Operating Model & Org Design

Branch: Implementation, Operating Models & Change (Consulting galaxy) · Cluster CN7 (cn-implementation) · Node CN7.01 · ~10 focused hours · Understanding gate ≥85% + mapped Case Lab

Almost every organization that has ever felt slow, confused, or politically exhausting has, at some point, tried to fix itself by redrawing a picture. A new CEO arrives, or a board grows impatient, or a strategy shifts, and within a few months a memo goes out with a new org chart attached: new boxes, new dotted lines, a title elevated here, a function merged there, a town hall where everyone claps. Six months later, the same decisions take the same number of weeks, the same people are still waiting on the same other people, and the same complaint ("nothing ever actually changes here") resurfaces. This is not bad luck. An org chart is a drawing of who reports to whom, and almost nothing about how an organization actually behaves is determined by who reports to whom. What determines behavior is who has the authority to decide, how work actually moves from one person to the next, what people are genuinely capable of doing, and what gets measured and rewarded when they do it. Redraw the picture and leave those four things untouched, and you have redecorated the building, not renovated it. This module exists to teach you the difference, by name, and to make sure you are never the person standing in the after photo wondering why nothing feels different.

Here is the whole module compressed into one sentence, worth holding through the ten hours ahead: an operating model is the joint design of five things (structure, process, governance and decision rights, capabilities, and metrics) and changing only the first one while leaving the other four exactly as they were is the single most common, most expensive, and most avoidable failure in organizational design. This module gives that failure its name (the boxes-and-lines trap) because a name is what lets you catch it in a client's proposed reorg before you burn a quarter proving it does not work. The five levers are not exotic: a defensible span of control, telling real work apart from mere relaying; a functional, divisional, or matrix structure chosen on purpose rather than copied from a competitor; an end-to-end process map that finds the handoffs quietly costing days; a RACI matrix and a delegation-of-authority (DOA) table assigning exactly one accountable owner to every decision instead of the four or five it typically accumulates by accident; a capability audit that tells you honestly whether the people about to receive new authority can use it; and a metrics cascade measuring what you actually want, paired carefully enough that nobody wins by gaming it. None of these five is difficult in isolation; what separates a real operating-model designer from someone good at PowerPoint is doing all five together, in the right order, and refusing to declare victory after doing only the first.

Two synthetic organizations carry this module's full worked redesign from diagnosis to fix: Nandavan Retail, a 180-store Indian multi-category retail chain whose regional pricing decisions take six weeks to approve (long enough to miss the three-week festival windows that matter most) and whose leadership already tried, eighteen months ago, to fix this with exactly the redecoration this module warns against; and Cascade Mutual Insurance, a synthetic US regional property-and-casualty insurer whose claims process is slow for reasons that turn out to be decision-rights problems, not people problems, and whose first attempt at a fix teaches a second, quieter trap: a metric that improves for the wrong reason. Both companies are built from the ground up in this module, with every figure either stated as a plain assumption or derived from one, so that by the end you can rebuild every number yourself rather than having memorized someone else's.

App link. The span-of-control math, the RACI-and-DOA construction, and the capability/metrics cascade you practice here are the source material for this branch's mapped Case-Engine entry: CC-OPS-02 (Lakeview Health OR Throughput, interviewer-led, US, synthetic: a hospital operating-room utilization problem of bottlenecks, scheduling, and capacity), authored in a later wave of this same workstream. Until it ships, the Applied mini-project below is this module's hands-on equivalent, built to the same rubric.

One honest note, stated fully in CN0.01 and worth a single sentence here: a partner's instinct for exactly how much redesign a specific, political, human organization can absorb in one go compounds from dozens of real engagements this module cannot manufacture. What it can teach (to a real, defensible, advisory-grade bar) is the teachable floor beneath that instinct: the five levers, the tools that make each one concrete, and the trap that catches everyone who tries to skip straight to the chart. That floor is what the gate at the end of this module actually tests.

Learning objectives

By the end you can:

1. Define an operating model, from first principles, as the joint design of five levers (structure, process, governance and decision rights, capabilities, and metrics) and state precisely why redesigning structure alone (the boxes-and-lines trap) changes nothing real, naming the mechanism, not just the symptom.
2. Diagnose and redesign organizational structure using span-of-control and layer-count arithmetic, recognize a "ghost layer" (a layer whose average span is too thin to justify its own coordination overhead), and make a defensible centralize-versus-decentralize call using a stated test.
3. Choose, on purpose, among functional, divisional, and matrix structure archetypes for a given strategy and coordination need, naming each archetype's core trade-off rather than selecting one by imitation or fashion.
4. Map a core end-to-end business process, identify handoff and queueing waste, and redesign it with a quantified cycle-time improvement you can defend line by line.
5. Build a correct RACI matrix and a delegation-of-authority (DOA) table for a real decision, diagnosing and repairing the "too many Accountables, no real Responsible" failure pattern, and design the governance cadence (the standing forums) that keeps decision rights current as the organization grows.
6. Run a capability audit (a skills heatmap against the authority you are about to delegate) and make a defensible train-versus-hire-versus-redeploy call, correctly distinguished from CN6.02's deal-scale build-buy-partner decision.
7. Build a metrics cascade linking a strategic objective down to team-level measures, and name and defend against the vanity-metric trap and Goodhart's law (a single unpaired metric invites gaming the measure instead of the outcome it stands for).
8. Run a full worked operating-model redesign end to end (structure, governance/RACI, capabilities, metrics) on a single organization, correctly sequenced, and hand off the rollout/change-management journey to CN7.02.
9. (Productivity objective: R10 duality.) Use AI to draft first-pass RACI matrices, structure-option memos, and KPI-tree scaffolds, while owning every decision-rights judgment yourself and verifying every borrowed structural fact against a primary source before it reaches a client.

