Skip to content

Vol. I, Ch. 11

DCT Laboratory — Volume I, Chapter 11

Enterprise Performance Architecture

This laboratory demonstrates

Seed 26111 · Companion to Chapter 11 and AXIOM Lab 1.11 (module AXIOM-11) · Mirrored in DCT_V1_Ch11_Lab.xlsx

This laboratory reproduces Worked Example 11.1, Who Really Chose Digital?, with the book's declarations, and works the five steps of AXIOM Lab 1.11, Auditing Meridian's Scorecard: build the dashboard, sweep the weights, change the yardstick, roll up a ratio, walk the frontier and move the window. The 200,000 weightings that measure how much of the space of weights each option wins are drawn, as in the book, with seed 26111, the lab's seed (Online Appendix 11.B.6). Every Meridian number below is the book's (Worked Example 11.1, Tables 11.4–11.6, Figures 11.1–11.3, Section 11.8, Online Appendix 11.B); the final cell checks them. Money is in $ million.

The declarations (Worked Example 11.1, the situation; Worked Example 1.1; Online Appendix 1.B.4)

In Worked Example 1.1 the chief financial officer's scorecard, Score=14 (Liquidity+Technology+Efficiency−Strategic risk)\text{Score} = \tfrac14\,(\text{Liquidity} + \text{Technology} + \text{Efficiency} - \text{Strategic risk}), ranked the digital option first at Year 5. The paths come from the Chapter 1 model (Online Appendix 1.B.4): four of Meridian's eight readings move, liquidity (indexed to today =100= 100, that is $620 million), the technology platform, operational efficiency and strategic risk (lower is better). Each reading relaxes toward an option-specific target at its own speed, with a trough for temporary strain and a jump when the restructuring's sale closes at Year 1, xi(t)=xˉi+(xi(0)−xˉi)e−kit+σiDi tτi e1−t/τi+Ji 1{t≥1},x_i(t) = \bar x_i + \bigl(x_i(0) - \bar x_i\bigr)e^{-k_i t} + \sigma_i D_i\,\frac{t}{\tau_i}\,e^{1 - t/\tau_i} + J_i\,\mathbf{1}\{t \ge 1\}, from the state declared in Chapter 3: liquidity 100, technology 41, efficiency 71, strategic risk 64, a score of 37.0. The head of Digital Solutions presents the ranking as the case for digital; the chair asks whether the data chose digital, or the design of the scorecard.

Python Code
import numpy as np
import matplotlib.pyplot as plt
from matplotlib.colors import ListedColormap
from matplotlib.patches import Patch
plt.rcParams.update({"figure.dpi": 110, "axes.spines.top": False, "axes.spines.right": False})
SEED = 26111      # the lab's seed; Online Appendix 11.B.6 draws its 200,000 weightings with it

# The Chapter 1 model (Online Appendix 1.B.4); t in years
READINGS = ["liquidity", "technology", "efficiency", "strategic risk"]
TODAY = np.array([100.0, 41.0, 71.0, 64.0])     # liquidity indexed to today = 100 ($620 million); indices 0-100
SIGN = np.array([1.0, 1.0, 1.0, -1.0])          # strategic risk enters with a minus sign: lower is better
OPTIONS = ["digital", "restructuring", "turnaround"]
COLORS = ["#1F4E79", "#E65100", "#B71C1C"]      # digital, restructuring, turnaround
SALE = 1.0                                      # the restructuring's sale closes at Year 1
PARAMS = {  # each reading: target, speed a year, trough sigma*D, trough time tau (years), jump at the sale
    "digital":       [(112, 0.35, -28, 2.0, 0), (78, 0.55, 0, None, 0), (77, 0.50, -9, 1.5, 0), (51, 0.40, 14, 1.2, 0)],
    "restructuring": [(104, 0.30, 0, None, 18), (38, 0.25, 0, None, 0), (74, 0.45, 0, None, 0), (56, 0.35, 0, None, -8)],
    "turnaround":    [(106, 0.45, 0, None, 0), (33, 0.20, 0, None, 0), (79, 0.60, 0, None, 0), (67, 0.30, 0, None, 0)]}

def readings(option, t):
    # the four readings of an option at times t (years), one row per time
    t = np.atleast_1d(np.asarray(t, float))
    cols = []
    for x0, (target, k, depth, tau, jump) in zip(TODAY, PARAMS[option]):
        x = target + (x0 - target) * np.exp(-k * t) + jump * (t >= SALE)
        if depth:
            x = x + depth * (t / tau) * np.exp(1 - t / tau)
        cols.append(x)
    return np.column_stack(cols)

EQUAL = np.full(4, 0.25)
def score(x, w=EQUAL):
    # the CFO's scorecard: w . (L, T, E, -R); equal weights give (L + T + E - R)/4
    return (np.asarray(x, float) * SIGN) @ w

Y5 = np.vstack([readings(o, 5.0) for o in OPTIONS])                 # Year-5 readings, one row per option
MONTHS = np.arange(61)                                               # five years, month by month
S = np.vstack([score(readings(o, MONTHS / 12)) for o in OPTIONS])    # the scorecard of each option, monthly
print(f"today: liquidity {TODAY[0]:.0f} ($620 million), technology {TODAY[1]:.0f}, efficiency {TODAY[2]:.0f}, "
      f"strategic risk {TODAY[3]:.0f} -> score {score(TODAY):.1f}")
print(f"Chapter 1 model: {len(OPTIONS)} options x {len(READINGS)} moving readings, month by month for five years; "
      f"the sale closes at Year {SALE:.0f}")
today: liquidity 100 ($620 million), technology 41, efficiency 71, strategic risk 64 -> score 37.0
Chapter 1 model: 3 options x 4 moving readings, month by month for five years; the sale closes at Year 1

Panel 1 — Build the dashboard (Tables 11.2 and 11.4; Section 11.8: $23 million a year, below 55 within three quarters)

AXIOM Lab 1.11, step 1. The performance vector has ten dimensions (Equation 11.2, Table 11.2), each a flow read off the state and the decisions (Definition 11.1). Meridian's dashboard declares six of them with their supports on Chapter 9's architecture (Table 11.4): financial and resilience performance at treasury, operational at Industrial Systems and Materials, strategic at the three divisions, technological at Digital, innovation at Digital and engineering. Each dimension needs an indicator with its error and a band. Meridian's own readings declare three performance indicators (Table 5.2 tags efficiency, ROIC and market share as performance) and the liquidity buffer that resilience reads (Definition 11.12); technological and innovation performance have no declared indicator yet, and a dimension without one is blind by construction (Section 11.4). Of the CFO's four readings, only efficiency is a performance reading: Table 5.2 tags liquidity and the technology platform as capital and strategic risk as risk.

The supports also fix Table 11.4's lag floors: the distance between two supports along Chapter 9's fourteen arrows is the fewest quarters in which one dimension can move another through the state (Theorem 11.3(i)). Nothing done in Digital can reach Industrial Systems' operations in less than two quarters, and no path leads from Digital to Materials at all. Section 11.8 adds the sustainability line: skipping Meridian's capability rent, about $23 million a year (Worked Example 10.1), would add that much to reported EBITDA while capability, decaying at 2.5 percent a quarter, falls below the floor of 55 within three quarters: performance delivered, and borrowed from a stock.

Python Code
# Table 11.2: the ten dimensions of the performance vector (Equation 11.2)
DIMENSIONS = [("P_F", "financial", "earnings, cash, returns on capital", "margin, ROIC, free cash flow"),
              ("P_O", "operational", "throughput, quality, delivery, cost", "equipment effectiveness, defects"),
              ("P_S", "strategic", "position gained: share, options", "share change, win rate"),
              ("P_H", "human", "capability applied and grown", "output per employee, retention"),
              ("P_T", "technological", "platform uptime and speed of change", "availability, cycle time"),
              ("P_I", "innovation", "validated novelty shipped", "new-product revenue share"),
              ("P_G", "governance", "quality of decisions and controls", "audit findings, incidents"),
              ("P_E", "environmental", "footprint per unit of output", "intensity metrics"),
              ("P_R", "resilience", "shock absorption, shown or ready", "recovery half-life, buffer days"),
              ("P_ESG", "sustainability", "the above, without depleting capital", "Definition 11.11")]
print(f"Table 11.2: the {len(DIMENSIONS)} dimensions of the performance vector")
for sym, name, what, ind in DIMENSIONS:
    print(f"  {sym:6s} {name:15s} {what:38s} {ind}")

# Meridian's dashboard: Table 11.4's six dimensions with their supports (Chapter 9's components), the indicator
# Meridian declares for each (Table 5.2: reading and error) and today's band, the reading plus or minus its error
PLACE = {"TRS": "treasury", "IND": "Industrial Systems", "DIG": "Digital", "MAT": "Materials", "ENG": "engineering"}
SUPPORT = {"P_F": ["TRS"], "P_O": ["IND", "MAT"], "P_S": ["IND", "DIG", "MAT"], "P_T": ["DIG"],
           "P_I": ["DIG", "ENG"], "P_R": ["TRS"]}
INDICATOR = {"P_F": ("ROIC (x6)", 8.3, 0.4, "%"), "P_O": ("operational efficiency (x5)", 71.0, 3.0, ""),
             "P_S": ("market share (x8)", 19.0, 0.8, "%"),
             "P_R": ("liquidity above the $350m floor (x1)", 620.0 - 350.0, 0.01 * 620.0, "$m")}
def show(v, unit, err=False):
    if unit == "%":
        return f"{v:.1f} pp" if err else f"{v:.1f}%"
    return f"${v:.1f}m" if unit == "$m" else f"{v:.0f}"
print("\nMeridian's dashboard: Table 11.4's six dimensions, with the indicators Meridian declares (Table 5.2)")
print(f"  {'':5s} {'support':30s} {'indicator':37s} {'today':>7s} {'error':>10s}   band today")
for dim, sup in SUPPORT.items():
    where = "the three divisions" if len(sup) == 3 else ", ".join(PLACE[c] for c in sup)
    if dim in INDICATOR:
        name, v, err, unit = INDICATOR[dim]
        print(f"  {dim:5s} {where:30s} {name:37s} {show(v, unit):>7s} {'+/-' + show(err, unit, True):>10s}   "
              f"{show(v - err, unit)} to {show(v + err, unit)}")
    else:
        print(f"  {dim:5s} {where:30s} none declared: blind until an indicator reads its support")
TAGS = [("liquidity", "ECA, capital"), ("technology platform", "ECA, capital"),
        ("operational efficiency", "EPA, performance"), ("strategic risk", "ERRA, risk")]
print("the CFO's scorecard reads " + "; ".join(f"{r} ({t})" for r, t in TAGS) + ": one performance reading in four")

# Proposition 11.1 (Online Appendix 11.B): the same financial performance, different resilience
EBITDA = 0.118 * 4200.0                  # 11.8 percent of $4.2 billion
for who, liq in (("Meridian", 620.0), ("a twin", 450.0)):
    after = liq - 150.0
    print(f"  {who:9s} EBITDA ${EBITDA:.0f} million, liquidity ${liq:.0f} million -> ${after:.0f} million after a "
          f"$150 million shock: {'above' if after >= 350 else 'below'} the $350 million floor")

