170, 0.668, 10.44), (6.0, 160, 0.502, 8.82), (8.0, 160, 0.182, 6.25), (10.0, 135, 0.015, 4.17),
       (12.5, 100, 0.0, 3.00), (15.0, 75, 0.0, 2.25)]
FEE, VQ = 0.03, 8.0              # Chapter 9's levers: line fees of 3% over the program; $8m a quarter of delay
FAMILY_5 = 24.00                 # Chapter 11's records' family keeps 5% at $24.00m (Worked Example 11.1)
A9 = np.array([r[0] for r in CH9])
C9 = np.array([r[3] for r in CH9])
LEVERS_GAP = max(abs(FEE * l + VQ * d - c) for _, l, d, c in CH9)
print(f"each declared cost is 3% of the line plus $8m a quarter of delay, within {LEVERS_GAP:.3f} (delays printed to three decimals)")
SL = -np.diff(C9) / np.diff(A9)                  # $m of cost per point of appetite, between neighbours
print("appetite (%)    " + "".join(f"{a:7g}" for a in A9))
print("least cost      " + "".join(f"{c:7.2f}" for c in C9))
print("price a point      " + "".join(f"{s:7.2f}" for s in SL))
P45, P56 = C9[3] - C9[4], C9[4] - C9[5]
print(f"near the board's 5%: a point of appetite is worth {P56:.2f} (5% to 6%) to {P45:.2f} (4% to 5%): about "
      f"{P56:.1f} to {P45:.1f} million")
MONO, FALLING = bool(np.all(np.diff(C9) < 0)), bool(np.all(np.diff(SL) < 0))
print(f"Proposition 12.3(i): the least cost never rises with the appetite: {MONO}; on these ten points its price falls "
      f"as the appetite grows: {FALLING}")
print(f"Proposition 12.3(ii): the records' family keeps 5% at {FAMILY_5:.2f}, against {C9[4]:.2f} under one law: the "
      f"family moves the point up by {FAMILY_5 - C9[4]:.2f}, Chapter 11's price of ambiguity")

def chance(z):
    # Online Appendix 12.B's example: a line of size z in [0, 2] costs z; the need is 1 or 2 with chances 0.3 and 0.7
    return 1.0 if z < 1 else (0.7 if z < 2 else 0.0)

AVG = (0.55 * 0.0 + 0.45 * 2.0, 0.55 * chance(0.0) + 0.45 * chance(2.0))
print(f"Proposition 12.3(iii): the line z = 1 (cost 1, chance {chance(1.0):g}) is efficient, yet 0.55 x (0, 1) + "
      f"0.45 x (2, 0) = ({AVG[0]:.1f}, {AVG[1]:.2f}) is better in both: a dent, which no weighted sum reaches")

fig, ax = plt.subplots(figsize=(7.6, 3.6))
ax.plot(A9, C9, "o-", color="#1F4E79", lw=1.8, ms=4, label="Chapter 9's frontier: the cheapest committed package")
ax.plot([5.0], [FAMILY_5], marker="s", ms=7, color="#E65100", ls="none", label="Chapter 11's records' family, at 5%")
ax.annotate("", xy=(5.0, FAMILY_5 - 0.6), xytext=(5.0, C9[4] + 0.6),
            arrowprops=dict(arrowstyle="->", color="#E65100", lw=0.9))
ax.text(5.3, (FAMILY_5 + C9[4]) / 2, f"family: +\\${FAMILY_5 - C9[4]:.2f}m", fontsize=8, color="#E65100", va="center")
ax.text(4.4, 7.6, f"4% to 5%:\n\\${P45:.2f}m a point", fontsize=8, color="#1F4E79", ha="right", va="center")
ax.text(6.2, C9[5] + 1.1, f"5% to 6%: \\${P56:.2f}m a point", fontsize=8, color="#1F4E79", ha="left", va="center")
ax.axvline(5.0, color="#999999", lw=0.8, ls=":")
ax.set(xlim=(0, 16), ylim=(0, 26), xlabel="appetite: chance of a breach in some quarter (%)",
       ylabel="cost of the package ($ million)", title="Near 5%, a point of appetite is worth \\$1.6m to \\$1.9m (seed 26212)")
ax.legend(frameon=False, fontsize=8, loc="upper right"); plt.tight_layout(); plt.show()
each declared cost is 3% of the line plus $8m a quarter of delay, within 0.008 (delays printed to three decimals)
appetite (%)          1      2      3      4      5      6      8     10   12.5     15
least cost        23.09  17.86  14.68  12.35  10.44   8.82   6.25   4.17   3.00   2.25
price a point         5.23   3.18   2.33   1.91   1.62   1.29   1.04   0.47   0.30
near the board's 5%: a point of appetite is worth 1.62 (5% to 6%) to 1.91 (4% to 5%): about 1.6 to 1.9 million
Proposition 12.3(i): the least cost never rises with the appetite: True; on these ten points its price falls as the appetite grows: True
Proposition 12.3(ii): the records' family keeps 5% at 24.00, against 10.44 under one law: the family moves the point up by 13.56, Chapter 11's price of ambiguity
Proposition 12.3(iii): the line z = 1 (cost 1, chance 0.7) is efficient, yet 0.55 x (0, 1) + 0.45 x (2, 0) = (0.9, 0.55) is better in both: a dent, which no weighted sum reaches
Near 5%, a point of appetite is worth $1.6m to $1.9m (seed 26212)

Validation — the book's numbers, and agreement with DCT_V2_Ch12_Lab.xlsx

Each checkpoint is a number printed in Chapter 12 (Sections 12.1–12.5, Worked Example 12.1, Table 12.3, Figures 12.1–12.3, the LOS Guidance) or in Online Appendix Example 12.B.2, with a tolerance of half a unit in its last printed digit; a few check what a theorem or proposition promises (1 = holds). The five-round plan's savings, $135.625 million a year, print as 135.63 in Step 2 and pass at its last printed digit. Chapter 9's frontier is the book's declared values (Online Appendix Example 10.B.2), so its checkpoints check the prices the chapter reads from it. The workbook recomputes the checkpoints marked workbook with spreadsheet formulas and shows PASS against the same engine values on its Checks sheet.