The duality, stated once (R10). Objectives 1–8 are the understanding objective the mastery gate rewards, you earn this node by designing a defensible operating model by hand, through a case or drill, with no tool doing the judgment for you. Objective 9 is the productivity payoff you keep afterward: AI can draft a serviceable first-pass RACI skeleton or a KPI-tree template in seconds; knowing exactly which parts, every decision-rights assignment, every DOA threshold, every capability call, remain yours to make and defend is the point.

Prerequisites & connections

Builds on. CN3.02 (Corporate, Growth & Customer Frameworks) already taught the McKinsey 7-S framework as a diagnostic lens (Strategy, Structure, Systems, Staff, Skills, Style, Shared Values) and its own practice set already showed you the symptom this module names: an "agile" strategy announced while the chart, promotion criteria, and management behavior all stay frozen (CN3.02 P10). This module does not re-teach 7-S's diagnostic categories; it teaches the construction toolkit that turns a 7-S diagnosis into an actually-built operating model: span-of-control arithmetic, a real RACI and DOA table, process redesign, a capability audit, and a metrics cascade, the mechanics 7-S names but does not itself supply. CN0.02 (The Engagement) taught delegation at the individual-task level, matching one task to one team member, with a success criterion and a check-in cadence. This module operates one level up: decision rights across a permanent organizational structure, who is accountable for a whole category of recurring decisions, not who does one task this week. The two are cousins, never the same tool: CN0.02's formula tells a Principal how to hand a market-sizing model to a Consultant; this module's RACI tells an entire company who owns pricing authority for the next three years. CN6.02 (Portfolio, Build-Buy-Partner, PMI & Divestitures) already previewed this module's toolkit under a deal clock: its Integration Management Office and 100-day PMI plan are a specific, time-boxed application of the same governance-and-capability machinery this module teaches in general form. CN6.02's build-buy-partner framework operates at the scale of acquiring an entire capability through a transaction; this module's capability lever operates at the scale of developing or reassigning the people you already employ, organic, not transactional, and Worked Example 4 states the distinction with numbers. From the Finance galaxy, the M5.01–M5.10 sector playbooks own each industry's specific operating rhythms (a bank's three-lines-of-defense, a retailer's merchandising calendar) in the depth only a specialist needs; this module teaches the generic method that adapts to any of them, the same relationship CN4.01 established for industry-reading. M4.01–M4.02 supply the "what the organization must execute" input a structure design has to serve: assumed already decided here.

Feeds forward. CN7.02 (Transformation Roadmaps, Governance & Leading the Client) is this module's direct and necessary sequel, and the boundary between the two is load-bearing: this module designs the target state, what structure, decision rights, capabilities, and metrics should be once a redesign is finished. CN7.02 designs the journey to get there, sequencing initiatives, cascading KPIs through a transformation roadmap, standing up program governance, and running the working sessions and steering committees (agenda design, pre-wiring, facilitating to a decision) that carry an organization from today's operating model to the one this module teaches you to design. Kotter-style change management, multi-month stakeholder communication, and SteerCo mechanics all belong to CN7.02, not here. CN7.03 (the Partner Capstone) may hand you a case with an operating-model thread woven in, and this module is where you built the muscle for its design half. The hand-off into a client-ready deliverable belongs to M10.03 and the future Presentation galaxy, this module teaches the content of a defensible redesign, never the deck craft that presents it.

The one-sentence version of this module. An operating model is not a picture of who reports to whom; it is five interlocking decisions (structure, process, governance and decision rights, capabilities, and metrics) and an organization that changes only the picture, while leaving the other four exactly as they were, has spent a budget and a quarter changing nothing real at all.

Core teaching content

Process, lever two: mapping how work actually flows

A process is the actual, observable sequence of steps a piece of work travels through from trigger to completion, a customer order from placement to cash collected, a claim from first notice of loss to payment. Structure tells you who sits where on the chart; process tells you what actually happens to work as it moves between those boxes, and the two frequently disagree, a beautifully designed structure can still run a slow, waste-riddled process across its own boundaries. Mapping a process for redesign means listing its real steps in order, the time each actually takes (not the time it is supposed to take), and (the step most redesigns skip) every handoff, because handoffs are disproportionately where waste accumulates: a queue waiting for attention, a duplicate re-entry of information a previous step already captured, an approval reviewing the same fact a different approver just reviewed.