# Table 11.4's lag floors: distances between supports along Chapter 9's fourteen arrows (Theorem 11.3(i))
COMPONENTS = ["MKT", "GOV", "MAT", "IND", "DIG", "ENG", "TRS"]
ARROWS = [("MKT", "IND"), ("MKT", "DIG"), ("MKT", "MAT"), ("IND", "TRS"), ("DIG", "TRS"), ("MAT", "TRS"),
          ("TRS", "IND"), ("TRS", "DIG"), ("ENG", "IND"), ("IND", "ENG"), ("ENG", "DIG"), ("DIG", "ENG"),
          ("GOV", "TRS"), ("GOV", "DIG")]
IX, NEVER = {c: i for i, c in enumerate(COMPONENTS)}, 99
DIST = np.full((7, 7), NEVER); np.fill_diagonal(DIST, 0)
for u, v in ARROWS:
    DIST[IX[u], IX[v]] = 1
for k in range(7):                       # fewest arrows from every component to every other (Floyd-Warshall)
    DIST = np.minimum(DIST, DIST[:, [k]] + DIST[[k], :])
def lag(a, b):
    # the distance from dimension a's support to dimension b's
    return int(min(DIST[IX[u], IX[v]] for u in SUPPORT[a] for v in SUPPORT[b]))
TABLE114 = {("P_F", "P_O"): ("+", 1), ("P_F", "P_S"): ("+", 1), ("P_F", "P_T"): ("+", 1), ("P_F", "P_I"): ("+", 1),
            ("P_F", "P_R"): ("+", 0), ("P_O", "P_F"): ("+", 1), ("P_O", "P_S"): ("+", 0), ("P_O", "P_R"): ("+", 1),
            ("P_S", "P_F"): ("+", 1), ("P_S", "P_I"): ("+", 0), ("P_T", "P_F"): ("+", 1), ("P_T", "P_O"): ("+", 2),
            ("P_T", "P_S"): ("+", 0), ("P_T", "P_I"): ("+", 0), ("P_T", "P_R"): ("+", 1), ("P_I", "P_F"): ("+", 1),
            ("P_I", "P_S"): ("+", 0), ("P_I", "P_T"): ("+", 0), ("P_R", "P_F"): ("-", 0), ("P_R", "P_O"): ("-", 1)}
DIMS6 = list(SUPPORT)
print("\nTable 11.4 from the architecture: declared sign / lag floor in quarters, the distance between the supports")
print("  improving " + "".join(f"{b:>7s}" for b in DIMS6))
for a in DIMS6:
    cells = ["---" if a == b else (f"{TABLE114[a, b][0]}/{lag(a, b)}" if (a, b) in TABLE114 else "") for b in DIMS6]
    print(f"  {a:9s}" + "".join(f"{c:>7s}" for c in cells))
LAG_MATCH = sum(lag(a, b) == q for (a, b), (sgn, q) in TABLE114.items())
DIG_IND, DIG_MAT = int(DIST[IX["DIG"], IX["IND"]]), int(DIST[IX["DIG"], IX["MAT"]])
print(f"{LAG_MATCH} of the {len(TABLE114)} declared lag floors equal the distances between the supports")
print(f"Digital to Industrial Systems: at least {DIG_IND} quarters; Digital to Materials: "
      f"{'no path, never' if DIG_MAT >= NEVER else DIG_MAT}")

# Section 11.8: the sustainability line (the rent of Worked Example 10.1)
DELTA_Q, C_STAR, PRICE_PT = 0.025, 58.0, 4.0     # attrition a quarter; capability held; $ million a point
RENT_YEAR = 4 * DELTA_Q * C_STAR * PRICE_PT
CAP_NO_RENT = C_STAR * (1 - DELTA_Q) ** np.arange(5)
Q_BELOW = int(np.argmax(CAP_NO_RENT < 55.0))
print(f"\nSection 11.8: the capability rent is {DELTA_Q * C_STAR:.2f} points a quarter, ${RENT_YEAR:.1f} million a year;"
      f" skipping it adds that much to reported EBITDA")
print(f"  capability with no rent, quarters 0-4: " + ", ".join(f"{c:.2f}" for c in CAP_NO_RENT)
      + f" -> below the floor of 55 in quarter {Q_BELOW}")
Table 11.2: the 10 dimensions of the performance vector
  P_F    financial       earnings, cash, returns on capital     margin, ROIC, free cash flow
  P_O    operational     throughput, quality, delivery, cost    equipment effectiveness, defects
  P_S    strategic       position gained: share, options        share change, win rate
  P_H    human           capability applied and grown           output per employee, retention
  P_T    technological   platform uptime and speed of change    availability, cycle time
  P_I    innovation      validated novelty shipped              new-product revenue share
  P_G    governance      quality of decisions and controls      audit findings, incidents
  P_E    environmental   footprint per unit of output           intensity metrics
  P_R    resilience      shock absorption, shown or ready       recovery half-life, buffer days
  P_ESG  sustainability  the above, without depleting capital   Definition 11.11

Meridian's dashboard: Table 11.4's six dimensions, with the indicators Meridian declares (Table 5.2)
        support                        indicator                               today      error   band today
  P_F   treasury                       ROIC (x6)                                8.3%  +/-0.4 pp   7.9% to 8.7%
  P_O   Industrial Systems, Materials  operational efficiency (x5)                71       +/-3   68 to 74
  P_S   the three divisions            market share (x8)                       19.0%  +/-0.8 pp   18.2% to 19.8%
  P_T   Digital                        none declared: blind until an indicator reads its support
  P_I   Digital, engineering           none declared: blind until an indicator reads its support
  P_R   treasury                       liquidity above the $350m floor (x1)  $270.0m   +/-$6.2m   $263.8m to $276.2m
the CFO's scorecard reads liquidity (ECA, capital); technology platform (ECA, capital); operational efficiency (EPA, performance); strategic risk (ERRA, risk): one performance reading in four
  Meridian  EBITDA $496 million, liquidity $620 million -> $470 million after a $150 million shock: above the $350 million floor
  a twin    EBITDA $496 million, liquidity $450 million -> $300 million after a $150 million shock: below the $350 million floor

Table 11.4 from the architecture: declared sign / lag floor in quarters, the distance between the supports
  improving     P_F    P_O    P_S    P_T    P_I    P_R
  P_F          ---    +/1    +/1    +/1    +/1    +/0
  P_O          +/1    ---    +/0                  +/1
  P_S          +/1           ---           +/0       
  P_T          +/1    +/2    +/0    ---    +/0    +/1
  P_I          +/1           +/0    +/0    ---       
  P_R          -/0    -/1                         ---
20 of the 20 declared lag floors equal the distances between the supports
Digital to Industrial Systems: at least 2 quarters; Digital to Materials: no path, never

Section 11.8: the capability rent is 1.45 points a quarter, $23.2 million a year; skipping it adds that much to reported EBITDA
  capability with no rent, quarters 0-4: 58.00, 56.55, 55.14, 53.76, 52.41 -> below the floor of 55 in quarter 3

Panel 2 — Check for dominance (Worked Example 11.1, Step 1; Table 11.6: 47.1, 46.1 and 38.4)

Worked Example 11.1, Step 1. At Year 5 the scorecard gives digital 47.1, restructuring 46.1 and the turnaround 38.4 (Online Appendix 11.B.6 lists the readings). No option dominates another: digital leads restructuring by 36.8 points on technology; restructuring leads digital by 26.8 on liquidity, 5.8 on strategic risk and 0.1 on efficiency; the turnaround has the best efficiency of the three. By Theorem 11.2(ii), positive weights exist that put either leader first, so the ranking cannot be read as a finding about the options until its weights are read as a choice.

Python Code
print(f"{'Year 5':14s}" + "".join(f"{r:>16s}" for r in READINGS) + f"{'score':>9s}")
print(f"{'today':14s}" + "".join(f"{v:16.2f}" for v in TODAY) + f"{score(TODAY):9.2f}")
for o, x in zip(OPTIONS, Y5):
    print(f"{o:14s}" + "".join(f"{v:16.2f}" for v in x) + f"{score(x):9.2f}")

def dominates(a, b):
    # at least as good on every reading and better on at least one (strategic risk reversed)
    a, b = np.asarray(a) * SIGN, np.asarray(b) * SIGN
    return bool(np.all(a >= b) and np.any(a > b))
DOM = [(OPTIONS[i], OPTIONS[j]) for i in range(3) for j in range(3) if i != j and dominates(Y5[i], Y5[j])]
print("\npairs in which one option dominates another:", DOM if DOM else "none")
D = (Y5[0] - Y5[1]) * SIGN           # digital less restructuring, strategic risk with its minus sign
print("digital less restructuring, reading by reading (risk reversed):", ", ".join(f"{r} {v:+.2f}" for r, v in zip(READINGS, D)))
print(f"digital leads on technology by {D[1]:.1f}; restructuring leads on liquidity by {-D[0]:.1f}, "
      f"on strategic risk by {-D[3]:.1f}, on efficiency by {-D[2]:.1f}")
print(f"best efficiency: the {OPTIONS[int(np.argmax(Y5[:, 2]))]}, {Y5[:, 2].max():.2f}; efficiency differs between the "
      f"leaders by {abs(D[2]):.2f}, inside the reading's +/-3 error (Table 5.2)")
print(f"digital's lead at equal weights: {D @ EQUAL:.2f} points")
Year 5               liquidity      technology      efficiency  strategic risk    score
today                   100.00           41.00           71.00           64.00    37.00
digital                  94.30           75.63           73.60           55.22    47.08
restructuring           121.11           38.86           73.68           49.39    46.07
turnaround              105.37           35.94           78.60           66.33    38.40

pairs in which one option dominates another: none
digital less restructuring, reading by reading (risk reversed): liquidity -26.81, technology +36.78, efficiency -0.09, strategic risk -5.83
digital leads on technology by 36.8; restructuring leads on liquidity by 26.8, on strategic risk by 5.8, on efficiency by 0.1
best efficiency: the turnaround, 78.60; efficiency differs between the leaders by 0.09, inside the reading's +/-3 error (Table 5.2)
digital's lead at equal weights: 1.01 points

Panel 3 — Sweep the weights (Theorem 11.2(ii); Figure 11.1; Online Appendix 11.B.2: a tie 1.6 points away)

Worked Example 11.1, Step 2, and AXIOM Lab 1.11, step 2. Write d\mathbf{d} for digital's readings less restructuring's, strategic risk with its minus sign: d=(−26.81,36.78,−0.09,−5.83)\mathbf{d} = (-26.81, 36.78, -0.09, -5.83), and w⋅d=1.01\mathbf{w}\cdot\mathbf{d} = 1.01 at equal weights. Moving tt of weight from technology to liquidity changes the lead by −t(dT−dL)-t(d_T - d_L): 0.64 points for each point of weight, so the lead of 1.0 is gone after 1.6 points (26.6 percent on liquidity, 23.4 percent on technology), and with 2 points moved restructuring leads, 47.7 to 47.5. Moving weight from technology ties the options after 2.4 points to strategic risk or 2.7 to efficiency; since any change of weights lowers the lead by at most 12∑k∣Δwk∣ (max⁡kdk−min⁡kdk)\tfrac12\sum_k\lvert\Delta w_k\rvert\,(\max_k d_k - \min_k d_k), no smaller change can produce a tie (Online Appendix 11.B.2). Drawing 200,000 weightings uniformly from the simplex of nonnegative weights summing to one, with seed 26111 as in the book, digital ranks first under 49.4 percent, restructuring under 49.3 percent and the turnaround under 1.3 percent, and only with at least 68 percent on efficiency. That count measures how much of the space of weights each option wins; it is not a probability that either option is better.