Python Code
def bk(s):
    # a number as the book prints it: its value and half a unit in its last printed digit
    t = s.replace(",", "")
    dp = len(t.split(".")[1]) if "." in t else 0
    return float(t), 0.5 * 10.0 ** -dp + 1e-9

def c_(name, got, printed):
    v, tol = bk(printed)
    return (name, float(got), v, tol)

def flag(name, ok):
    return (name, float(bool(ok)), 1.0, 0.0)

def count(name, got, n):
    return (name, float(got), float(n), 0.0)

WORD = {0: "no rounds", 1: "one round", 2: "two rounds", 3: "three rounds", 4: "four rounds", 5: "five rounds", 6: "six rounds"}
OA_BEST = {1: ("370.67", "70", ("220", "70", "70", "70", "20", "0")), 2: ("384.17", "105", ("100", "105", "105", "105", "90", "0")),
           3: ("330.45", "122.5", ("85", "0", "122.5", "122.5", "122.5", "2.5")),
           4: ("260.88", "131.25", ("85", "0", "122.5", "122.5", "2.5", "11.25")),
           5: ("186.82", "135.625", ("85", "0", "122.5", "2.5", "11.25", "15.625")),
           6: ("106.59", "137.8125", ("85", "0", "2.5", "11.25", "15.625", "17.8125"))}
checks = [
    # the declarations and the payout programs
    count("payout programs, 64 schedules x 151 reserves (workbook)", N_PROGRAMS, 9664),
    count("payout programs that keep every commitment (workbook)", len(LP), 7824),
    flag("war chest = linear program: the same feasibility in all 9,664", WC_SAME),
    ("war chest = linear program: largest difference", WC_GAP, 0.0, 1e-9),
    flag("savings depend only on the rounds: 140 - 140/2^n (workbook)", np.allclose(S6, 140 - 140 / 2.0 ** NR)),
    c_("Chapter 5's price of savings, Section 12.2 (workbook)", W_BOARD, "5.645"),
    c_("Chapter 5's price of savings, Online Appendix (workbook)", W_BOARD, "5.6447"),
    # Step 2; Figure 12.1(a) (Panel 1)
    count("schedules keeping every commitment, no reserve (workbook)", N_FEAS0, 64),
    count("dominated schedules (workbook)", DOM2.sum(), 59),
    count("efficient schedules (workbook)", len(EFF2), 5),
    flag("the efficient plans are the best with two to six rounds", [NR[i] for i in EFF2] == [2, 3, 4, 5, 6]
         and all(BEST_N[n] in EFF2 for n in range(2, 7))),
    c_("no rounds: payouts (workbook)", P0[BEST_N[0]], "136.36"),
    c_("no rounds: payout in year 1 (workbook)", LP[BEST_N[0], 0][1][0], "150"),
]
for n, (pv, sv, pays) in OA_BEST.items():
    i = BEST_N[n]
    checks += [c_(f"{WORD[n]}: payouts (workbook)", P0[i], pv), c_(f"{WORD[n]}: savings (workbook)", S6[i], sv)]
    checks += [c_(f"{WORD[n]}: payout in year {k + 1} (workbook)", LP[i, 0][1][k], s) for k, s in enumerate(pays)]
checks += [c_(f"three rounds: cash at the end of year {k + 1} (workbook)", LP[I3, 0][2][k], s)
           for k, s in enumerate(("485", "470", "350", "350", "350", "470"))]
checks += [
    c_("five rounds: savings as Step 2 prints them", S6[BEST_N[5]], "135.63"),
    c_("six rounds: savings as Step 2 prints them", S6[BEST_N[6]], "137.81"),
    c_("three rounds: paid in all, Worked Example 12.1 (workbook)", LP[I3, 0][1].sum(), "455"),
    flag("one round is dominated by two (workbook)", dominates(F2[BEST_N[2]], F2[BEST_N[1]])),
    flag("Theorem 12.1(i): every dominated schedule is dominated by an efficient one", COVERED),
    c_("ideal point, two objectives: payouts (workbook)", IDEAL2[0], "384.17"),
    c_("ideal point, two objectives: savings (workbook)", IDEAL2[1], "137.81"),
    flag("Theorem 12.1(iii): the ideal point is out of reach (workbook)", not IDEAL2_HIT),
    c_("bound: savings of at least 122.5, payouts (workbook)", LP[EPS_A, 0][0], "330.45"),
    flag("bound: savings of at least 122.5 picks years 1-3, the only answer", SCHEDULES[EPS_A] == (1, 1, 1, 0, 0, 0) and EPS_UNIQUE),
    c_("bound: savings 122.5 and a reserve of 50, payouts (workbook)", LP[EPS_B, 50][0], "285.00"),
]
for r, (dp, ds, rate) in zip(RATES, (("53.72", "17.5", "3.07"), ("69.57", "8.75", "7.95"), ("74.06", "4.375", "16.93"),
                                     ("80.23", "2.1875", "36.68"))):
    nm = f"{r['frm']} to {r['to']} rounds"
    checks += [c_(f"{nm}: payouts given up (workbook)", r["dP"], dp), c_(f"{nm}: savings gained (workbook)", r["dS"], ds),
               c_(f"{nm}: price of savings (workbook)", r["rate"], rate)]
checks += [c_(f"board value, {WORD[n]} (workbook)", BOARD_V[n], s)
           for n, s in ((2, "976.87"), (3, "1,021.93"), (4, "1,001.75"), (5, "952.39"), (6, "884.51"))]
checks += [
    count("the board's weighted sum: rounds (workbook)", NR[BOARD_I], 3),
    flag("the board's weighted sum: years 1-3", SCHEDULES[BOARD_I] == (1, 1, 1, 0, 0, 0)),
    flag("the prices rise along the two-objective frontier (workbook)", INCREASING),
    # Step 4; Figure 12.2 (Panel 3)
    c_("the reserve's price: smallest multiplier (workbook)", PRICES.min(), "0.9091"),
    c_("the reserve's price: largest multiplier (workbook)", PRICES.max(), "0.9091"),
    c_("the reserve's price, Step 4 (workbook)", BETA, "0.909"),
    c_("cash before a year-2 round, after a year-1 round (workbook)", CASH2, "570"),
    c_("rounds in years 1 and 2 need a reserve of at most (workbook)", CASH2 - PRE, "100"),
    c_("cash before a year-3 round, after rounds in years 1 and 2 (workbook)", CASH3, "555"),
    c_("rounds in years 1-3 need a reserve of at most (workbook)", CASH3 - PRE, "85"),
    count("no plan keeps a reserve above (workbook)", BMAX.max(), 150),
    flag("the trade-off surface for savings of 122.5 is the three-round plans'", PHI_IS_3),
    count("three rounds: first switch (workbook)", SW3[0], 86),
    count("three rounds: second switch (workbook)", SW3[1], 101),
    count("three rounds: switches", len(SW3), 2),
    flag("three rounds: years 1-3 up to 85", ENV3[0] == ENV3[85] == SCHEDULES.index((1, 1, 1, 0, 0, 0))),
    flag("three rounds: years 1, 2, 4 from 86 to 100", ENV3[86] == ENV3[100] == SCHEDULES.index((1, 1, 0, 1, 0, 0))),
    flag("three rounds: years 1, 3, 4 from 101 to 150", ENV3[101] == ENV3[150] == SCHEDULES.index((1, 0, 1, 1, 0, 0))),
]
checks += [c_(f"three rounds at a reserve of {b} (workbook)", PHI[b], s) for b, s in
           ((25, "307.72"), (50, "285.00"), (85, "253.18"), (86, "248.25"), (90, "244.61"), (100, "235.52"),
            (101, "210.23"), (120, "192.96"), (150, "165.69"))]
checks += [
    c_("drop beyond the price at 86 (workbook)", DROPS[86], "4.02"),
    c_("drop beyond the price at 101 (workbook)", DROPS[101], "24.38"),
    ("Theorem 12.2(iv): the surface falls at the multiplier where smooth", max(abs(s - BETA) for s in SLOPES), 0.0, 1e-9),
    # three objectives (Panel 4)
    count("plans best for their rounds and reserve (workbook)", len(PTS), 992),
    count("efficient plans, three objectives (workbook)", M, 740),
    count("supported plans (workbook)", SUP.sum(), 453),
    count("plans in dents (workbook)", M - SUP.sum(), 287),
]
for n, (lo, hi, k) in {1: (101, 149, 49), 2: (101, 149, 49), 3: (86, 149, 63), 4: (86, 149, 63), 5: (86, 149, 63)}.items():
    checks += [count(f"dents, {WORD[n]}: plans (workbook)", len(DENTS[n]), k),
               count(f"dents, {WORD[n]}: smallest reserve (workbook)", min(DENTS[n]), lo),
               count(f"dents, {WORD[n]}: largest reserve (workbook)", max(DENTS[n]), hi)]
checks += [
    count("dents, six rounds: none (workbook)", len(DENTS[6]), 0),
    count("dents: smallest reserve, Step 4 (workbook)", min(ALL_DENT_B), 86),
    count("dents: largest reserve, Step 4 (workbook)", max(ALL_DENT_B), 149),
    c_("ideal point: payouts (workbook)", IDEAL[0], "384.17"),
    c_("ideal point: savings (workbook)", IDEAL[1], "137.81"),
    c_("ideal point: reserve (workbook)", IDEAL[2], "150"),