Two waste patterns recur often enough to name on sight. Queueing waste is time work spends waiting for attention rather than being actively worked on, and it is often the largest component of total cycle time precisely because it is invisible on an org chart; Worked Example 5 finds and fixes exactly this. Duplicative or redundant-review waste is a step re-examining a fact a prior step already resolved, usually surviving because "we've always done it that way," not because it still catches real errors; Worked Example 1's frozen five-step approval and Worked Example 5's blanket cross-line review are the same pattern in different clothes. Redesigning a process well means removing steps that add time without adding a decision, automating mechanical steps, and making sure any step removed was not secretly carrying a decision right nobody has explicitly reassigned, that reassignment is governance's job, taken up next.

Capabilities, lever four: the audit and the train-hire-redeploy call

Redesigning structure and governance to hand a manager real authority accomplishes nothing if that manager cannot actually exercise it well, the fourth lever, capabilities, is the honest check on whether the people about to be empowered are ready, and if not, what to do before, not after, authority changes hands. A usable capability audit is a small, specific heatmap: name the two or three skills the new authority actually requires (margin literacy, local competitive judgment, system proficiency, not generic "leadership"), rate each relevant person against each skill, and let the pattern tell you where the real gaps sit.

Once gaps are identified, the choice is train, hire, or redeploy, decided by the same kind of explicit trade-off every lever here runs on: the size of the gap (a light refresh is a training problem; a fundamentally different skill set is usually a hiring problem), the value already in the room (an experienced manager's local knowledge and relationships are real assets a replacement would take years to rebuild, and a cost comparison ignoring this systematically over-favors hiring), and the plain arithmetic of training versus replacement cost. Worked Example 4 runs this comparison with real numbers. This decision is easily confused with CN6.02's build-buy-partner framework: hold the distinction sharply: CN6.02 asks whether to acquire an entire new capability the organization lacks anywhere, at deal scale; this module asks whether to develop, redeploy, or replace the specific people already on the team who need a skill for authority the redesign just gave them. Confusing the two produces exactly the mismatch Common Mistake 5 names, training a team for authority nobody delegated, or delegating authority to a team nobody checked was ready.

Culture and incentives: the signal, and the sixth trap

A fair question at this point is where culture (the shared, often unstated norms about how people actually behave) fits among the five levers just taught. The honest answer: culture is not a sixth lever you design directly; it is the emergent signal that tells you whether the other five are pulling together. An organization whose structure grants real authority, whose processes are efficient, whose RACI has one clear owner per decision, whose people are capable, and whose metrics reward the right behavior will, over time, develop a culture that matches, culture is downstream of what actually gets rewarded and decided, not a thing you announce into being with a values poster. This is also where incentives and rewards, compensation, promotion criteria, performance-review weighting, belong, and they deserve their own explicit check precisely because they are the quiet, common exception to this module's discipline: a redesign can genuinely move structure, process, governance, and metrics, and still fail if the bonus plan underneath it all keeps rewarding the old behavior. Call this the boxes-and-lines trap's milder cousin, not "we never touched the chart," but "we touched everything except what people are actually paid to do", and Common Mistake 11 and Practice P13 both drill it directly, because it is quiet enough to slip past a redesign that got the other four levers right.

Sequencing the five levers: and the hand-off to CN7.02

A real redesign runs the five levers in a defensible order, not simultaneously and not in whatever order feels most urgent. Diagnose first: trace the actual symptom (a slow decision, a missed target) to the specific lever actually causing it. Worked Example 1's cycle-time problem turns out to be almost entirely a decision-rights problem, not a structural one, which is precisely why the prior "fix" (structure alone) failed. Design structure and governance together, since they interact most: a structural change that does not reassign the decision rights sitting inside the old structure leaves them exactly where they always were. Check capabilities against the new decision rights before they go live, not after someone has already made a bad call with authority they were not ready for. Design metrics last, against the finished structure and governance, since a metric designed before you know who is accountable will measure the wrong owner. Throughout, check incentives explicitly, the lever most likely to be silently forgotten.

This module has taught you what the target operating model should be. It has deliberately not taught you how to get a real organization, with real politics and real fear of change, from today's operating model to that target, that is CN7.02's job: sequencing the transformation into a roadmap, cascading the KPIs through named initiatives, standing up program governance, and running the actual steering committees and working sessions that carry a redesign from a memo to a lived reality. Whatever this module's redesign concludes, it earns the same answer-first discipline as any other consulting deliverable: cross-link M10.03 for the craft, never re-derived here: lead with the governing recommendation, support it with the evidence from each lever, and state the risks and the next concrete step, never bury the recommendation under a narrative of how you got there.