Python Code
LEAD = float(D @ EQUAL)
PER_POINT = (D[1] - D[0]) / 100                    # lead lost per point of weight moved from technology to liquidity
T_TIE = LEAD / (D[1] - D[0])                       # weight moved at the tie
def moved(t, frm=1, to=0):
    # equal weights with t moved from one reading to another (default: technology to liquidity)
    w = EQUAL.copy(); w[frm] -= t; w[to] += t
    return w
print(f"lead {LEAD:.4f}; each point of weight moved from technology to liquidity cuts it by {PER_POINT:.4f}")
print(f"tie after {100 * T_TIE:.2f} points: {100 * (0.25 + T_TIE):.1f}% on liquidity, {100 * (0.25 - T_TIE):.1f}% on technology")
print(f"{'points moved':>13s}" + "".join(f"{o:>15s}" for o in OPTIONS) + "   first")
for pts in (0.0, 1.0, 100 * T_TIE, 2.0, 5.0):
    sc = score(Y5, moved(pts / 100))
    print(f"{pts:13.2f}" + "".join(f"{v:15.2f}" for v in sc) + f"   {OPTIONS[int(np.argmax(sc))] if pts != 100 * T_TIE else 'a tie'}")
MOVES = {READINGS[j]: LEAD / (D[1] - D[j]) for j in (0, 2, 3)}
print("ties when weight moves from technology to: " + ", ".join(f"{r} {100 * t:.2f} points" for r, t in MOVES.items()))
print(f"any weight change lowers the lead by at most (1/2) sum|dw| x (max d - min d) = (1/2) sum|dw| x {D.max() - D.min():.2f};"
      f" moving t from technology to liquidity uses all of it, so {100 * T_TIE:.2f} points is the smallest tie")

rng = np.random.default_rng(SEED)                  # seed 26111, as in Online Appendix 11.B.6
WD = rng.dirichlet(np.ones(4), size=200_000)       # weightings uniform on the simplex
FIRST = np.argmax((Y5 * SIGN) @ WD.T, axis=0)
SHARES = np.bincount(FIRST, minlength=3) / len(FIRST)
TURN = WD[FIRST == 2]
print("\nshare of 200,000 weightings ranking each option first (seed 26111): "
      + ", ".join(f"{o} {100 * s:.1f}%" for o, s in zip(OPTIONS, SHARES)))
print(f"the turnaround's weightings: at least {100 * TURN[:, 2].min():.1f}% on efficiency, at most "
      f"{100 * TURN[:, 1].max():.1f}% on technology")

# Figure 11.1's slice: weight on technology and on liquidity; efficiency and strategic risk share the rest equally
g = np.linspace(0, 1, 601)
WT, WL = np.meshgrid(g, g)
REST = (1 - WT - WL) / 2
WIN = np.argmax(np.einsum("oj,abj->abo", Y5 * SIGN, np.stack([WL, WT, REST, REST], axis=-1)), axis=-1).astype(float)
WIN[REST < -1e-12] = np.nan
print(f"on the slice the turnaround ranks first at {int(np.nansum(WIN == 2))} of {int(np.sum(~np.isnan(WIN)))} grid points")
fig, ax = plt.subplots(figsize=(7.6, 3.6))
ax.pcolormesh(g, g, np.ma.masked_invalid(WIN), cmap=ListedColormap(["#DCE6F0", "#FBE3D3", "#F3D6D6"]), vmin=-0.5,
              vmax=2.5, shading="nearest", rasterized=True)
c = (D[2] + D[3]) / 2                              # the tie on the slice: (D_T - c) w_T + (D_L - c) w_L + c = 0
wt = np.linspace(0, 1, 801); wl = -(c + (D[1] - c) * wt) / (D[0] - c)
keep = (wl >= 0) & (wl + wt <= 1)
ax.plot(wt[keep], wl[keep], color="#616161", lw=1)
ax.plot([0, 1], [1, 0], color="#999999", lw=0.6)
ax.plot(0.25, 0.25, "o", color=COLORS[0], ms=5, zorder=5)
ax.plot(0.25 - T_TIE, 0.25 + T_TIE, "o", mfc="white", mec="#616161", ms=5, zorder=5)
ax.annotate("CFO's equal weights:\ndigital 47.1, restructuring 46.1", xy=(0.25, 0.25), xytext=(0.30, 0.07),
            fontsize=8, color=COLORS[0], arrowprops=dict(arrowstyle="-", color=COLORS[0], lw=0.6))
ax.annotate(f"tie: {100 * T_TIE:.1f} points moved\nfrom technology to liquidity", xy=(0.25 - T_TIE, 0.25 + T_TIE),
            xytext=(0.02, 0.50), fontsize=8, color="#616161", arrowprops=dict(arrowstyle="-", color="#616161", lw=0.6))
ax.text(0.37, 0.80, "weights would sum beyond 1", fontsize=8, color="#999999")
ax.text(0.615, 0.66, "efficiency and strategic risk\nshare the rest of the weight\nequally on this slice", fontsize=8,
        va="top", color="#616161", clip_on=False)
ax.text(0.615, 0.36, "all weightings (seed 26111):\n" + "\n".join(f"{o} first {100 * s:.1f}%" for o, s in zip(OPTIONS, SHARES)),
        fontsize=8, va="top", clip_on=False)
ax.legend(handles=[Patch(facecolor="#DCE6F0", edgecolor=COLORS[0], label="digital first"),
                   Patch(facecolor="#FBE3D3", edgecolor=COLORS[1], label="restructuring first")],
          frameon=False, fontsize=8, loc="upper left", bbox_to_anchor=(1.0, 1.0))
ax.set(xlim=(0, 0.6), ylim=(0, 1), xlabel="weight on technology (share of the total)",
       ylabel="weight on liquidity (share of the total)", title="Who ranks first is a choice of weights (seed 26111)")
plt.tight_layout(); plt.show()
lead 1.0126; each point of weight moved from technology to liquidity cuts it by 0.6359
tie after 1.59 points: 26.6% on liquidity, 23.4% on technology
 points moved        digital  restructuring     turnaround   first
         0.00          47.08          46.07          38.40   digital
         1.00          47.26          46.89          39.09   digital
         1.59          47.37          47.37          39.50   a tie
         2.00          47.45          47.71          39.78   restructuring
         5.00          48.01          50.18          41.87   restructuring
ties when weight moves from technology to: liquidity 1.59 points, efficiency 2.75 points, strategic risk 2.38 points
any weight change lowers the lead by at most (1/2) sum|dw| x (max d - min d) = (1/2) sum|dw| x 63.59; moving t from technology to liquidity uses all of it, so 1.59 points is the smallest tie

share of 200,000 weightings ranking each option first (seed 26111): digital 49.4%, restructuring 49.3%, turnaround 1.3%
the turnaround's weightings: at least 68.4% on efficiency, at most 13.1% on technology
on the slice the turnaround ranks first at 0 of 180901 grid points
Who ranks first is a choice of weights (seed 26111)

Panel 4 — Change the yardstick (Theorem 11.2(iii); Online Appendix 11.B.3: 52.3 to 41.4, then 24.0 to 10.6)

Worked Example 11.1, Step 3, and AXIOM Lab 1.11, step 3. The scorecard adds the readings in their own index points, and that is itself a choice. Rescale each reading so that the best option scores 100 and the worst 0, keep the equal weights, and restructuring wins, 52.3 to 41.4: the rescaling divides each reading by its spread across the options (26.8 for liquidity, 39.7 for technology, 5.0 for efficiency, 16.9 for strategic risk), so the 5.8-point difference in strategic risk becomes 34 points, while the 36.8-point technology lead becomes 93 and the 26.8-point liquidity deficit becomes 100. Score by percentage improvement on today instead and digital wins, 24.0 to 10.6, because technology starts lowest and its 84 percent gain outweighs every other change. A normalization pi↦(pi−mi)/rip_i \mapsto (p_i - m_i)/r_i turns the declared weights wiw_i into effective weights wi/riw_i/r_i: as shares, 12, 8, 62 and 18 percent on the first scale and 16, 38, 22 and 24 percent on the second (Online Appendix 11.B.3). The "equal" weights are equal on neither. Same data, same weights, opposite rankings.

The lab also switches the reference group, the set of options or peers the scale is drawn from. The last table below rescales against two other groups, computed from the same Year-5 readings: the options with today's company, and the two leaders alone.

Python Code
def best_to_worst(X, ref=None):
    # each reading rescaled so the best of the reference group scores 100 and the worst 0 (strategic risk reversed)
    V = np.atleast_2d(X) * SIGN
    R = V if ref is None else np.atleast_2d(ref) * SIGN
    lo, hi = R.min(axis=0), R.max(axis=0)
    return 100 * (V - lo) / (hi - lo)
def pct_on_today(X):
    # percentage improvement on today's reading (strategic risk reversed)
    return 100 * (np.atleast_2d(X) - TODAY) / TODAY * SIGN
def effective(r, w=EQUAL):
    # effective weights, as shares, of a normalization that divides reading i by r_i
    e = w / r
    return e / e.sum()
SPREAD = (Y5 * SIGN).max(axis=0) - (Y5 * SIGN).min(axis=0)
B2W, PCT = best_to_worst(Y5) @ EQUAL, pct_on_today(Y5) @ EQUAL
print(f"{'yardstick, equal weights':34s}" + "".join(f"{o:>15s}" for o in OPTIONS) + "   first")
for name, sc in (("index points (the CFO's)", score(Y5)), ("best option 100, worst 0", B2W), ("percentage improvement on today", PCT)):
    print(f"{name:34s}" + "".join(f"{v:15.2f}" for v in sc) + f"   {OPTIONS[int(np.argmax(sc))]}")
print("\nspreads across the options: " + ", ".join(f"{r} {v:.2f}" for r, v in zip(READINGS, SPREAD)))
print(f"rescaled differences, digital less restructuring: strategic risk {100 * D[3] / SPREAD[3]:.1f}, "
      f"technology {100 * D[1] / SPREAD[1]:+.1f}, liquidity {100 * D[0] / SPREAD[0]:.1f}, efficiency {100 * D[2] / SPREAD[2]:.1f}")
print(f"technology's gain on today under digital: {pct_on_today(Y5)[0, 1]:.1f}%")
EFF_B2W, EFF_PCT = effective(SPREAD), effective(TODAY / 100)
for name, e in (("best 100, worst 0", EFF_B2W), ("percentage on today", EFF_PCT)):
    print(f"effective weights, {name:20s}: " + ", ".join(f"{r} {100 * v:.1f}%" for r, v in zip(READINGS, e)))
print(f"on the percentage scale digital leads by {PCT[0] - PCT[1]:.1f} points")