    c_("nadir point: payouts (workbook)", NADIR[0], "22.06"),
    c_("nadir point: savings (workbook)", NADIR[1], "70"),
    c_("nadir point: reserve (workbook)", NADIR[2], "0"),
    flag("nadir payouts: five rounds in years 1, 3-6 with a reserve of 150", E3[KN][1] == 150 and SCHEDULES[E3[KN][2]] == (1, 0, 1, 1, 1, 1)),
    flag("certificates: every supported plan is the best at its weights (workbook)", CHECK_W <= 1e-9),
    flag("certificates: every plan in a dent is beaten by an average (workbook)", CHECK_A > 0),
    count("Section 12.4: support programs, one for each efficient plan", M, 740),
]
CE, C2 = CHOICE["equal"], CHOICE["cfo2"]
checks += [
    # Step 5; Table 12.3; Figure 12.3 (Panel 5)
    flag("equal weights: rounds in years 1, 2, 4 (workbook)", CE["z"] == (1, 1, 0, 1, 0, 0)),
    count("equal weights: reserve (workbook)", CE["b"], 92),
    c_("equal weights: payouts (workbook)", CE["P"], "242.80"),
    c_("equal weights: savings (workbook)", CE["S"], "122.5"),
] + [c_(f"equal weights: payout in year {k + 1} (workbook)", CE["p"][k], s)
     for k, s in enumerate(("8", "90", "0", "122.5", "122.5", "2.5"))] + [
    c_("equal weights: shortfall, payouts (workbook)", CE["d"][0], "0.390"),
    c_("equal weights: shortfall, savings (workbook)", CE["d"][1], "0.226"),
    c_("equal weights: shortfall, reserve (workbook)", CE["d"][2], "0.387"),
    flag("equal weights: in a dent (workbook)", not CE["supported"]),
    c_("equal weights: board value (workbook)", CE["J"], "934.28"),
    flag("CFO twice: rounds in years 1, 2 (workbook)", C2["z"] == (1, 1, 0, 0, 0, 0)),
    count("CFO twice: reserve (workbook)", C2["b"], 96),
    c_("CFO twice: payouts (workbook)", C2["P"], "296.90"),
    c_("CFO twice: savings (workbook)", C2["S"], "105"),
] + [c_(f"CFO twice: payout in year {k + 1} (workbook)", C2["p"][k], s)
     for k, s in enumerate(("4", "105", "105", "105", "90", "0"))] + [
    c_("CFO twice: weighted shortfall, payouts (workbook)", C2["d"][0], "0.482"),
    c_("CFO twice: weighted shortfall, savings (workbook)", C2["d"][1], "0.484"),
    c_("CFO twice: weighted shortfall, reserve (workbook)", C2["d"][2], "0.360"),
    flag("CFO twice: supported (workbook)", C2["supported"]),
    c_("CFO twice: board value (workbook)", C2["J"], "889.60"),
    count("Theorem 12.3(iii): efficient plans each the only choice for its weights (workbook)", REACHED, 740),
    flag("CFO's best: years 1, 2 (workbook)", OWN["CFO"]["z"] == (1, 1, 0, 0, 0, 0)),
    flag("COO's best: all six (workbook)", OWN["COO"]["z"] == (1, 1, 1, 1, 1, 1)),
    flag("treasury's best: year 1 (workbook)", OWN["treasury"]["z"] == (1, 0, 0, 0, 0, 0)),
] + [c_(f"treasury's best: payout in year {k + 1} (workbook)", OWN["treasury"]["p"][k], s)
     for k, s in enumerate(("70", "70", "70", "70", "20", "0"))] + [
    flag("the board's weighted sum, three objectives: years 1-3, no reserve", BOARD3["z"] == (1, 1, 1, 0, 0, 0) and BOARD3["b"] == 0),
    c_("the board's weighted sum, three objectives: value (workbook)", BOARD3["J"], "1,021.93"),
    flag("goals: first answer, years 1, 2, 5 (workbook)", SCHEDULES[G1] == (1, 1, 0, 0, 1, 0)),
    c_("goals: first answer pays (workbook)", G1_PAID, "300.00"),
    c_("goals: first answer, available (workbook)", G1_AVAIL, "300.05"),
    flag("goals: the first answer is dominated (workbook)", G1_DOMINATED),
    flag("goals: second stage, years 1-3 (workbook)", SCHEDULES[G2] == (1, 1, 1, 0, 0, 0)),
    c_("goals: second stage pays (workbook)", G2_P, "307.72"),
    c_("goals: left unpaid, LOS Guidance 12.4 (workbook)", G2_P - G1_PAID, "7.72"),
]
T123_BOOK = {"CFO's best": ("1, 2", "0", "384.17", "105.00", "976.87"), "COO's best": ("1–6", "0", "106.59", "137.81", "884.51"),
             "Treasury's best": ("1", "150", "234.31", "70.00", "629.44"),
             "Board value (Chapter 5)": ("1–3", "0", "330.45", "122.50", "1,021.93"),
             "Chebyshev, equal weights": ("1, 2, 4", "92", "242.80", "122.50", "934.28"),
             "Chebyshev, CFO twice": ("1, 2", "96", "296.90", "105.00", "889.60"),
             "Goals, first answer": ("1, 2, 5", "25", "300.00", "122.50", "991.48"),
             "Goals, second stage": ("1–3", "25", "307.72", "122.50", "999.21")}
for r in TABLE:
    yb, rb, pb, sb, jb = T123_BOOK[r["name"]]
    checks += [flag(f"Table 12.3, {r['name']}: rounds", r["years"] == yb),
               c_(f"Table 12.3, {r['name']}: reserve (workbook)", r["reserve"], rb),
               c_(f"Table 12.3, {r['name']}: payouts (workbook)", r["P"], pb),
               c_(f"Table 12.3, {r['name']}: savings (workbook)", r["S"], sb),
               c_(f"Table 12.3, {r['name']}: board value (workbook)", r["J"], jb)]
checks += [
    c_("Figure 12.3: board value from (workbook)", J_RANGE[0], "629.44"),
    c_("Figure 12.3: board value to (workbook)", J_RANGE[1], "1,021.93"),
    # Section 12.4; Proposition 12.2 (Panel 6)
    flag("Proposition 12.2(i): bounds box the frontier", BOXED),
    flag("Proposition 12.2(ii): sets of 1-3 schedules with the most hypervolume are efficient", HV_OK),
    flag("Proposition 12.2(ii): an undominated schedule adds hypervolume, a dominated one none", HV_ADD),
    flag("Proposition 12.2(iii): random positive weights return only supported plans", SWEEP_OK),
    flag("Proposition 12.2(iv): the unbounded archive ends with the efficient set", T_EVO is not None),
    # Section 12.5; Proposition 12.3 (Panel 7)
    c_("Chapter 9's frontier: a point of appetite, 4% to 5% (workbook)", P45, "1.9"),
    c_("Chapter 9's frontier: a point of appetite, 5% to 6% (workbook)", P56, "1.6"),
    flag("Proposition 12.3(i): the least cost never rises with the appetite (workbook)", MONO),
    flag("declared costs: 3% of the line plus $8m a quarter, within the delays' rounding (workbook)", LEVERS_GAP <= 8 * 0.0005 + 0.005),
    c_("Proposition 12.3(ii): the family above one law, Worked Example 11.1 (workbook)", FAMILY_5 - C9[4], "13.56"),
    c_("Proposition 12.3(iii): the average's cost (workbook)", AVG[0], "0.9"),
    c_("Proposition 12.3(iii): the average's chance (workbook)", AVG[1], "0.55"),
]
ok = 0
for name, got, want, tol in checks:
    status = "PASS" if abs(got - want) <= tol else "FAIL"
    ok += status == "PASS"
    print(f"{name:84s} {got:11.4f}   book {want:10.4f}   {status}")
print(f"\n{ok} of {len(checks)} checkpoints PASS")
payout programs, 64 schedules x 151 reserves (workbook)                                9664.0000   book  9664.0000   PASS
payout programs that keep every commitment (workbook)                                  7824.0000   book  7824.0000   PASS
war chest = linear program: the same feasibility in all 9,664                             1.0000   book     1.0000   PASS
war chest = linear program: largest difference                                            0.0000   book     0.0000   PASS
savings depend only on the rounds: 140 - 140/2^n (workbook)                               1.0000   book     1.0000   PASS
Chapter 5's price of savings, Section 12.2 (workbook)                                     5.6447   book     5.6450   PASS
Chapter 5's price of savings, Online Appendix (workbook)                                  5.6447   book     5.6447   PASS
schedules keeping every commitment, no reserve (workbook)                                64.0000   book    64.0000   PASS
dominated schedules (workbook)                                                           59.0000   book    59.0000   PASS
efficient schedules (workbook)                                                            5.0000   book     5.0000   PASS