Common mistakes & how experts think differently

1. The boxes-and-lines trap. Redrawing the org chart (new titles, new reporting lines) while decision rights, process, capabilities, and metrics all stay exactly as they were. It is the easiest of the five levers to touch and, for that reason, the one most often mistaken for the whole redesign. Experts ask, of every structural change, "what decision right, what process step, and what metric moved with it?", and treat "none" as unfinished, not finished.
2. A RACI with too many Accountables and no real Responsible. Every additional "A" beyond one is a full sequential veto gate, not extra rigor, it multiplies cycle time without improving decision quality. Experts insist on exactly one Accountable and a Responsible with something real to do, not merely someone to blame if the single Accountable is slow.
3. Span-of-control extremes in either direction. Too wide starves people of coaching and lets errors compound unnoticed; too narrow creates a ghost layer relaying messages a wider span would carry just as well, at the cost of one more hop every decision must clear. Experts test the actual span, not assume "more layers" always means "more control."
4. Choosing a structure archetype by imitation or fashion rather than by the specific coordination need the strategy creates. A matrix with no tie-breaker between its two lines is slower and more political than the hierarchy it replaced, not more sophisticated.
5. Mismatching capability work to the decision rights being delegated. Training a team for authority nobody delegated wastes the budget; delegating authority nobody checked was ready to use wastes the authority. Worked Example 4 shows the audit that catches this.
6. Mistaking activity for outcome, the vanity-metric trap. "Promotions run" or "tickets closed" can rise sharply while the outcome those activities serve stays flat or worsens. Experts insist a cascade measures outcome, or a defensible leading driver of it, never mere activity volume.
7. Shipping a metric with no guard partner: inviting Goodhart's law. A single unpaired speed metric, tied to a bonus, gets gamed by definition, because people optimize the number, not the goal it stands for. Experts pair every speed or activity metric with a quality guard before either goes live.
8. Confusing this module's decision rights with CN0.02's task delegation. Cousins, not the same tool, one governs a single task this week, the other a whole category of recurring decisions for years. Using either formula at the other's scale builds the wrong tool.
9. Building a DOA table once and never revisiting it. A threshold sized for one scale silently mis-delegates once the organization grows, quietly reintroducing the bottleneck the redesign removed. Experts put the DOA table, and the whole RACI, on a stated review cadence.
10. Skipping diagnosis and jumping straight to a target-state chart. A redesign that never asked which lever is actually causing the symptom risks fixing a lever that was never broken while leaving the real cause (usually governance) untouched.
11. Getting structure, process, governance, and metrics right, and forgetting incentives. A bonus plan still rewarding the old behavior quietly undoes an otherwise well-executed redesign, the boxes-and-lines trap's quieter, harder-to-spot cousin.

The thread through all eleven: the expert treats an operating model as five separately-checkable artifacts, never a single picture, and refuses to call a redesign finished until every one, and the incentives underneath them: has actually, verifiably moved.

Worked examples

Worked example 1: Nandavan Retail: diagnosing the trap (India, ₹)

Nandavan Retail (synthetic, Indian multi-category retail chain) has grown from 50 to 180 stores across four regions (North, South, East, West) over six years. Every regional promotional-pricing request must pass through five sequential steps before it can go live: a Regional Category Manager drafts it, an HQ Category Manager reviews it, the HQ Merchandising Head approves the concept, Finance signs off on the margin impact, and, for anything above a threshold, the CEO gives final sign-off. Each step averages six business days (calendar conflicts, email back-and-forth, meetings to schedule) before moving to the next. Regional festival windows (the highest-value promotional opportunity of the year) are live for only three weeks (fifteen business days) at a time. Last year, promotions missed the festival window entirely in three of the four regions; each missed window forfeits an estimated ₹1.8 crore of incremental gross margin the promotion would otherwise have captured.

Solution. Step 1, compute the baseline cycle time and its cost. Five steps × six business days = 30 business days = six weeks, twice the fifteen-business-day festival window, so any promotion routed through the full chain is structurally certain to miss the window regardless of how quickly any single approver happens to move. Foregone margin from the three missed regions: 3 × ₹1.8 crore = ₹5.4 crore in one year.

Step 2, examine what the 18-month-old "fix" actually changed. Eighteen months ago, leadership hired a boutique firm that redesigned the org chart: each region's operations head was retitled "Regional Managing Director" (RMD) and elevated to a solid-line report to the CEO, with a town hall announcing "greater regional empowerment." Check what moved: the five-step approval workflow did not change (the same four gates still exist, in the same order); the decision rights did not change (no RMD gained formal authority to approve a promotion); the system access did not change (only HQ category managers hold the approval login). Only the box on the chart and the title moved.