# Theorem 11.2(iii)'s own example: weights (1, 1), A = (3, 1), B = (1, 2), rescaled by a = (1, 3)
A_, B_, a_ = np.array([3.0, 1.0]), np.array([1.0, 2.0]), np.array([1.0, 3.0])
print(f"\nTheorem 11.2(iii): A scores {A_.sum():.0f} > {B_.sum():.0f} raw; after the rescaling (1, 3), {a_ @ A_:.0f} < {a_ @ B_:.0f}")

# Two other reference groups (computed from the book's Year-5 readings; not printed in the book)
GROUPS = {"the three options (the book's)": (Y5, [0, 1, 2]),
          "the options and today's company": (np.vstack([Y5, TODAY]), [0, 1, 2]),
          "the two leaders alone": (Y5[:2], [0, 1])}
REF = {}
print(f"\n{'best 100, worst 0, against':34s}" + "".join(f"{o:>15s}" for o in OPTIONS) + "   first   effective weight on efficiency")
for name, (grp, members) in GROUPS.items():
    sc = best_to_worst(Y5[members], grp) @ EQUAL
    full = np.full(3, np.nan); full[members] = sc
    r = (np.atleast_2d(grp) * SIGN).max(axis=0) - (np.atleast_2d(grp) * SIGN).min(axis=0)
    REF[name] = (full, effective(r))
    print(f"{name:34s}" + "".join(f"{v:15.2f}" if not np.isnan(v) else f"{'-':>15s}" for v in full)
          + f"   {OPTIONS[int(np.nanargmax(full))]:13s} {100 * effective(r)[2]:.1f}%")
print(f"against the two leaders, efficiency's {abs(D[2]):.2f}-point gap, inside its +/-3 error, counts as much as the "
      f"{D[1]:.1f}-point technology lead")

fig, ax = plt.subplots(figsize=(7.6, 3.6))
groups = [("index points\n(the CFO's)", score(Y5)), ("best 100, worst 0\n(three options)", B2W),
          ("percentage improvement\non today", PCT), ("best 100, worst 0\n(two leaders alone)", REF["the two leaders alone"][0])]
for j, (lab, sc) in enumerate(groups):
    shown = [i for i in range(3) if not np.isnan(sc[i])]
    for pos, i in enumerate(shown):
        x = j + (pos - (len(shown) - 1) / 2) * 0.26
        ax.bar(x, sc[i], width=0.24, color=COLORS[i], label=OPTIONS[i] if j == 0 else None)
        ax.text(x, sc[i] + 1.2, f"{sc[i]:.1f}", ha="center", va="bottom", fontsize=7)
ax.set_xticks(range(len(groups))); ax.set_xticklabels([lab for lab, _ in groups], fontsize=8)
ax.set(ylim=(0, 85), ylabel="equal-weight score (points of each scale)",
       title="Same data, same weights, opposite rankings (seed 26111)")
ax.legend(frameon=False, fontsize=8, loc="upper left", ncol=3)
plt.tight_layout(); plt.show()
yardstick, equal weights                  digital  restructuring     turnaround   first
index points (the CFO's)                    47.08          46.07          38.40   digital
best option 100, worst 0                    41.40          52.26          35.32   restructuring
percentage improvement on today             24.04          10.62           0.02   digital

spreads across the options: liquidity 26.81, technology 39.69, efficiency 5.00, strategic risk 16.94
rescaled differences, digital less restructuring: strategic risk -34.4, technology +92.7, liquidity -100.0, efficiency -1.7
technology's gain on today under digital: 84.5%
effective weights, best 100, worst 0   : liquidity 11.6%, technology 7.8%, efficiency 62.2%, strategic risk 18.4%
effective weights, percentage on today : liquidity 15.6%, technology 38.1%, efficiency 22.0%, strategic risk 24.4%
on the percentage scale digital leads by 13.4 points

Theorem 11.2(iii): A scores 4 > 3 raw; after the rescaling (1, 3), 6 < 7

best 100, worst 0, against                digital  restructuring     turnaround   first   effective weight on efficiency
the three options (the book's)              41.40          52.26          35.32   restructuring 62.2%
the options and today's company             49.95          60.66          35.32   restructuring 52.0%
the two leaders alone                       25.00          75.00              -   restructuring 98.0%
against the two leaders, efficiency's 0.09-point gap, inside its +/-3 error, counts as much as the 36.8-point technology lead
Same data, same weights, opposite rankings (seed 26111)

Panel 5 — Roll up a ratio (Theorem 11.5(i); Table 11.5; Online Appendix 11.B.5: 81.7 to 82.0 percent)

AXIOM Lab 1.11, step 4. A group ratio is the divisional ratios weighted by their shares of the denominator, R=∑vsvrvR = \sum_v s_v r_v, so between two periods it moves by a rate effect ∑vsv′(rv′−rv)\sum_v s_v'(r_v' - r_v) plus a mix effect ∑v(sv′−sv) rv\sum_v (s_v' - s_v)\,r_v (Theorem 11.5(i)). The theorem's two divisions raise their margins from 20.0 to 22.0 percent and from 15.0 to 16.0 percent while revenue shifts toward the second, and the group margin falls from 18.3 to 18.0 percent: a rate effect of +1.33+1.33 points and a mix effect of −1.67-1.67. The bank of Table 11.5 runs the same arithmetic on a cost–income ratio, where lower is better: retail shrinks on purpose (80 to 78 percent), markets grows (85 to 84 percent), and the group ratio worsens from 81.7 to 82.0 percent, a rate effect of −1.33-1.33 points and a mix effect of +1.67+1.67. The unweighted average of the two divisional ratios falls from 82.5 to 81.0 percent and would report an improvement of 1.5 points (Online Appendix 11.B.5). The safe rule: add numerators and denominators separately, then divide, and show the mix as its own exhibit.

The book declares no divisional margins for Meridian, so the mix bench runs on the theorem's two divisions; for Meridian the trap is not hypothetical, since the digital option shifts revenue toward a growing division whose margins are still being built. Parts (ii) and (iii) close the audit: three targets on a shared capacity of 10 can each fit pairwise and fail together, and a ratio rolls up only with its denominators.

Python Code
def bridge(N0, D0, N1, D1):
    # group ratio before and after, and Theorem 11.5(i)'s rate and mix effects
    N0, D0, N1, D1 = (np.asarray(a, float) for a in (N0, D0, N1, D1))
    r0, r1, s0, s1 = N0 / D0, N1 / D1, D0 / D0.sum(), D1 / D1.sum()
    return {"R0": N0.sum() / D0.sum(), "R1": N1.sum() / D1.sum(), "r0": r0, "r1": r1, "s0": s0, "s1": s1,
            "rate": float(s1 @ (r1 - r0)), "mix": float((s1 - s0) @ r0), "avg0": r0.mean(), "avg1": r1.mean()}
TWO = bridge([40, 15], [200, 100], [22, 32], [100, 200])       # Theorem 11.5(i): profit and revenue of two divisions
BANK = bridge([160, 85], [200, 100], [78, 168], [100, 200])    # Table 11.5: costs over income, retail and markets
for name, b, what in (("Theorem 11.5(i), two divisions", TWO, "margin"), ("Table 11.5, the bank", BANK, "cost-income ratio")):
    print(f"{name}: divisional {what}s " + ", ".join(f"{100 * x:.1f}% -> {100 * y:.1f}%" for x, y in zip(b["r0"], b["r1"]))
          + f"; shares of the denominator " + ", ".join(f"{x:.2f} -> {y:.2f}" for x, y in zip(b["s0"], b["s1"])))
    print(f"  group {100 * b['R0']:.2f}% -> {100 * b['R1']:.2f}% ({100 * (b['R1'] - b['R0']):+.2f} points) = rate effect "
          f"{100 * b['rate']:+.2f} + mix effect {100 * b['mix']:+.2f}; the unweighted average says "
          f"{100 * b['avg0']:.1f}% -> {100 * b['avg1']:.1f}%")

# The mix bench: hold this year's divisional margins and move the share of revenue
SHARE = np.linspace(0, 1, 101)
GROUP_NEW = SHARE * TWO["r1"][0] + (1 - SHARE) * TWO["r1"][1]
GROUP_OLD = SHARE * TWO["r0"][0] + (1 - SHARE) * TWO["r0"][1]
CROSS = (TWO["R0"] - TWO["r1"][1]) / (TWO["r1"][0] - TWO["r1"][1])
print(f"\nmix bench (computed from the theorem's example): at this year's margins the group beats last year's "
      f"{100 * TWO['R0']:.2f}% only while division 1 keeps at least {100 * CROSS:.1f}% of revenue; it keeps "
      f"{100 * TWO['s1'][0]:.1f}%")

# Theorem 11.5(ii): targets on a capacity shared by three units; (iii): ratios roll up only with denominators
K_CAP, TAU = 10, np.array([4, 4, 4])
PAIRS = [int(TAU[i] + TAU[j]) for i in range(3) for j in range(i + 1, 3)]
print(f"shared capacity {K_CAP}: every pair of targets fits ({max(PAIRS)} <= {K_CAP}); all three do not ({TAU.sum()} > {K_CAP})")
c_ = TWO["r0"]
print(f"same divisional ratios {100 * c_[0]:.0f}% and {100 * c_[1]:.0f}%: group {100 * c_.mean():.2f}% with equal revenue, "
      f"{100 * (2 * c_[0] + c_[1]) / 3:.2f}% when division 1's revenue doubles: no certificate without the denominators")

fig, ax = plt.subplots(figsize=(7.6, 3.6))
ax.axvspan(0, 100 * CROSS, color="#F3D6D6", lw=0)
ax.plot(100 * SHARE, 100 * GROUP_OLD, color="#616161", lw=1.4, ls="--", label="last year's margins: 20.0% and 15.0%")
ax.plot(100 * SHARE, 100 * GROUP_NEW, color=COLORS[0], lw=2, label="this year's margins: 22.0% and 16.0%")
ax.axhline(100 * TWO["R0"], color="#999999", lw=0.6)
ax.plot(100 * TWO["s0"][0], 100 * TWO["R0"], "o", color="#616161", ms=6, zorder=5)
ax.plot(100 * TWO["s1"][0], 100 * TWO["R1"], "o", color=COLORS[2], ms=6, zorder=5)
ax.annotate(f"last year: share {100 * TWO['s0'][0]:.0f}%, group {100 * TWO['R0']:.1f}%", xy=(100 * TWO["s0"][0], 100 * TWO["R0"]),
            xytext=(70, 17.0), fontsize=8, color="#616161", arrowprops=dict(arrowstyle="-", color="#616161", lw=0.6))
ax.annotate(f"this year: share {100 * TWO['s1'][0]:.0f}%, group {100 * TWO['R1']:.1f}%", xy=(100 * TWO["s1"][0], 100 * TWO["R1"]),
            xytext=(1.5, 20.2), fontsize=8, color=COLORS[2], arrowprops=dict(arrowstyle="-", color=COLORS[2], lw=0.6))
ax.text(100 * CROSS / 2, 22.35, f"group below last year's {100 * TWO['R0']:.1f}%\nwhile the share is under {100 * CROSS:.1f}%",
        ha="center", va="top", fontsize=8, color=COLORS[2])
ax.set(xlim=(0, 100), ylim=(15, 22.5), xlabel="share of revenue in division 1, the higher-margin division (%)",
       ylabel="group margin (%)", title="Every division improves while the group falls (seed 26111)")
ax.legend(frameon=False, fontsize=8, loc="lower right")
plt.tight_layout(); plt.show()
Theorem 11.5(i), two divisions: divisional margins 20.0% -> 22.0%, 15.0% -> 16.0%; shares of the denominator 0.67 -> 0.33, 0.33 -> 0.67
  group 18.33% -> 18.00% (-0.33 points) = rate effect +1.33 + mix effect -1.67; the unweighted average says 17.5% -> 19.0%
Table 11.5, the bank: divisional cost-income ratios 80.0% -> 78.0%, 85.0% -> 84.0%; shares of the denominator 0.67 -> 0.33, 0.33 -> 0.67
  group 81.67% -> 82.00% (+0.33 points) = rate effect -1.33 + mix effect +1.67; the unweighted average says 82.5% -> 81.0%

mix bench (computed from the theorem's example): at this year's margins the group beats last year's 18.33% only while division 1 keeps at least 38.9% of revenue; it keeps 33.3%
shared capacity 10: every pair of targets fits (8 <= 10); all three do not (12 > 10)
same divisional ratios 20% and 15%: group 17.50% with equal revenue, 18.33% when division 1's revenue doubles: no certificate without the denominators
Every division improves while the group falls (seed 26111)