the efficient plans are the best with two to six rounds                                   1.0000   book     1.0000   PASS
no rounds: payouts (workbook)                                                           136.3636   book   136.3600   PASS
no rounds: payout in year 1 (workbook)                                                  150.0000   book   150.0000   PASS
one round: payouts (workbook)                                                           370.6726   book   370.6700   PASS
one round: savings (workbook)                                                            70.0000   book    70.0000   PASS
one round: payout in year 1 (workbook)                                                  220.0000   book   220.0000   PASS
one round: payout in year 2 (workbook)                                                   70.0000   book    70.0000   PASS
one round: payout in year 3 (workbook)                                                   70.0000   book    70.0000   PASS
one round: payout in year 4 (workbook)                                                   70.0000   book    70.0000   PASS
one round: payout in year 5 (workbook)                                                   20.0000   book    20.0000   PASS
one round: payout in year 6 (workbook)                                                    0.0000   book     0.0000   PASS
two rounds: payouts (workbook)                                                          384.1733   book   384.1700   PASS
two rounds: savings (workbook)                                                          105.0000   book   105.0000   PASS
two rounds: payout in year 1 (workbook)                                                 100.0000   book   100.0000   PASS
two rounds: payout in year 2 (workbook)                                                 105.0000   book   105.0000   PASS
two rounds: payout in year 3 (workbook)                                                 105.0000   book   105.0000   PASS
two rounds: payout in year 4 (workbook)                                                 105.0000   book   105.0000   PASS
two rounds: payout in year 5 (workbook)                                                  90.0000   book    90.0000   PASS
two rounds: payout in year 6 (workbook)                                                   0.0000   book     0.0000   PASS
three rounds: payouts (workbook)                                                        330.4520   book   330.4500   PASS
three rounds: savings (workbook)                                                        122.5000   book   122.5000   PASS
three rounds: payout in year 1 (workbook)                                                85.0000   book    85.0000   PASS
three rounds: payout in year 2 (workbook)                                                 0.0000   book     0.0000   PASS
three rounds: payout in year 3 (workbook)                                               122.5000   book   122.5000   PASS
three rounds: payout in year 4 (workbook)                                               122.5000   book   122.5000   PASS
three rounds: payout in year 5 (workbook)                                               122.5000   book   122.5000   PASS
three rounds: payout in year 6 (workbook)                                                 2.5000   book     2.5000   PASS
four rounds: payouts (workbook)                                                         260.8806   book   260.8800   PASS
four rounds: savings (workbook)                                                         131.2500   book   131.2500   PASS
four rounds: payout in year 1 (workbook)                                                 85.0000   book    85.0000   PASS
four rounds: payout in year 2 (workbook)                                                  0.0000   book     0.0000   PASS
four rounds: payout in year 3 (workbook)                                                122.5000   book   122.5000   PASS
four rounds: payout in year 4 (workbook)                                                122.5000   book   122.5000   PASS
four rounds: payout in year 5 (workbook)                                                  2.5000   book     2.5000   PASS
four rounds: payout in year 6 (workbook)                                                 11.2500   book    11.2500   PASS
five rounds: payouts (workbook)                                                         186.8216   book   186.8200   PASS
five rounds: savings (workbook)                                                         135.6250   book   135.6250   PASS
five rounds: payout in year 1 (workbook)                                                 85.0000   book    85.0000   PASS
five rounds: payout in year 2 (workbook)                                                  0.0000   book     0.0000   PASS
five rounds: payout in year 3 (workbook)                                                122.5000   book   122.5000   PASS
five rounds: payout in year 4 (workbook)                                                  2.5000   book     2.5000   PASS
five rounds: payout in year 5 (workbook)                                                 11.2500   book    11.2500   PASS
five rounds: payout in year 6 (workbook)                                                 15.6250   book    15.6250   PASS
six rounds: payouts (workbook)                                                          106.5915   book   106.5900   PASS
six rounds: savings (workbook)                                                          137.8125   book   137.8125   PASS
six rounds: payout in year 1 (workbook)                                                  85.0000   book    85.0000   PASS
six rounds: payout in year 2 (workbook)                                                   0.0000   book     0.0000   PASS
six rounds: payout in year 3 (workbook)                                                   2.5000   book     2.5000   PASS
six rounds: payout in year 4 (workbook)                                                  11.2500   book    11.2500   PASS
six rounds: payout in year 5 (workbook)                                                  15.6250   book    15.6250   PASS
six rounds: payout in year 6 (workbook)                                                  17.8125   book    17.8125   PASS
three rounds: cash at the end of year 1 (workbook)                                      485.0000   book   485.0000   PASS
three rounds: cash at the end of year 2 (workbook)                                      470.0000   book   470.0000   PASS
three rounds: cash at the end of year 3 (workbook)                                      350.0000   book   350.0000   PASS
three rounds: cash at the end of year 4 (workbook)                                      350.0000   book   350.0000   PASS
three rounds: cash at the end of year 5 (workbook)                                      350.0000   book   350.0000   PASS
three rounds: cash at the end of year 6 (workbook)                                      470.0000   book   470.0000   PASS
five rounds: savings as Step 2 prints them                                              135.6250   book   135.6300   PASS
six rounds: savings as Step 2 prints them                                               137.8125   book   137.8100   PASS
three rounds: paid in all, Worked Example 12.1 (workbook)                               455.0000   book   455.0000   PASS
one round is dominated by two (workbook)                                                  1.0000   book     1.0000   PASS
Theorem 12.1(i): every dominated schedule is dominated by an efficient one                1.0000   book     1.0000   PASS
ideal point, two objectives: payouts (workbook)                                         384.1733   book   384.1700   PASS