Step 3: measure the result. One year after the "fix," average cycle time is measured again: roughly 28 business days (5.6 weeks), a two-day improvement, entirely attributable to newly-retitled RMDs chasing HQ a little more assertively, not to any step being removed from the chain. Festival-window misses: still three of four regions. Foregone margin: roughly ₹5.2 crore, a (5.4 − 5.2) ÷ 5.4 ≈ 3.7% improvement: noise, not a fix.

Read it. This is the boxes-and-lines trap, named exactly: a new chart and a town hall, while every other lever (process, decision rights, capabilities, metrics) stayed frozen, so the number that actually mattered barely moved. The true five-step, thirty-day model is fixed in Worked Examples 2 and 3, not by drawing a different picture but by changing what the picture never touched.

Worked example 2: Nandavan Retail: structure done right (India, ₹)

Nandavan Retail's 180 stores run through five management layers: 30 Cluster Managers (about six stores each) report to 6 Area Managers (about five Cluster Managers each), who report to the 4 Regional Heads (an average of just 1.5 Area Managers per Regional Head), who report to the CEO.

Solution. Step 1: test each layer's span. Cluster Managers: 180 ÷ 30 = 6 stores each, reasonable for fairly standardized store operations. Area Managers: 30 ÷ 6 = 5 Cluster Managers each: reasonable. Regional Heads: 6 Area Managers ÷ 4 regions = 1.5 Area Managers per Regional Head on average, a span far below the roughly-four-person floor that signals a real coordination role, and a strong signal that the Area Manager layer, sitting between Cluster Managers and Regional Heads, is doing more relaying than real supervision.

Step 2, remove the ghost layer and recompute. Fold the Area Manager layer out, letting the 30 Cluster Managers report directly to the 4 Regional Heads: new span = 30 ÷ 4 = 7.5, roughly 7–8 Cluster Managers per Regional Head, squarely inside the six-to-ten range appropriate for moderately routine, playbook-driven store-operations supervision, which Cluster Manager work largely is (note the condition explicitly: this delayering is defensible because Cluster Manager work follows a standardized playbook; if it required heavy unstructured judgment, jumping a Regional Head's span from 1.5 to 7.5 at once would be reckless, not efficient).

Step 3, compute the direct cost saved. Eliminating 6 Area Manager roles at a fully-loaded cost of ₹28 lakh/year each saves 6 × ₹28 lakh = ₹1.68 crore/year in direct headcount cost, and, more importantly than the headcount saving alone, removes one full relay step that every decision routed up that chain used to pass through, directly attacking the cycle-time problem Worked Example 1 diagnosed (a saving Worked Example 3 compounds with the governance fix).

Read it. Structure alone did not solve Nandavan's cycle-time problem (the five-step approval workflow still exists after this delayering) but it removes a genuine ghost layer and saves real cost, because the span arithmetic said so, not because a consultant felt the chart looked too tall. Notice, too, that this worked example found only one problem layer, not five; a real diagnosis sometimes concludes most of a structure is fine, itself a useful, defensible finding.

Worked example 3: Nandavan Retail: governance and RACI done right (India, ₹)

Continuing directly from Worked Examples 1–2: the underlying decision (approve a regional promotional pricing plan) still has no single owner. Rebuild its RACI and DOA table from scratch.

Solution. Step 1: diagnose the old RACI. The old pattern effectively assigned Accountable to four different roles in sequence (HQ Category Manager, HQ Merchandising Head, Finance, and CEO each functioned as an independent, blocking gate) while the Regional Category Manager who actually drafted the plan held no real authority, only the "R" of doing the paperwork. Four sequential A's, at six days each, is exactly the 24 (plus the draft step's 6 days = 30) business days computed in Worked Example 1.

Step 2, build the fixed RACI and a three-tier DOA table. Regional Category Manager: R, always, and A at the smallest tier. Regional Managing Director (the real structural role from Worked Example 2, now given real authority to match its real title): A at the middle tier. CEO: A only above the top threshold. HQ Merchandising Head and Finance: C, they set margin and brand guardrail policy on a standing quarterly cadence, and are consulted case-by-case only on genuine exceptions, never a blocking sign-off on every plan. CEO below the top threshold: I, via a monthly pricing dashboard.

Step 3, compute the blended cycle time. Historically, 40% of promotions fall in tier 1, 45% in tier 2, and 15% in tier 3. Weighted average = (0.40 × 2.5) + (0.45 × 5.0) + (0.15 × 7.5) = 1.0 + 2.25 + 1.125 = 4.375 ≈ 4.4 business days, against the fifteen-business-day festival window, this leaves roughly 10.6 business days of slack for actual execution and marketing, a complete reversal of the original process, which consumed twice the window before a promotion could even go live.

Read it. The chart did not change again here, the fix is entirely in decision rights: one Accountable per tier, real authority for the person who always did the work, and specialist functions moved from blocking gates to standing-policy consultants. Combined with Worked Example 2's delayering, this structure-plus-governance pair, not a new title, is what actually buys back the six weeks.