Panel 6 — Walk the frontier (Theorem 11.4; Proposition 11.2; Figure 11.2: 0.84 months of cover a point)

AXIOM Lab 1.11, step 5, first part. Every option that maximizes a positively weighted index is Pareto optimal (Theorem 11.4(ii)), but the converse fails when programs are indivisible. Of the programs (0,10)(0, 10), (4,4)(4, 4) and (10,0)(10, 0), in points of margin and months of resilience gained, the balanced one is Pareto optimal, yet every positive weighting scores an end program higher, since max⁡(10w1,10w2)≥5(w1+w2)>4(w1+w2)\max(10w_1, 10w_2) \ge 5(w_1 + w_2) > 4(w_1 + w_2): no index ever ranks it first (Theorem 11.4(iii)). Three programs (0,m)(0, m), (a,a)(a, a) and (m,0)(m, 0) leave the balanced one visible exactly when 2a≥m2a \ge m: at (5,5)(5, 5) it ties at equal weights, at (6,6)(6, 6) equal weights rank it first (Online Appendix 11.B.4). On Figure 11.2's smooth frontier of margin against liquidity cover, an operating point inside the frontier can improve both before any trade-off arises (Proposition 11.2); at the frontier point reached from it, the slope prices the trade: one point of margin costs 0.84 months of cover, the exchange rate −w1/w2-w_1/w_2 of the weights that choose it (Theorem 11.4(iv)). Meridian's three options are all Pareto optimal on the scorecard's four readings, and each ranks first under some weights (Panel 3): none of them is invisible.

Python Code
def pareto(P):
    # indices of the points no other point dominates
    P = np.asarray(P, float)
    return [i for i in range(len(P)) if not any(np.all(P[j] >= P[i]) and np.any(P[j] > P[i]) for j in range(len(P)))]
def first_count(P, n=999):
    # how many of n positive weightings (w, 1 - w) rank each program first
    w1 = np.linspace(0.001, 0.999, n)
    top = np.argmax(np.asarray(P, float) @ np.vstack([w1, 1 - w1]), axis=0)
    return np.bincount(top, minlength=len(P))
PROGRAMS = np.array([[0.0, 10.0], [4.0, 4.0], [10.0, 0.0]])          # Theorem 11.4(iii): margin, resilience gain
print("Pareto optimal: " + ", ".join(f"({PROGRAMS[i][0]:.0f}, {PROGRAMS[i][1]:.0f})" for i in pareto(PROGRAMS)))
BAL_FIRST = int(first_count(PROGRAMS)[1])
print(f"of 999 positive weightings, {BAL_FIRST} rank the balanced program first; at equal weights the scores are "
      + ", ".join(f"{v:.0f}" for v in PROGRAMS @ [0.5, 0.5]))
for a in (4.0, 5.0, 6.0):                                             # Online Appendix 11.B.4: visible iff 2a >= m
    P = np.array([[0.0, 10.0], [a, a], [10.0, 0.0]])
    eq = P @ [0.5, 0.5]
    print(f"  balanced program ({a:.0f}, {a:.0f}): 2a {'>=' if 2 * a >= 10 else '<'} 10, equal weights score it {eq[1]:.0f} against "
          f"{eq[0]:.0f} -> {'first' if eq[1] > eq[0] else ('a tie' if eq[1] == eq[0] else 'never first')}")

# Figure 11.2, left: the frontier (m/16)^2 + (r/12)^2 = 1 of EBITDA margin (%) and liquidity cover (months)
MX, RY, P0 = 16.0, 12.0, np.array([9.0, 6.0])
GROW = 1 / np.sqrt((P0[0] / MX) ** 2 + (P0[1] / RY) ** 2)
PF = GROW * P0
SLOPE = -(RY ** 2 * PF[0]) / (MX ** 2 * PF[1])
print(f"\noperating point ({P0[0]:.0f}, {P0[1]:.0f}) is inside: ({P0[0] / MX:.4f})^2 + ({P0[1] / RY:.4f})^2 = "
      f"{(P0[0] / MX) ** 2 + (P0[1] / RY) ** 2:.4f} < 1, so both can improve (Proposition 11.2)")
print(f"expanding both by {GROW:.4f} reaches the frontier at ({PF[0]:.2f}% margin, {PF[1]:.2f} months of cover)")
print(f"slope there {SLOPE:.4f}: one point of margin costs {-SLOPE:.2f} months of cover; the weights that choose this "
      f"point are in the ratio w1/w2 = {-SLOPE:.4f}")
print("\nMeridian's options on the scorecard's four readings, all Pareto optimal: "
      + ", ".join(f"{OPTIONS[i]} (first under {100 * SHARES[i]:.1f}% of weightings)" for i in pareto(Y5 * SIGN)))
Pareto optimal: (0, 10), (4, 4), (10, 0)
of 999 positive weightings, 0 rank the balanced program first; at equal weights the scores are 5, 4, 5
  balanced program (4, 4): 2a < 10, equal weights score it 4 against 5 -> never first
  balanced program (5, 5): 2a >= 10, equal weights score it 5 against 5 -> a tie
  balanced program (6, 6): 2a >= 10, equal weights score it 6 against 5 -> first

operating point (9, 6) is inside: (0.5625)^2 + (0.5000)^2 = 0.5664 < 1, so both can improve (Proposition 11.2)
expanding both by 1.3287 reaches the frontier at (11.96% margin, 7.97 months of cover)
slope there -0.8438: one point of margin costs 0.84 months of cover; the weights that choose this point are in the ratio w1/w2 = 0.8438

Meridian's options on the scorecard's four readings, all Pareto optimal: digital (first under 49.4% of weightings), restructuring (first under 49.3% of weightings), turnaround (first under 1.3% of weightings)

Panel 7 — Move the window (Theorem 11.3(iii); Figure 11.3; Online Appendix 11.B.6: months 11, 29, 33 and 56)

Worked Example 11.1, Step 4, and AXIOM Lab 1.11, step 5, second part. Run the scorecard month by month. Digital's score first falls, to 32.11 at month 11, and does not regain today's 37.0 until month 29: the J-curve of Theorem 11.3(iii), driven by the liquidity trough and the efficiency dip while the platform is built. It passes the turnaround at month 33 and restructuring only at month 56. Annual reviews in Years 1 and 2 rank digital last, those in Years 3 and 4 rank it second; only a five-year window ranks it first, so a board that compares the options at its annual reviews will see digital trail for four years. Restructuring's score jumps by 6.57 points when the sale closes at Year 1; before that, restructuring and the turnaround differ by less than 0.1 point.

The theorem prices the window itself. A program that spends uˉ\bar u in one period, at a cost cuˉc\bar u, and gains Δ\Delta in every period after it shows a net gain over HH periods exactly when (H−1)Δ>cuˉ(H - 1)\Delta > c\bar u (Equation 11.3), so any window shorter than 1+cuˉ/Δ1 + c\bar u/\Delta periods reports a sound program as a loss. When the built state decays, the window lengthens: a program that costs 10 and adds 3 in its first period of benefit pays back over windows longer than 4.3 periods if the benefit lasts, 4.8 if it decays at 10 percent a period and 5.9 at 20 percent; at 30 percent or more it never pays back (Online Appendix 11.B.1).

Python Code
TODAY_SCORE = float(score(TODAY))
LOW = int(np.argmin(S[0]))
BELOW = MONTHS[S[0] < TODAY_SCORE]
BACK = int(BELOW[-1] + 1)
PASS_T, PASS_R = int(np.argmax(S[0] > S[2])), int(np.argmax(S[0] > S[1]))
print(f"digital: low {S[0, LOW]:.2f} at month {LOW}; below today's {TODAY_SCORE:.1f} from month {BELOW[0]} to month "
      f"{BELOW[-1]}; {S[0, BACK]:.2f} at month {BACK}")
print(f"passes the turnaround at month {PASS_T} ({S[0, PASS_T]:.2f} against {S[2, PASS_T]:.2f}) and restructuring at "
      f"month {PASS_R} ({S[0, PASS_R]:.2f} against {S[1, PASS_R]:.2f})")
JUMP = S[1, 12] - S[1, 11]
PRE_GAP = float(np.abs(S[1, :12] - S[2, :12]).max())
print(f"restructuring jumps by {JUMP:.2f} points when the sale closes; before it, restructuring and the turnaround differ "
      f"by at most {PRE_GAP:.3f}")
RANK = 1 + (S > S[0]).sum(axis=0)                  # digital's rank at each month
print(f"\n{'review':>8s}" + "".join(f"{o:>15s}" for o in OPTIONS) + "   first          digital ranks")
for yr in range(1, 6):
    m = 12 * yr
    print(f"{'Year ' + str(yr):>8s}" + "".join(f"{v:15.2f}" for v in S[:, m]) + f"   {OPTIONS[int(np.argmax(S[:, m]))]:14s} {RANK[m]}")
print("\nslide the review date (month): digital against today and against the leader")
for m in (12, 24, 28, 29, 32, 33, 55, 56, 60):
    print(f"  month {m:2d}: digital {S[0, m]:6.2f}, {'a gain' if S[0, m] > TODAY_SCORE else 'a loss'} on today's "
          f"{TODAY_SCORE:.1f}; rank {RANK[m]}; leader {OPTIONS[int(np.argmax(S[:, m]))]} {S[:, m].max():.2f}")