ideal point, two objectives: savings (workbook)                                         137.8125   book   137.8100   PASS
Theorem 12.1(iii): the ideal point is out of reach (workbook)                             1.0000   book     1.0000   PASS
bound: savings of at least 122.5, payouts (workbook)                                    330.4520   book   330.4500   PASS
bound: savings of at least 122.5 picks years 1-3, the only answer                         1.0000   book     1.0000   PASS
bound: savings 122.5 and a reserve of 50, payouts (workbook)                            284.9974   book   285.0000   PASS
2 to 3 rounds: payouts given up (workbook)                                               53.7214   book    53.7200   PASS
2 to 3 rounds: savings gained (workbook)                                                 17.5000   book    17.5000   PASS
2 to 3 rounds: price of savings (workbook)                                                3.0698   book     3.0700   PASS
3 to 4 rounds: payouts given up (workbook)                                               69.5714   book    69.5700   PASS
3 to 4 rounds: savings gained (workbook)                                                  8.7500   book     8.7500   PASS
3 to 4 rounds: price of savings (workbook)                                                7.9510   book     7.9500   PASS
4 to 5 rounds: payouts given up (workbook)                                               74.0590   book    74.0600   PASS
4 to 5 rounds: savings gained (workbook)                                                  4.3750   book     4.3750   PASS
4 to 5 rounds: price of savings (workbook)                                               16.9278   book    16.9300   PASS
5 to 6 rounds: payouts given up (workbook)                                               80.2301   book    80.2300   PASS
5 to 6 rounds: savings gained (workbook)                                                  2.1875   book     2.1875   PASS
5 to 6 rounds: price of savings (workbook)                                               36.6766   book    36.6800   PASS
board value, two rounds (workbook)                                                      976.8710   book   976.8700   PASS
board value, three rounds (workbook)                                                   1021.9325   book  1021.9300   PASS
board value, four rounds (workbook)                                                    1001.7526   book  1001.7500   PASS
board value, five rounds (workbook)                                                     952.3894   book   952.3900   PASS
board value, six rounds (workbook)                                                      884.5071   book   884.5100   PASS
the board's weighted sum: rounds (workbook)                                               3.0000   book     3.0000   PASS
the board's weighted sum: years 1-3                                                       1.0000   book     1.0000   PASS
the prices rise along the two-objective frontier (workbook)                               1.0000   book     1.0000   PASS
the reserve's price: smallest multiplier (workbook)                                       0.9091   book     0.9091   PASS
the reserve's price: largest multiplier (workbook)                                        0.9091   book     0.9091   PASS
the reserve's price, Step 4 (workbook)                                                    0.9091   book     0.9090   PASS
cash before a year-2 round, after a year-1 round (workbook)                             570.0000   book   570.0000   PASS
rounds in years 1 and 2 need a reserve of at most (workbook)                            100.0000   book   100.0000   PASS
cash before a year-3 round, after rounds in years 1 and 2 (workbook)                    555.0000   book   555.0000   PASS
rounds in years 1-3 need a reserve of at most (workbook)                                 85.0000   book    85.0000   PASS
no plan keeps a reserve above (workbook)                                                150.0000   book   150.0000   PASS
the trade-off surface for savings of 122.5 is the three-round plans'                      1.0000   book     1.0000   PASS
three rounds: first switch (workbook)                                                    86.0000   book    86.0000   PASS
three rounds: second switch (workbook)                                                  101.0000   book   101.0000   PASS
three rounds: switches                                                                    2.0000   book     2.0000   PASS
three rounds: years 1-3 up to 85                                                          1.0000   book     1.0000   PASS
three rounds: years 1, 2, 4 from 86 to 100                                                1.0000   book     1.0000   PASS
three rounds: years 1, 3, 4 from 101 to 150                                               1.0000   book     1.0000   PASS
three rounds at a reserve of 25 (workbook)                                              307.7247   book   307.7200   PASS
three rounds at a reserve of 50 (workbook)                                              284.9974   book   285.0000   PASS
three rounds at a reserve of 85 (workbook)                                              253.1793   book   253.1800   PASS
three rounds at a reserve of 86 (workbook)                                              248.2506   book   248.2500   PASS
three rounds at a reserve of 90 (workbook)                                              244.6143   book   244.6100   PASS
three rounds at a reserve of 100 (workbook)                                             235.5234   book   235.5200   PASS
three rounds at a reserve of 101 (workbook)                                             210.2341   book   210.2300   PASS
three rounds at a reserve of 120 (workbook)                                             192.9614   book   192.9600   PASS
three rounds at a reserve of 150 (workbook)                                             165.6886   book   165.6900   PASS
drop beyond the price at 86 (workbook)                                                    4.0195   book     4.0200   PASS
drop beyond the price at 101 (workbook)                                                  24.3802   book    24.3800   PASS
Theorem 12.2(iv): the surface falls at the multiplier where smooth                        0.0000   book     0.0000   PASS
plans best for their rounds and reserve (workbook)                                      992.0000   book   992.0000   PASS
efficient plans, three objectives (workbook)                                            740.0000   book   740.0000   PASS
supported plans (workbook)                                                              453.0000   book   453.0000   PASS
plans in dents (workbook)                                                               287.0000   book   287.0000   PASS
dents, one round: plans (workbook)                                                       49.0000   book    49.0000   PASS
dents, one round: smallest reserve (workbook)                                           101.0000   book   101.0000   PASS
dents, one round: largest reserve (workbook)                                            149.0000   book   149.0000   PASS
dents, two rounds: plans (workbook)                                                      49.0000   book    49.0000   PASS
dents, two rounds: smallest reserve (workbook)                                          101.0000   book   101.0000   PASS
dents, two rounds: largest reserve (workbook)                                           149.0000   book   149.0000   PASS