Worked example 4: Nandavan Retail: capabilities and the metrics cascade (India, ₹)

The new DOA table hands each region's Regional Category Manager (RCM) real pricing authority for the first time. Are they ready? A capability audit rates each of the four RCMs (1 = needs training, 2 = developing, 3 = ready) on margin/P&L literacy, local competitive judgment, and system proficiency: North (2, 3, 1), South (3, 3, 2), East (1, 2, 1), West (2, 2, 2).

Solution. Step 1, read the heatmap. South is nearly ready (a light system refresh is enough). North, East, and West each show a real gap, most acutely East (a newer RCM), on multiple skills at once.

Step 2: cost the train-versus-hire call. Training North, East, and West in a structured six-week program costs ₹4 lakh per manager: 3 × ₹4 lakh = ₹12 lakh total. Replacing those three with new hires who already have the skill costs an estimated ₹18 lakh each: 3 × ₹18 lakh = ₹54 lakh, more than four times the training cost, and forfeits each incumbent's local relationships and competitor knowledge, built over years. Train, do not replace: the arithmetic and the retained knowledge both point the same way (South needs only a brief refresher). Note the distinction from CN6.02: this is developing people already on the team for newly-delegated authority, not acquiring a capability the organization lacks anywhere, a different question at a different scale.

Step 3: cascade the metrics. The objective: "grow regional relevance and same-store sales without eroding gross margin." Corporate: gross margin % (hold ≥38%) and same-store sales growth (≥8%/year). Regional: promotional sell-through rate (≥70%) and festival-window capture rate (100%, directly measuring whether Worked Example 1's original problem is fixed). Flag the vanity-metric trap: "number of promotions run" must never stand alone, since a newly-empowered RCM could hit an activity target by running many low-value promotions while quietly eroding margin, every activity metric is paired with the margin guard.

Step 4: score a live quarter. Region North reports: gross margin 37.5% (target 38%, achievement 37.5 ÷ 38 = 98.7%), sell-through 74% (target 70%, achievement capped at 100%), festival capture 100% (achievement 100%). Weighted scorecard (margin 50%, sell-through 30%, capture 20%): (0.50 × 98.7) + (0.30 × 100) + (0.20 × 100) = 49.35 + 30 + 20 = 99.3/100, an excellent quarter, and one the paired margin metric confirms was not bought by running promotions indiscriminately.

Read it. Capability and metrics are the last two levers precisely because they depend on the first three being settled: you cannot honestly audit a skill gap until you know what authority the skill is for, and you cannot cascade a metric until you know who is actually accountable for moving it.

Worked example 5: Cascade Mutual Insurance: decision rights and process (US, $)

Cascade Mutual Insurance (synthetic US regional property-and-casualty insurer) has an average claims-handling time (AHT) of 14 calendar days. Of all claims, 55% are under $5,000 (an adjuster may settle alone), 30% are $5,000–$25,000 (a Team Lead must sign off), and 15% exceed $25,000 (a Claims Manager must sign off). Separately, a blanket policy (instituted years ago after one bad multi-line claim, never revisited) requires a second, full Team Lead review for any claim touching more than one coverage line, regardless of size; this hits roughly 40% of all claims and adds an average of three calendar days each, because it depends on a Team Lead's calendar availability for a full, formal review. New unassigned claims also sit in a queue an average of 1.5 days before an adjuster picks them up.

Solution. Step 1, fix the RACI, not just the workflow. The blanket cross-line review is a classic redundant gate: it treats every cross-line claim as equally risky regardless of size, when the actual risk driver is complexity. Redesign it as a C, not a repeated A: the Team Lead is consulted only when the adjuster affirmatively flags a question, expected to trigger for about 8% of claims, at 0.5 days each rather than a full 3-day review.

Step 2, compute the governance saving. Expected days added by the old blanket policy = 0.40 × 3 = 1.2 days, averaged across all claims. Expected days added by the new, flagged-only policy = 0.08 × 0.5 = 0.04 days. Saving = 1.2 − 0.04 = 1.16 days. New AHT = 14 − 1.16 = 12.84 ≈ 12.8 days, a (1.16 ÷ 14) ≈ 8.3% reduction from a single decision-rights fix, before touching anything else.

Step 3: fix the process, separately. The 1.5-day unassigned-claims queue is pure queueing waste, not a decision-rights problem: an automated, rules-based assignment engine routes a new claim to an available adjuster immediately, cutting the wait to an estimated 0.2 days, a saving of 1.5 − 0.2 = 1.3 days.

Step 4: combine both fixes. Total saving = 1.16 (governance) + 1.3 (process) = 2.46 days. New AHT = 14 − 2.46 = 11.54 ≈ 11.5 days, a (2.46 ÷ 14) ≈ 17.6% reduction in cycle time.