# Theorem 11.3(iii) and Online Appendix 11.B.1: the window a one-off program needs
COST, GAIN = 10.0, 3.0                             # the program costs 10 and adds 3 in its first period of benefit
def payback(delta):
    # the window beyond which the program shows a net gain (Equation 11.3 when delta = 0); None if it never does
    if delta == 0:
        return 1 + COST / GAIN
    x = COST * delta / GAIN
    return None if x >= 1 - 1e-12 else 1 + np.log(1 / (1 - x)) / np.log(1 / (1 - delta))
def cumulative(H, delta):
    # the program's cumulative effect over periods 1 to H: the cost now, then a gain that decays at delta
    return -COST + sum(GAIN * (1 - delta) ** (k - 2) for k in range(2, H + 1))
WINDOWS = {delta: payback(delta) for delta in (0.0, 0.1, 0.2, 0.3)}
print(f"\nOnline Appendix 11.B.1, cost 10 and first gain 3: {'decay':>6s} {'window':>8s}   cumulative effect over H = 1 ... 8")
for delta, h in WINDOWS.items():
    print(f"{'':44s}{delta:6.0%} {('never' if h is None else f'{h:.2f}'):>8s}   "
          + " ".join(f"{cumulative(H, delta):6.2f}" for H in range(1, 9)))

fig, ax = plt.subplots(figsize=(7.6, 3.6))
for m in (12, 24, 36, 48, 60):
    ax.axvline(m, color="#E0E0E0", lw=0.8, zorder=0)
ax.axhline(TODAY_SCORE, color="#999999", lw=0.8, ls="--")
ax.fill_between(MONTHS, S[0], TODAY_SCORE, where=S[0] < TODAY_SCORE, color="#DCE6F0", lw=0)
ax.plot(MONTHS, S[0], color=COLORS[0], lw=2, label="digital")
ax.plot(MONTHS[:12], S[1, :12], color=COLORS[1], lw=2, label="restructuring")
ax.plot(MONTHS[11:13], S[1, 11:13], color=COLORS[1], lw=1, ls=":")
ax.plot(MONTHS[12:], S[1, 12:], color=COLORS[1], lw=2)
ax.plot(MONTHS, S[2], color=COLORS[2], lw=2, label="turnaround")
ax.plot(LOW, S[0, LOW], "o", color=COLORS[0], ms=5)
ax.annotate(f"low {S[0, LOW]:.1f} at month {LOW}; back to\ntoday's {TODAY_SCORE:.1f} at month {BACK}", xy=(LOW, S[0, LOW]),
            xytext=(16, 31.0), fontsize=8, color=COLORS[0], arrowprops=dict(arrowstyle="-", color=COLORS[0], lw=0.6))
for m, other in ((PASS_T, 2), (PASS_R, 1)):
    ax.plot(m, S[0, m], "o", mfc="white", mec=COLORS[0], ms=5, zorder=5)
ax.annotate(f"passes the turnaround\nat month {PASS_T}", xy=(PASS_T, S[0, PASS_T]), xytext=(34, 34.2), fontsize=8,
            color=COLORS[0], arrowprops=dict(arrowstyle="-", color=COLORS[0], lw=0.6))
ax.annotate(f"passes restructuring\nat month {PASS_R}", xy=(PASS_R, S[0, PASS_R]), xytext=(38, 49.3), fontsize=8,
            color=COLORS[0], arrowprops=dict(arrowstyle="-", color=COLORS[0], lw=0.6))
ax.annotate(f"sale closes: +{JUMP:.1f}", xy=(12, 41.5), xytext=(1.5, 44.5), fontsize=8, color=COLORS[1],
            arrowprops=dict(arrowstyle="-", color=COLORS[1], lw=0.6))
ax.text(60, TODAY_SCORE - 0.4, f"today's score {TODAY_SCORE:.1f}", ha="right", va="top", fontsize=8, color="#616161")
ax.set_xticks([0, 12, 24, 36, 48, 60]); ax.set_yticks([30, 35, 40, 45, 50])
ax.set(xlim=(0, 60), ylim=(30, 52), xlabel="months from today (annual reviews at 12, 24, 36, 48 and 60)",
       ylabel="CFO's scorecard (index points)", title="Digital trails restructuring until month 56 (seed 26111)")
ax.legend(frameon=False, fontsize=8, loc="lower right", ncol=3)
plt.tight_layout(); plt.show()
digital: low 32.11 at month 11; below today's 37.0 from month 1 to month 28; 37.19 at month 29
passes the turnaround at month 33 (38.74 against 38.41) and restructuring at month 56 (46.15 against 46.00)
restructuring jumps by 6.57 points when the sale closes; before it, restructuring and the turnaround differ by at most 0.051

  review        digital  restructuring     turnaround   first          digital ranks
  Year 1          32.15          44.46          37.89   restructuring  3
  Year 2          35.27          45.11          38.29   restructuring  3
  Year 3          39.88          45.55          38.43   restructuring  2
  Year 4          43.96          45.86          38.45   restructuring  2
  Year 5          47.08          46.07          38.40   digital        1

slide the review date (month): digital against today and against the leader
  month 12: digital  32.15, a loss on today's 37.0; rank 3; leader restructuring 44.46
  month 24: digital  35.27, a loss on today's 37.0; rank 3; leader restructuring 45.11
  month 28: digital  36.80, a loss on today's 37.0; rank 3; leader restructuring 45.28
  month 29: digital  37.19, a gain on today's 37.0; rank 3; leader restructuring 45.31
  month 32: digital  38.36, a gain on today's 37.0; rank 3; leader restructuring 45.42
  month 33: digital  38.74, a gain on today's 37.0; rank 2; leader restructuring 45.46
  month 55: digital  45.90, a gain on today's 37.0; rank 2; leader restructuring 45.99
  month 56: digital  46.15, a gain on today's 37.0; rank 1; leader digital 46.15
  month 60: digital  47.08, a gain on today's 37.0; rank 1; leader digital 47.08

Online Appendix 11.B.1, cost 10 and first gain 3:  decay   window   cumulative effect over H = 1 ... 8
                                                0%     4.33   -10.00  -7.00  -4.00  -1.00   2.00   5.00   8.00  11.00
                                               10%     4.85   -10.00  -7.00  -4.30  -1.87   0.32   2.29   4.06   5.65
                                               20%     5.92   -10.00  -7.00  -4.60  -2.68  -1.14   0.08   1.07   1.85
                                               30%    never   -10.00  -7.00  -4.90  -3.43  -2.40  -1.68  -1.18  -0.82
Digital trails restructuring until month 56 (seed 26111)

Panel 8 — Declare the apparatus (Worked Example 11.1, Step 5; Tables 11.3 and 11.6; Worked Example 8.1: 24.6 and 3.0 percent)

Worked Example 11.1, Step 5. Table 11.6 collects the readings: the same three options, six declarations, and the first place moving between digital and restructuring. The strategy team's recommendation is therefore not "digital scores 47.1" but a declaration the board can own: the weights, with their 1.6-point margin from a tie; the yardstick and the options or peers it rescales against; the aggregation rule and any gates; and the review window, at least five years for the digital program. Each is a legitimate choice. None is a fact about the options.

The worked example's discussion asks whether the board's risk appetite, at most a 5 percent chance of breaching the liquidity floor, should enter the scorecard as a weight or as a gate that an option must pass before it is scored (Table 11.3). The odds are Worked Example 8.1's, drawn as in the book with random seed 8 (Volume II, Online Appendix 1.B.4) from Chapter 8's uncertainty model: the digital program's quarterly liquidity plan over three years, a deviation that carries 60 percent of each quarter's surprise (standard deviation $50 million) into the next, and funding-window closures about once a decade that remove on average $120 million. The program as planned breaches the $350 million floor in 24.6 percent of 100,000 paths; a standby line that replaces up to $100 million of any closure, together with a phased build, brings the chance to 3.0 percent. A weight would let technology buy back the breach risk at a fixed rate; a gate, the lexicographic rule of Table 11.3, fails the program as planned in its tier and admits it once the levers bring it within the appetite.

Python Code
TABLE116 = [("The CFO's scorecard", score(Y5)),
            ("2 points of weight moved from technology to liquidity", score(Y5, moved(0.02))),
            ("Best option 100, worst 0", B2W),
            ("Percentage improvement on today", PCT),
            ("Review at Year 1", S[:, 12]),
            ("Review at Year 3", S[:, 36])]
print(f"Table 11.6 {'declaration':55s}" + "".join(f"{o:>15s}" for o in OPTIONS) + "   first")
for label, sc in TABLE116:
    print(f"  {label:63s}" + "".join(f"{v:15.1f}" for v in sc) + f"   {OPTIONS[int(np.argmax(sc))]}")
print("first place, row by row: " + ", ".join(OPTIONS[int(np.argmax(sc))] for _, sc in TABLE116))

# Worked Example 8.1's uncertainty model, drawn as in the book with random seed 8 (Volume II, Online Appendix 1.B.4)
BOOK_SEED, N_PATHS, QUARTERS = 8, 100_000, 12
FLOOR, APPETITE = 350.0, 0.05                    # $ million; the board's appetite for a breach
A_E, SD_W = 0.6, 50.0                            # deviation E' = 0.6 E + w, surprises w with sd $50 million
LAM, J_MEAN, J_SD = 0.10, 120.0, 50.0            # closures a year; their size, mean and sd ($ million, gamma)
LINE, DEPTH, DEPTH_PHASED = 100.0, 28.0, 20.0    # standby line per closure; the plan's draw depth, planned and phased

def plan(depth=DEPTH):
    # the digital program's liquidity, quarterly for three years: Chapter 1's index at $6.2 million a point
    t = np.arange(QUARTERS + 1) / 4
    return 6.2 * (112 + (100 - 112) * np.exp(-0.35 * t) - depth * (t / 2) * np.exp(1 - t / 2))

def breach_odds(line=0.0, depth=DEPTH, n=N_PATHS, seed=BOOK_SEED):
    # share of paths below the floor in some quarter; each quarter the n surprises, the n closure counts, then the
    # sizes of that quarter's closures in path order, as in the book's engine
    rng = np.random.default_rng(seed)
    pl, E, J, low = plan(depth), np.zeros(n), np.zeros(n), np.full(n, np.inf)
    shape, scale = (J_MEAN / J_SD) ** 2, J_SD ** 2 / J_MEAN
    for k in range(QUARTERS):
        w = SD_W * rng.standard_normal(n)
        cnt = rng.poisson(LAM / 4, n)
        sizes = rng.gamma(shape, scale, cnt.sum())
        J = J - np.bincount(np.repeat(np.arange(n), cnt), np.maximum(sizes - line, 0.0), minlength=n)
        E = A_E * E + w
        low = np.minimum(low, pl[k + 1] + E + J)
    return float((low < FLOOR).mean())

P_PLAN, P_LEVERS = breach_odds(), breach_odds(line=LINE, depth=DEPTH_PHASED)
print(f"\nliquidity plan: ${plan()[0]:.0f} million today, trough ${plan().min():.0f} million in quarter {plan().argmin()}; "
      f"phased build trough ${plan(DEPTH_PHASED).min():.0f} million")
print(f"{'gate: P(breach) <= 5% (seed 8, 100,000 paths)':50s} {'P(breach)':>10s}   verdict")
for name, p in (("digital program as planned", P_PLAN), ("with the standby line and the phased build", P_LEVERS)):
    print(f"  {name:48s} {100 * p:9.1f}%   {'passes: scored' if p <= APPETITE else 'fails its tier: not scored'}")
Table 11.6 declaration                                                    digital  restructuring     turnaround   first
  The CFO's scorecard                                                       47.1           46.1           38.4   digital
  2 points of weight moved from technology to liquidity                     47.5           47.7           39.8   restructuring
  Best option 100, worst 0                                                  41.4           52.3           35.3   restructuring
  Percentage improvement on today                                           24.0           10.6            0.0   digital
  Review at Year 1                                                          32.2           44.5           37.9   restructuring
  Review at Year 3                                                          39.9           45.6           38.4   restructuring
first place, row by row: digital, restructuring, restructuring, digital, restructuring, restructuring

liquidity plan: $620 million today, trough $482 million in quarter 7; phased build trough $531 million
gate: P(breach) <= 5% (seed 8, 100,000 paths)       P(breach)   verdict
  digital program as planned                            24.6%   fails its tier: not scored
  with the standby line and the phased build             3.0%   passes: scored

Validation — the book's numbers, and agreement with DCT_V1_Ch11_Lab.xlsx

Each checkpoint is a number printed in Chapter 11 (Worked Example 11.1, Tables 11.4–11.6, Figures 11.1–11.3, the theorems' examples, Section 11.8) or in Online Appendix 11.B, with the readings of Online Appendix 11.B.6 and the odds of Worked Example 8.1 that its discussion cites. The workbook recomputes the ones marked workbook with Excel formulas and shows PASS against the same engine values on its Checks sheet.