dents, three rounds: plans (workbook)                                                    63.0000   book    63.0000   PASS
dents, three rounds: smallest reserve (workbook)                                         86.0000   book    86.0000   PASS
dents, three rounds: largest reserve (workbook)                                         149.0000   book   149.0000   PASS
dents, four rounds: plans (workbook)                                                     63.0000   book    63.0000   PASS
dents, four rounds: smallest reserve (workbook)                                          86.0000   book    86.0000   PASS
dents, four rounds: largest reserve (workbook)                                          149.0000   book   149.0000   PASS
dents, five rounds: plans (workbook)                                                     63.0000   book    63.0000   PASS
dents, five rounds: smallest reserve (workbook)                                          86.0000   book    86.0000   PASS
dents, five rounds: largest reserve (workbook)                                          149.0000   book   149.0000   PASS
dents, six rounds: none (workbook)                                                        0.0000   book     0.0000   PASS
dents: smallest reserve, Step 4 (workbook)                                               86.0000   book    86.0000   PASS
dents: largest reserve, Step 4 (workbook)                                               149.0000   book   149.0000   PASS
ideal point: payouts (workbook)                                                         384.1733   book   384.1700   PASS
ideal point: savings (workbook)                                                         137.8125   book   137.8100   PASS
ideal point: reserve (workbook)                                                         150.0000   book   150.0000   PASS
nadir point: payouts (workbook)                                                          22.0583   book    22.0600   PASS
nadir point: savings (workbook)                                                          70.0000   book    70.0000   PASS
nadir point: reserve (workbook)                                                           0.0000   book     0.0000   PASS
nadir payouts: five rounds in years 1, 3-6 with a reserve of 150                          1.0000   book     1.0000   PASS
certificates: every supported plan is the best at its weights (workbook)                  1.0000   book     1.0000   PASS
certificates: every plan in a dent is beaten by an average (workbook)                     1.0000   book     1.0000   PASS
Section 12.4: support programs, one for each efficient plan                             740.0000   book   740.0000   PASS
equal weights: rounds in years 1, 2, 4 (workbook)                                         1.0000   book     1.0000   PASS
equal weights: reserve (workbook)                                                        92.0000   book    92.0000   PASS
equal weights: payouts (workbook)                                                       242.7961   book   242.8000   PASS
equal weights: savings (workbook)                                                       122.5000   book   122.5000   PASS
equal weights: payout in year 1 (workbook)                                                8.0000   book     8.0000   PASS
equal weights: payout in year 2 (workbook)                                               90.0000   book    90.0000   PASS
equal weights: payout in year 3 (workbook)                                                0.0000   book     0.0000   PASS
equal weights: payout in year 4 (workbook)                                              122.5000   book   122.5000   PASS
equal weights: payout in year 5 (workbook)                                              122.5000   book   122.5000   PASS
equal weights: payout in year 6 (workbook)                                                2.5000   book     2.5000   PASS
equal weights: shortfall, payouts (workbook)                                              0.3904   book     0.3900   PASS
equal weights: shortfall, savings (workbook)                                              0.2258   book     0.2260   PASS
equal weights: shortfall, reserve (workbook)                                              0.3867   book     0.3870   PASS
equal weights: in a dent (workbook)                                                       1.0000   book     1.0000   PASS
equal weights: board value (workbook)                                                   934.2767   book   934.2800   PASS
CFO twice: rounds in years 1, 2 (workbook)                                                1.0000   book     1.0000   PASS
CFO twice: reserve (workbook)                                                            96.0000   book    96.0000   PASS
CFO twice: payouts (workbook)                                                           296.9006   book   296.9000   PASS
CFO twice: savings (workbook)                                                           105.0000   book   105.0000   PASS
CFO twice: payout in year 1 (workbook)                                                    4.0000   book     4.0000   PASS
CFO twice: payout in year 2 (workbook)                                                  105.0000   book   105.0000   PASS
CFO twice: payout in year 3 (workbook)                                                  105.0000   book   105.0000   PASS
CFO twice: payout in year 4 (workbook)                                                  105.0000   book   105.0000   PASS
CFO twice: payout in year 5 (workbook)                                                   90.0000   book    90.0000   PASS
CFO twice: payout in year 6 (workbook)                                                    0.0000   book     0.0000   PASS
CFO twice: weighted shortfall, payouts (workbook)                                         0.4820   book     0.4820   PASS
CFO twice: weighted shortfall, savings (workbook)                                         0.4839   book     0.4840   PASS
CFO twice: weighted shortfall, reserve (workbook)                                         0.3600   book     0.3600   PASS
CFO twice: supported (workbook)                                                           1.0000   book     1.0000   PASS
CFO twice: board value (workbook)                                                       889.5982   book   889.6000   PASS
Theorem 12.3(iii): efficient plans each the only choice for its weights (workbook)      740.0000   book   740.0000   PASS
CFO's best: years 1, 2 (workbook)                                                         1.0000   book     1.0000   PASS
COO's best: all six (workbook)                                                            1.0000   book     1.0000   PASS
treasury's best: year 1 (workbook)                                                        1.0000   book     1.0000   PASS
treasury's best: payout in year 1 (workbook)                                             70.0000   book    70.0000   PASS
treasury's best: payout in year 2 (workbook)                                             70.0000   book    70.0000   PASS
treasury's best: payout in year 3 (workbook)                                             70.0000   book    70.0000   PASS
treasury's best: payout in year 4 (workbook)                                             70.0000   book    70.0000   PASS
treasury's best: payout in year 5 (workbook)                                             20.0000   book    20.0000   PASS
treasury's best: payout in year 6 (workbook)                                              0.0000   book     0.0000   PASS
the board's weighted sum, three objectives: years 1-3, no reserve                         1.0000   book     1.0000   PASS