Read it. Two different levers produced two savings that stack cleanly: the governance fix removed a redundant decision gate; the process fix removed a queueing delay unrelated to who was allowed to decide what. Diagnosing which lever caused which delay (rather than redrawing Cascade's chart) is the discipline Worked Example 1 showed Nandavan skipping.

Worked example 6: Cascade Mutual Insurance: when a metric gets gamed (US, $)

Cascade Mutual rolls out a new KPI following Worked Example 5's fix: each adjuster's quarterly bonus is tied to Average Handling Time, with a target of 11.5 days (matching the new benchmark). The following quarter, on a base of 2,000 claims, AHT falls to 11.2 days: beating the target. But the reopened-claim rate (claims formally closed, then reopened within 60 days because they were closed with incomplete documentation), previously unmeasured, rises from a historical 4% to 11% of closed claims. Each reopened claim costs an average $340 in rework.

Solution. Step 1, see the mechanism. Adjusters, evaluated on AHT alone, began closing ambiguous claims (concentrated in the $5,000–$25,000 tier) before documentation was genuinely complete, in order to hit the bonus-linked speed target, exactly Goodhart's law: once AHT became the target, it stopped being a trustworthy measure of genuinely-resolved claims, because people began optimizing the number itself.

Step 2: cost the failure. Reopened claims before: 4% × 2,000 = 80. Reopened claims after: 11% × 2,000 = 220. Increase: 220 − 80 = 140 additional reopened claims. Cost: 140 × $340 = $47,600 in additional quarterly rework, a real, dollar-denominated cost the AHT dashboard, viewed alone, never shows, because AHT actually improved (11.2 versus the 11.5 target) in the very quarter this cost was created.

Step 3, fix it with a guard metric. Pair AHT with the reopened-claim rate as a joint condition: the bonus pays only if AHT is at or below target and the reopened-claim rate stays at or below a stated cap (say, 5%), removing the incentive to trade documentation quality for speed, without abandoning the legitimate speed goal.

Read it. This mirrors Worked Example 4: there, the gaming trap was designed around in advance; here, Cascade fell into it, fixed with the same discipline applied after the fact, no metric ships alone, every speed metric is paired with a guard that catches the specific way a bonus-motivated person could game it.

Practice set

Work each problem fully before reading its solution. Currency figures are synthetic and internally consistent; India uses ₹, the US uses $, as stated per problem.

P1 (guided). Span, layers, and delayering. A distribution company has 240 warehouse staff, organized into shifts of 8 staff per Shift Supervisor, and 5 Shift Supervisors per Warehouse Manager. (a) Compute the number of Shift Supervisors and Warehouse Managers required. (b) The company considers flattening by widening each Shift Supervisor's span to 12 staff instead of 8. Recompute, and state how many supervisory roles this eliminates in total.

Solution. (a) Shift Supervisors = 240 ÷ 8 = 30. Warehouse Managers = 30 ÷ 5 = 6. (b) At a span of 12: Shift Supervisors = 240 ÷ 12 = 20; Warehouse Managers = 20 ÷ 5 = 4. Roles eliminated = (30 − 20) + (6 − 4) = 10 + 2 = 12 supervisory roles. Whether this is wise depends on whether the work is routine enough to support the wider span, the arithmetic alone never answers that; it only tells you the headcount consequence of the choice.

P2 (guided). Spot and fix a broken RACI. A product-return approval process currently assigns: Store Associate = R, Store Manager = A, District Manager = A, Regional VP = A, Finance = A. Name the flaw and redesign it with a two-tier DOA (returns ≤$200; returns >$200).

Solution. The flaw is four sequential Accountables for a single decision, every return, regardless of size, clears four independent approval gates, and the Store Associate who actually handles the customer has no real authority, only the "R." Fixed design: Store Associate = R, always. Store Manager = sole A for returns ≤$200 (Finance and District Manager = I, not gates). Regional VP = sole A for returns >$200 (Store Manager and District Manager = C, providing context, not a blocking sign-off; Finance = I via a monthly return-value dashboard, escalated only if reserve-worthy fraud is suspected). One Accountable per tier, not four Accountables for every case.

P3 (independent). DOA-weighted cycle time. A claims process has three DOA tiers: tier 1 (60% of claims, 1 approval step, 1 day) tier 2 (30% of claims, 2 steps at 1.5 days/step) tier 3 (10% of claims, 3 steps at 1.5 days/step). Compute the blended average cycle time.

Solution. Tier 1: 1 × 1 = 1 day. Tier 2: 2 × 1.5 = 3 days. Tier 3: 3 × 1.5 = 4.5 days. Weighted average = (0.60 × 1) + (0.30 × 3) + (0.10 × 4.5) = 0.6 + 0.9 + 0.45 = 1.95 days.

P4 (independent). Centralize or decentralize? A 60-outlet restaurant chain must decide, decision by decision, whether to centralize or decentralize: (i) core recipe standards and food-safety protocol; (ii) daily staff scheduling around local foot-traffic patterns; (iii) national advertising campaign creative; (iv) response to a specific local competitor's price cut. For each, state centralize or decentralize and justify with the "who has better information, who bears the consequence" test.