Python Code
YR = {yr: S[:, 12 * yr] for yr in range(1, 6)}
checks = [
    ("today's score (workbook)", TODAY_SCORE, 37.0, 0.05),
    ("Year 5: digital liquidity (workbook)", Y5[0, 0], 94.30, 0.005),
    ("Year 5: digital technology, Table 1.9 (workbook)", Y5[0, 1], 75.6, 0.05),
    ("Year 5: digital efficiency (workbook)", Y5[0, 2], 73.60, 0.005),
    ("Year 5: digital strategic risk (workbook)", Y5[0, 3], 55.22, 0.005),
    ("Year 5: restructuring liquidity (workbook)", Y5[1, 0], 121.11, 0.005),
    ("Year 5: restructuring technology (workbook)", Y5[1, 1], 38.86, 0.005),
    ("Year 5: restructuring efficiency (workbook)", Y5[1, 2], 73.68, 0.005),
    ("Year 5: restructuring strategic risk (workbook)", Y5[1, 3], 49.39, 0.005),
    ("Year 5: turnaround liquidity (workbook)", Y5[2, 0], 105.37, 0.005),
    ("Year 5: turnaround technology (workbook)", Y5[2, 1], 35.94, 0.005),
    ("Year 5: turnaround efficiency (workbook)", Y5[2, 2], 78.60, 0.005),
    ("Year 5: turnaround strategic risk (workbook)", Y5[2, 3], 66.33, 0.005),
    ("Year 5 score: digital (workbook)", score(Y5[0]), 47.08, 0.005),
    ("Year 5 score: restructuring (workbook)", score(Y5[1]), 46.07, 0.005),
    ("Year 5 score: turnaround (workbook)", score(Y5[2]), 38.40, 0.005),
    ("pairs with one option dominating (workbook)", len(DOM), 0, 0),
    ("digital's technology lead (workbook)", D[1], 36.8, 0.05),
    ("restructuring's liquidity lead (workbook)", -D[0], 26.8, 0.05),
    ("restructuring's strategic-risk lead (workbook)", -D[3], 5.8, 0.05),
    ("restructuring's efficiency lead (workbook)", -D[2], 0.1, 0.05),
    ("digital's lead at equal weights (workbook)", LEAD, 1.01, 0.005),
    ("lead lost per point of weight moved (workbook)", PER_POINT, 0.64, 0.005),
    ("points of weight moved to tie (workbook)", 100 * T_TIE, 1.6, 0.05),
    ("tie: weight on liquidity, % (workbook)", 100 * (0.25 + T_TIE), 26.6, 0.05),
    ("tie: weight on technology, % (workbook)", 100 * (0.25 - T_TIE), 23.4, 0.05),
    ("2 points moved: digital (workbook)", score(Y5[0], moved(0.02)), 47.5, 0.05),
    ("2 points moved: restructuring (workbook)", score(Y5[1], moved(0.02)), 47.7, 0.05),
    ("2 points moved: turnaround (workbook)", score(Y5[2], moved(0.02)), 39.8, 0.05),
    ("tie moving weight to strategic risk, points (workbook)", 100 * MOVES["strategic risk"], 2.4, 0.05),
    ("tie moving weight to efficiency, points (workbook)", 100 * MOVES["efficiency"], 2.7, 0.05),
    ("weightings ranking digital first, % (seed 26111)", 100 * SHARES[0], 49.4, 0.05),
    ("weightings ranking restructuring first, % (seed 26111)", 100 * SHARES[1], 49.3, 0.05),
    ("weightings ranking the turnaround first, % (seed 26111)", 100 * SHARES[2], 1.3, 0.05),
    ("turnaround first: least weight on efficiency, %", 100 * TURN[:, 2].min(), 68, 0.5),
    ("turnaround first: most weight on technology, %", 100 * TURN[:, 1].max(), 13, 0.5),
    ("best 100, worst 0: digital (workbook)", B2W[0], 41.4, 0.05),
    ("best 100, worst 0: restructuring (workbook)", B2W[1], 52.3, 0.05),
    ("best 100, worst 0: turnaround (workbook)", B2W[2], 35.3, 0.05),
    ("spread of liquidity (workbook)", SPREAD[0], 26.8, 0.05),
    ("spread of technology (workbook)", SPREAD[1], 39.7, 0.05),
    ("spread of efficiency (workbook)", SPREAD[2], 5.0, 0.05),
    ("spread of strategic risk (workbook)", SPREAD[3], 16.9, 0.05),
    ("rescaled strategic-risk difference (workbook)", -100 * D[3] / SPREAD[3], 34, 0.5),
    ("rescaled technology lead (workbook)", 100 * D[1] / SPREAD[1], 93, 0.5),
    ("rescaled liquidity deficit (workbook)", -100 * D[0] / SPREAD[0], 100, 0.5),
    ("percentage on today: digital (workbook)", PCT[0], 24.0, 0.05),
    ("percentage on today: restructuring (workbook)", PCT[1], 10.6, 0.05),
    ("percentage on today: turnaround (workbook)", PCT[2], 0.0, 0.05),
    ("technology's gain on today under digital, % (workbook)", pct_on_today(Y5)[0, 1], 84, 0.5),
    ("percentage scale: digital's lead (workbook)", PCT[0] - PCT[1], 13, 0.5),
    ("effective weight, best-worst: liquidity, % (workbook)", 100 * EFF_B2W[0], 12, 0.5),
    ("effective weight, best-worst: technology, % (workbook)", 100 * EFF_B2W[1], 8, 0.5),
    ("effective weight, best-worst: efficiency, % (workbook)", 100 * EFF_B2W[2], 62, 0.5),
    ("effective weight, best-worst: risk, % (workbook)", 100 * EFF_B2W[3], 18, 0.5),
    ("effective weight, percentage: liquidity, % (workbook)", 100 * EFF_PCT[0], 16, 0.5),
    ("effective weight, percentage: technology, % (workbook)", 100 * EFF_PCT[1], 38, 0.5),
    ("effective weight, percentage: efficiency, % (workbook)", 100 * EFF_PCT[2], 22, 0.5),
    ("effective weight, percentage: risk, % (workbook)", 100 * EFF_PCT[3], 24, 0.5),
    ("Theorem 11.2(iii): A after rescaling (workbook)", a_ @ A_, 6, 0),
    ("Theorem 11.2(iii): B after rescaling (workbook)", a_ @ B_, 7, 0),
    ("two divisions: group margin before, % (workbook)", 100 * TWO["R0"], 18.33, 0.005),
    ("two divisions: group margin after, % (workbook)", 100 * TWO["R1"], 18.0, 0.05),
    ("two divisions: rate effect, points (workbook)", 100 * TWO["rate"], 1.33, 0.005),
    ("two divisions: mix effect, points (workbook)", 100 * TWO["mix"], -1.67, 0.005),
    ("bank: group cost-income before, % (workbook)", 100 * BANK["R0"], 81.7, 0.05),
    ("bank: group cost-income after, % (workbook)", 100 * BANK["R1"], 82.0, 0.05),
    ("bank: rate effect, points (workbook)", 100 * BANK["rate"], -1.33, 0.005),
    ("bank: mix effect, points (workbook)", 100 * BANK["mix"], 1.67, 0.005),
    ("bank: unweighted average before, % (workbook)", 100 * BANK["avg0"], 82.5, 0.05),
    ("bank: unweighted average after, % (workbook)", 100 * BANK["avg1"], 81.0, 0.05),
    ("shared capacity: three targets together (workbook)", TAU.sum(), 12, 0),
    ("indivisible: weightings ranking (4, 4) first (workbook)", BAL_FIRST, 0, 0),
    ("frontier: months of cover a margin point (workbook)", -SLOPE, 0.84, 0.005),
    ("digital's low (workbook)", S[0, LOW], 32.11, 0.005),
    ("month of digital's low (workbook)", LOW, 11, 0),
    ("month digital regains today's score (workbook)", BACK, 29, 0),
    ("digital at month 29 (workbook)", S[0, BACK], 37.19, 0.005),
    ("month digital passes the turnaround (workbook)", PASS_T, 33, 0),
    ("digital at month 33 (workbook)", S[0, PASS_T], 38.74, 0.005),
    ("turnaround at month 33 (workbook)", S[2, PASS_T], 38.41, 0.005),
    ("month digital passes restructuring (workbook)", PASS_R, 56, 0),
    ("digital at month 56 (workbook)", S[0, PASS_R], 46.15, 0.005),
    ("Year 1 review: digital (workbook)", YR[1][0], 32.15, 0.005),
    ("Year 1 review: restructuring (workbook)", YR[1][1], 44.46, 0.005),
    ("Year 1 review: turnaround (workbook)", YR[1][2], 37.89, 0.005),
    ("Year 2 review: digital (workbook)", YR[2][0], 35.27, 0.005),
    ("Year 2 review: restructuring (workbook)", YR[2][1], 45.11, 0.005),
    ("Year 2 review: turnaround (workbook)", YR[2][2], 38.29, 0.005),
    ("Year 3 review: digital (workbook)", YR[3][0], 39.88, 0.005),
    ("Year 3 review: restructuring (workbook)", YR[3][1], 45.55, 0.005),
    ("Year 3 review: turnaround (workbook)", YR[3][2], 38.43, 0.005),
    ("Year 4 review: digital (workbook)", YR[4][0], 43.96, 0.005),
    ("Year 4 review: restructuring (workbook)", YR[4][1], 45.86, 0.005),
    ("Year 4 review: turnaround (workbook)", YR[4][2], 38.45, 0.005),
    ("digital's rank, Year 1 review (workbook)", RANK[12], 3, 0),
    ("digital's rank, Year 2 review (workbook)", RANK[24], 3, 0),
    ("digital's rank, Year 3 review (workbook)", RANK[36], 2, 0),
    ("digital's rank, Year 4 review (workbook)", RANK[48], 2, 0),
    ("digital's rank, Year 5 review (workbook)", RANK[60], 1, 0),
    ("restructuring's jump at the sale (workbook)", JUMP, 6.57, 0.005),
    ("payback window, no decay (workbook)", WINDOWS[0.0], 4.3, 0.05),
    ("payback window, 10% decay (workbook)", WINDOWS[0.1], 4.8, 0.05),
    ("payback window, 20% decay (workbook)", WINDOWS[0.2], 5.9, 0.05),
    ("30% decay: windows that pay back (workbook)", 0 if WINDOWS[0.3] is None else 1, 0, 0),
    ("Table 11.4: lag floors reproduced (workbook)", LAG_MATCH, 20, 0),
    ("lag floor, Digital to Industrial Systems (workbook)", DIG_IND, 2, 0),
    ("Section 11.8: capability rent, $ million a year (workbook)", RENT_YEAR, 23, 0.5),
    ("Section 11.8: quarter capability falls below 55 (workbook)", Q_BELOW, 3, 0),
    ("Proposition 11.1: EBITDA, $ million (workbook)", EBITDA, 496, 0.5),
    ("breach, digital as planned, % (seed 8)", 100 * P_PLAN, 24.6, 0.05),
    ("breach, standby line and phased build, % (seed 8)", 100 * P_LEVERS, 3.0, 0.05),
]
ok, width = 0, max(len(name) for name, *_ in checks)
for name, got, want, tol in checks:
    status = "PASS" if abs(got - want) <= tol + 1e-9 else "FAIL"; ok += status == "PASS"
    print(f"{name:{width}s} {got:10.4f}   book {want:9.4f}   {status}")
print(f"\n{ok} of {len(checks)} checkpoints PASS")
today's score (workbook)                                      37.0000   book   37.0000   PASS
Year 5: digital liquidity (workbook)                          94.2956   book   94.3000   PASS
Year 5: digital technology, Table 1.9 (workbook)              75.6347   book   75.6000   PASS
Year 5: digital efficiency (workbook)                         73.5983   book   73.6000   PASS
Year 5: digital strategic risk (workbook)                     55.2177   book   55.2200   PASS
Year 5: restructuring liquidity (workbook)                   121.1075   book  121.1100   PASS
Year 5: restructuring technology (workbook)                   38.8595   book   38.8600   PASS
Year 5: restructuring efficiency (workbook)                   73.6838   book   73.6800   PASS
Year 5: restructuring strategic risk (workbook)               49.3902   book   49.3900   PASS
Year 5: turnaround liquidity (workbook)                      105.3676   book  105.3700   PASS
Year 5: turnaround technology (workbook)                      35.9430   book   35.9400   PASS