the board's weighted sum, three objectives: value (workbook)                           1021.9325   book  1021.9300   PASS
goals: first answer, years 1, 2, 5 (workbook)                                             1.0000   book     1.0000   PASS
goals: first answer pays (workbook)                                                     300.0000   book   300.0000   PASS
goals: first answer, available (workbook)                                               300.0511   book   300.0500   PASS
goals: the first answer is dominated (workbook)                                           1.0000   book     1.0000   PASS
goals: second stage, years 1-3 (workbook)                                                 1.0000   book     1.0000   PASS
goals: second stage pays (workbook)                                                     307.7247   book   307.7200   PASS
goals: left unpaid, LOS Guidance 12.4 (workbook)                                          7.7247   book     7.7200   PASS
Table 12.3, CFO's best: rounds                                                            1.0000   book     1.0000   PASS
Table 12.3, CFO's best: reserve (workbook)                                                0.0000   book     0.0000   PASS
Table 12.3, CFO's best: payouts (workbook)                                              384.1733   book   384.1700   PASS
Table 12.3, CFO's best: savings (workbook)                                              105.0000   book   105.0000   PASS
Table 12.3, CFO's best: board value (workbook)                                          976.8710   book   976.8700   PASS
Table 12.3, COO's best: rounds                                                            1.0000   book     1.0000   PASS
Table 12.3, COO's best: reserve (workbook)                                                0.0000   book     0.0000   PASS
Table 12.3, COO's best: payouts (workbook)                                              106.5915   book   106.5900   PASS
Table 12.3, COO's best: savings (workbook)                                              137.8125   book   137.8100   PASS
Table 12.3, COO's best: board value (workbook)                                          884.5071   book   884.5100   PASS
Table 12.3, Treasury's best: rounds                                                       1.0000   book     1.0000   PASS
Table 12.3, Treasury's best: reserve (workbook)                                         150.0000   book   150.0000   PASS
Table 12.3, Treasury's best: payouts (workbook)                                         234.3090   book   234.3100   PASS
Table 12.3, Treasury's best: savings (workbook)                                          70.0000   book    70.0000   PASS
Table 12.3, Treasury's best: board value (workbook)                                     629.4408   book   629.4400   PASS
Table 12.3, Board value (Chapter 5): rounds                                               1.0000   book     1.0000   PASS
Table 12.3, Board value (Chapter 5): reserve (workbook)                                   0.0000   book     0.0000   PASS
Table 12.3, Board value (Chapter 5): payouts (workbook)                                 330.4520   book   330.4500   PASS
Table 12.3, Board value (Chapter 5): savings (workbook)                                 122.5000   book   122.5000   PASS
Table 12.3, Board value (Chapter 5): board value (workbook)                            1021.9325   book  1021.9300   PASS
Table 12.3, Chebyshev, equal weights: rounds                                              1.0000   book     1.0000   PASS
Table 12.3, Chebyshev, equal weights: reserve (workbook)                                 92.0000   book    92.0000   PASS
Table 12.3, Chebyshev, equal weights: payouts (workbook)                                242.7961   book   242.8000   PASS
Table 12.3, Chebyshev, equal weights: savings (workbook)                                122.5000   book   122.5000   PASS
Table 12.3, Chebyshev, equal weights: board value (workbook)                            934.2767   book   934.2800   PASS
Table 12.3, Chebyshev, CFO twice: rounds                                                  1.0000   book     1.0000   PASS
Table 12.3, Chebyshev, CFO twice: reserve (workbook)                                     96.0000   book    96.0000   PASS
Table 12.3, Chebyshev, CFO twice: payouts (workbook)                                    296.9006   book   296.9000   PASS
Table 12.3, Chebyshev, CFO twice: savings (workbook)                                    105.0000   book   105.0000   PASS
Table 12.3, Chebyshev, CFO twice: board value (workbook)                                889.5982   book   889.6000   PASS
Table 12.3, Goals, first answer: rounds                                                   1.0000   book     1.0000   PASS
Table 12.3, Goals, first answer: reserve (workbook)                                      25.0000   book    25.0000   PASS
Table 12.3, Goals, first answer: payouts (workbook)                                     300.0000   book   300.0000   PASS
Table 12.3, Goals, first answer: savings (workbook)                                     122.5000   book   122.5000   PASS
Table 12.3, Goals, first answer: board value (workbook)                                 991.4806   book   991.4800   PASS
Table 12.3, Goals, second stage: rounds                                                   1.0000   book     1.0000   PASS
Table 12.3, Goals, second stage: reserve (workbook)                                      25.0000   book    25.0000   PASS
Table 12.3, Goals, second stage: payouts (workbook)                                     307.7247   book   307.7200   PASS
Table 12.3, Goals, second stage: savings (workbook)                                     122.5000   book   122.5000   PASS
Table 12.3, Goals, second stage: board value (workbook)                                 999.2053   book   999.2100   PASS
Figure 12.3: board value from (workbook)                                                629.4408   book   629.4400   PASS
Figure 12.3: board value to (workbook)                                                 1021.9325   book  1021.9300   PASS
Proposition 12.2(i): bounds box the frontier                                              1.0000   book     1.0000   PASS
Proposition 12.2(ii): sets of 1-3 schedules with the most hypervolume are efficient       1.0000   book     1.0000   PASS
Proposition 12.2(ii): an undominated schedule adds hypervolume, a dominated one none      1.0000   book     1.0000   PASS
Proposition 12.2(iii): random positive weights return only supported plans                1.0000   book     1.0000   PASS
Proposition 12.2(iv): the unbounded archive ends with the efficient set                   1.0000   book     1.0000   PASS
Chapter 9's frontier: a point of appetite, 4% to 5% (workbook)                            1.9100   book     1.9000   PASS
Chapter 9's frontier: a point of appetite, 5% to 6% (workbook)                            1.6200   book     1.6000   PASS
Proposition 12.3(i): the least cost never rises with the appetite (workbook)              1.0000   book     1.0000   PASS
declared costs: 3% of the line plus $8m a quarter, within the delays' rounding (workbook)      1.0000   book     1.0000   PASS
Proposition 12.3(ii): the family above one law, Worked Example 11.1 (workbook)           13.5600   book    13.5600   PASS
Proposition 12.3(iii): the average's cost (workbook)                                      0.9000   book     0.9000   PASS
Proposition 12.3(iii): the average's chance (workbook)                                    0.5500   book     0.5500   PASS