Year 5: turnaround efficiency (workbook)                      78.6017   book   78.6000   PASS
Year 5: turnaround strategic risk (workbook)                  66.3306   book   66.3300   PASS
Year 5 score: digital (workbook)                              47.0777   book   47.0800   PASS
Year 5 score: restructuring (workbook)                        46.0652   book   46.0700   PASS
Year 5 score: turnaround (workbook)                           38.3954   book   38.4000   PASS
pairs with one option dominating (workbook)                    0.0000   book    0.0000   PASS
digital's technology lead (workbook)                          36.7752   book   36.8000   PASS
restructuring's liquidity lead (workbook)                     26.8119   book   26.8000   PASS
restructuring's strategic-risk lead (workbook)                 5.8276   book    5.8000   PASS
restructuring's efficiency lead (workbook)                     0.0855   book    0.1000   PASS
digital's lead at equal weights (workbook)                     1.0126   book    1.0100   PASS
lead lost per point of weight moved (workbook)                 0.6359   book    0.6400   PASS
points of weight moved to tie (workbook)                       1.5924   book    1.6000   PASS
tie: weight on liquidity, % (workbook)                        26.5924   book   26.6000   PASS
tie: weight on technology, % (workbook)                       23.4076   book   23.4000   PASS
2 points moved: digital (workbook)                            47.4509   book   47.5000   PASS
2 points moved: restructuring (workbook)                      47.7101   book   47.7000   PASS
2 points moved: turnaround (workbook)                         39.7839   book   39.8000   PASS
tie moving weight to strategic risk, points (workbook)         2.3768   book    2.4000   PASS
tie moving weight to efficiency, points (workbook)             2.7470   book    2.7000   PASS
weightings ranking digital first, % (seed 26111)              49.3745   book   49.4000   PASS
weightings ranking restructuring first, % (seed 26111)        49.2915   book   49.3000   PASS
weightings ranking the turnaround first, % (seed 26111)        1.3340   book    1.3000   PASS
turnaround first: least weight on efficiency, %               68.3717   book   68.0000   PASS
turnaround first: most weight on technology, %                13.1094   book   13.0000   PASS
best 100, worst 0: digital (workbook)                         41.3999   book   41.4000   PASS
best 100, worst 0: restructuring (workbook)                   52.2640   book   52.3000   PASS
best 100, worst 0: turnaround (workbook)                      35.3238   book   35.3000   PASS
spread of liquidity (workbook)                                26.8119   book   26.8000   PASS
spread of technology (workbook)                               39.6916   book   39.7000   PASS
spread of efficiency (workbook)                                5.0034   book    5.0000   PASS
spread of strategic risk (workbook)                           16.9404   book   16.9000   PASS
rescaled strategic-risk difference (workbook)                 34.4003   book   34.0000   PASS
rescaled technology lead (workbook)                           92.6522   book   93.0000   PASS
rescaled liquidity deficit (workbook)                        100.0000   book  100.0000   PASS
percentage on today: digital (workbook)                       24.0381   book   24.0000   PASS
percentage on today: restructuring (workbook)                 10.6237   book   10.6000   PASS
percentage on today: turnaround (workbook)                     0.0246   book    0.0000   PASS
technology's gain on today under digital, % (workbook)        84.4748   book   84.0000   PASS
percentage scale: digital's lead (workbook)                   13.4144   book   13.0000   PASS
effective weight, best-worst: liquidity, % (workbook)         11.6050   book   12.0000   PASS
effective weight, best-worst: technology, % (workbook)         7.8392   book    8.0000   PASS
effective weight, best-worst: efficiency, % (workbook)        62.1884   book   62.0000   PASS
effective weight, best-worst: risk, % (workbook)              18.3674   book   18.0000   PASS
effective weight, percentage: liquidity, % (workbook)         15.6007   book   16.0000   PASS
effective weight, percentage: technology, % (workbook)        38.0505   book   38.0000   PASS
effective weight, percentage: efficiency, % (workbook)        21.9728   book   22.0000   PASS
effective weight, percentage: risk, % (workbook)              24.3761   book   24.0000   PASS
Theorem 11.2(iii): A after rescaling (workbook)                6.0000   book    6.0000   PASS
Theorem 11.2(iii): B after rescaling (workbook)                7.0000   book    7.0000   PASS
two divisions: group margin before, % (workbook)              18.3333   book   18.3300   PASS
two divisions: group margin after, % (workbook)               18.0000   book   18.0000   PASS
two divisions: rate effect, points (workbook)                  1.3333   book    1.3300   PASS
two divisions: mix effect, points (workbook)                  -1.6667   book   -1.6700   PASS
bank: group cost-income before, % (workbook)                  81.6667   book   81.7000   PASS
bank: group cost-income after, % (workbook)                   82.0000   book   82.0000   PASS
bank: rate effect, points (workbook)                          -1.3333   book   -1.3300   PASS
bank: mix effect, points (workbook)                            1.6667   book    1.6700   PASS
bank: unweighted average before, % (workbook)                 82.5000   book   82.5000   PASS
bank: unweighted average after, % (workbook)                  81.0000   book   81.0000   PASS
shared capacity: three targets together (workbook)            12.0000   book   12.0000   PASS
indivisible: weightings ranking (4, 4) first (workbook)        0.0000   book    0.0000   PASS
frontier: months of cover a margin point (workbook)            0.8438   book    0.8400   PASS
digital's low (workbook)                                      32.1065   book   32.1100   PASS
month of digital's low (workbook)                             11.0000   book   11.0000   PASS
month digital regains today's score (workbook)                29.0000   book   29.0000   PASS
digital at month 29 (workbook)                                37.1874   book   37.1900   PASS
month digital passes the turnaround (workbook)                33.0000   book   33.0000   PASS
digital at month 33 (workbook)                                38.7427   book   38.7400   PASS
turnaround at month 33 (workbook)                             38.4133   book   38.4100   PASS
month digital passes restructuring (workbook)                 56.0000   book   56.0000   PASS
digital at month 56 (workbook)                                46.1454   book   46.1500   PASS
Year 1 review: digital (workbook)                             32.1512   book   32.1500   PASS
Year 1 review: restructuring (workbook)                       44.4557   book   44.4600   PASS
Year 1 review: turnaround (workbook)                          37.8890   book   37.8900   PASS
Year 2 review: digital (workbook)                             35.2745   book   35.2700   PASS
Year 2 review: restructuring (workbook)                       45.1080   book   45.1100   PASS
Year 2 review: turnaround (workbook)                          38.2900   book   38.2900   PASS
Year 3 review: digital (workbook)                             39.8837   book   39.8800   PASS
Year 3 review: restructuring (workbook)                       45.5534   book   45.5500   PASS
Year 3 review: turnaround (workbook)                          38.4331   book   38.4300   PASS
Year 4 review: digital (workbook)                             43.9612   book   43.9600   PASS
Year 4 review: restructuring (workbook)                       45.8575   book   45.8600   PASS
Year 4 review: turnaround (workbook)                          38.4452   book   38.4500   PASS
digital's rank, Year 1 review (workbook)                       3.0000   book    3.0000   PASS
digital's rank, Year 2 review (workbook)                       3.0000   book    3.0000   PASS
digital's rank, Year 3 review (workbook)                       2.0000   book    2.0000   PASS
digital's rank, Year 4 review (workbook)                       2.0000   book    2.0000   PASS
digital's rank, Year 5 review (workbook)                       1.0000   book    1.0000   PASS
restructuring's jump at the sale (workbook)                    6.5664   book    6.5700   PASS
payback window, no decay (workbook)                            4.3333   book    4.3000   PASS
payback window, 10% decay (workbook)                           4.8484   book    4.8000   PASS
payback window, 20% decay (workbook)                           5.9233   book    5.9000   PASS
30% decay: windows that pay back (workbook)                    0.0000   book    0.0000   PASS
Table 11.4: lag floors reproduced (workbook)                  20.0000   book   20.0000   PASS
lag floor, Digital to Industrial Systems (workbook)            2.0000   book    2.0000   PASS
Section 11.8: capability rent, $ million a year (workbook)    23.2000   book   23.0000   PASS
Section 11.8: quarter capability falls below 55 (workbook)     3.0000   book    3.0000   PASS
Proposition 11.1: EBITDA, $ million (workbook)               495.6000   book  496.0000   PASS
breach, digital as planned, % (seed 8)                        24.6220   book   24.6000   PASS
breach, standby line and phased build, % (seed 8)              3.0030   book    3.0000   PASS

112 of 112 checkpoints PASS

Next. Chapter 12 puts this chapter's targets, floors and frontiers together with Chapter 8's distributions and calls the result risk: the exposure of declared objectives to the distribution of the company's state, with vulnerability, robustness and resilience kept apart. A target without a distribution is not yet a statement about risk. AXIOM-11 runs the same model interactively (Lab 1.11); 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