260 of 260 checkpoints PASS

Next. Chapter 13, Machine Learning for Enterprise Transformation, turns from models the board declares to models learned from the enterprise's own data. Chapters 1–12 optimized declared models: dynamics written down, laws and families of laws declared, and, in this chapter, objectives stated by their owners and chosen between on a frontier. AXIOM-12 runs the same instance interactively (Lab 2.12); this notebook and the workbook are its reproducible record.

Dynamic Corporate Transformation — Volume I: Foundations and Mathematical Framework; Volume II: Enterprise Optimization & Applications
Jump to any chapter

Volume I: Foundations and Mathematical Framework

  1. Ch. 1Introduction to Dynamic Corporate Transformation
  2. Ch. 2Enterprise Systems and Transformation
  3. Ch. 3Mathematical Foundations
  4. Ch. 4Enterprise Modeling Principles
  5. Ch. 5Enterprise State Representation
  6. Ch. 6Enterprise Transformation Operators
  7. Ch. 7Dynamic Enterprise Systems
  8. Ch. 8Stochastic Enterprise Dynamics
  9. Ch. 9Enterprise State Architecture
  10. Ch. 10Enterprise Capital Architecture
  11. Ch. 11Enterprise Performance Architecture
  12. Ch. 12Enterprise Risk, Resilience, and Robustness Architecture
  13. Ch. 13Unified Enterprise Transformation Architecture
  14. Ch. 14Mathematical Analysis of the Unified Enterprise Transformation Architecture
  15. Ch. 15Mathematical Formulation of Enterprise Optimization
  16. Ch. 16Synthesis, Future Directions, and Transition to Volume II

Volume II: Enterprise Optimization & Applications

  1. Ch. 1Introduction to Enterprise Optimization
  2. Ch. 2The General Enterprise Optimization Problem Revisited
  3. Ch. 3Convex Enterprise Optimization
  4. Ch. 4Nonlinear and Mixed-Integer Enterprise Optimization
  5. Ch. 5Dynamic Enterprise Optimization
  6. Ch. 6Optimal Control of Enterprise Systems
  7. Ch. 7Dynamic Programming for Enterprise Systems
  8. Ch. 8Hamilton–Jacobi–Bellman Enterprise Framework
  9. Ch. 9Stochastic Enterprise Optimization
  10. Ch. 10Neuro-Fuzzy Robust Enterprise Optimization
  11. Ch. 11Distributionally Robust Enterprise Optimization
  12. Ch. 12Multi-Objective Enterprise Optimization
  13. Ch. 13Machine Learning for Enterprise Transformation
  14. Ch. 14Artificial Intelligence for Enterprise Optimization
  15. Ch. 15Enterprise Digital Twins and Autonomous Enterprise Transformation
  16. Ch. 16Enterprise Applications and Integrated Transformation Case Studies

Dynamic Corporate Transformation

Springer Nature · Management for Professionals · Forthcoming 2027

  • Volume I — Foundations and Mathematical Framework
  • Volume II — Enterprise Optimization & Applications
Buy Volume I: Dynamic Corporate Transformation: Foundations and Mathematical FrameworkBuy Volume II: Dynamic Corporate Transformation: Enterprise Optimization & Applications
Jump to any chapter

Volume I: Foundations and Mathematical Framework

  1. Ch. 1Introduction to Dynamic Corporate Transformation
  2. Ch. 2Enterprise Systems and Transformation
  3. Ch. 3Mathematical Foundations
  4. Ch. 4Enterprise Modeling Principles
  5. Ch. 5Enterprise State Representation
  6. Ch. 6Enterprise Transformation Operators
  7. Ch. 7Dynamic Enterprise Systems
  8. Ch. 8Stochastic Enterprise Dynamics
  9. Ch. 9Enterprise State Architecture
  10. Ch. 10Enterprise Capital Architecture
  11. Ch. 11Enterprise Performance Architecture
  12. Ch. 12Enterprise Risk, Resilience, and Robustness Architecture
  13. Ch. 13Unified Enterprise Transformation Architecture
  14. Ch. 14Mathematical Analysis of the Unified Enterprise Transformation Architecture
  15. Ch. 15Mathematical Formulation of Enterprise Optimization
  16. Ch. 16Synthesis, Future Directions, and Transition to Volume II

Volume II: Enterprise Optimization & Applications

  1. Ch. 1Introduction to Enterprise Optimization
  2. Ch. 2The General Enterprise Optimization Problem Revisited
  3. Ch. 3Convex Enterprise Optimization
  4. Ch. 4Nonlinear and Mixed-Integer Enterprise Optimization
  5. Ch. 5Dynamic Enterprise Optimization
  6. Ch. 6Optimal Control of Enterprise Systems
  7. Ch. 7Dynamic Programming for Enterprise Systems
  8. Ch. 8Hamilton–Jacobi–Bellman Enterprise Framework
  9. Ch. 9Stochastic Enterprise Optimization
  10. Ch. 10Neuro-Fuzzy Robust Enterprise Optimization
  11. Ch. 11Distributionally Robust Enterprise Optimization
  12. Ch. 12Multi-Objective Enterprise Optimization
  13. Ch. 13Machine Learning for Enterprise Transformation
  14. Ch. 14Artificial Intelligence for Enterprise Optimization
  15. Ch. 15Enterprise Digital Twins and Autonomous Enterprise Transformation
  16. Ch. 16Enterprise Applications and Integrated Transformation Case Studies