Solution. (i) Centralize: brand and food-safety consistency matters far more than local variation, and the consequence of inconsistency is severe and shared across every outlet. (ii) Decentralize, the local manager has far better information about local foot traffic, and a scheduling misstep is small and local. (iii) Centralize, campaign economics and brand consistency dominate, and no single outlet has the scale-relevant information to do this well alone. (iv) Decentralize, within a bounded discount range, the local manager has the timeliest information on the specific threat, and a slow, centrally-routed response arrives after the moment has passed; a DOA cap protects against discounting away the brand's pricing integrity.

P5 (independent). Match the structure archetype. For each, name the best-fit archetype (functional, divisional, or matrix) and the core trade-off it accepts: (i) a single-product national retailer whose main advantage is deep merchandising and supply-chain expertise; (ii) a conglomerate running three genuinely distinct businesses (insurance, asset management, and a payments platform) each needing its own P&L accountability; (iii) a B2B software company whose largest clients demand both consistent global account management and deep country-specific regulatory execution.

Solution. (i) Functional: groups by specialist skill to maximize depth in the one thing that matters most, accepting slower cross-functional coordination, tolerable because the business itself is not fragmented. (ii) Divisional, each business needs distinct P&L accountability and runs different economics; the cost accepted is duplicated functional capability across divisions. (iii) Matrix, needs global consistency and local execution at once, which a single hierarchy cannot deliver; the cost accepted is dual reporting lines and real stall risk absent a tie-breaker.

P6 (independent). Process waste and redesign. An order-to-cash process runs: order entry (0.5 day), credit check (1 day), warehouse pick (1 day), pack (0.5 day), ship (0.5 day), invoice (1 day): total 4.5 days. Diagnosis finds the credit-check and invoice steps both manually re-key the same customer data the order-entry step already captured. Automating both cuts credit check to 0.2 day and invoicing to 0.3 day. Compute the new total cycle time and the percentage improvement.

Solution. Savings: credit check 1 − 0.2 = 0.8 day; invoicing 1 − 0.3 = 0.7 day; total saving = 1.5 days. New total = 4.5 − 1.5 = 3.0 days, a 1.5 ÷ 4.5 = 33.3% reduction, a process-lever fix with no structural or governance change at all, because the waste here was duplicative data re-entry, not a decision-rights problem.

P7 (independent). Train versus hire. Five technical-support team leads need training on a newly adopted CRM. A structured two-week program costs $1,800 per lead. Replacing all five with CRM-experienced hires costs an estimated $14,000 each in recruiting and onboarding, and forfeits each incumbent's average three years of accumulated product knowledge. Compute both total costs and recommend.

Solution. Training: 5 × $1,800 = $9,000. Replacement: 5 × $14,000 = $70,000, nearly eight times the training cost, before even counting the forfeited product knowledge. Train, not replace: both the direct arithmetic and the retained institutional knowledge point the same way.

P8 (independent). Build a metrics cascade. A company's strategic objective is "reduce customer churn while holding service cost flat." Draft a three-level cascade (corporate / business-unit / team) with one metric per level, and compute how much better than target the corporate metric performed if the target was 5.0% annual churn and actual churn was 4.6%.

Solution. Corporate: annual churn rate (target 5.0%, actual 4.6%). Business-unit: first-90-day churn rate (a leading indicator of the corporate outcome) paired with cost-per-service-interaction (the guard against fixing churn purely by spending more). Team: first-call resolution rate. Achievement: (5.0 − 4.6) ÷ 5.0 = 8% better than target, expressed as relative improvement, not the raw percentage-point gap.

P9 (independent). Spot the vanity metric. A customer-support team is evaluated solely on "tickets closed per day." Predict the failure mode, and name the fix.

Solution. Predicted failure: agents close tickets quickly and prematurely to maximize the count, marking issues resolved before the customer's actual problem is fixed, driving repeat contacts and eroding satisfaction, an outcome invisible in "tickets closed" alone. Fix: pair the activity metric with a guard-metric such as repeat-contact rate within 7 days, or a post-resolution customer satisfaction score, and condition any reward on both moving together, the same paired-metric discipline Worked Example 6 applies after the fact.

P10 (independent). Diagnose a boxes-and-lines scenario. A manufacturing company merges the management of two plants under a new "VP of Combined Operations" title, announced with a company-wide message about "unified operations." Six months later, the two plants still run separate approval workflows, separate IT systems, and separate bonus plans. Name what is missing and predict the outcome.

Solution. This is the boxes-and-lines trap: a new title and a merged reporting line, with process (separate workflows), governance/decision rights (nothing reassigned, whoever approved what before still does), capabilities (no capability question even asked), and metrics/incentives (separate bonus plans still rewarding separate-plant behavior) all untouched. Predicted outcome: no real operational unification occurs, decisions are still made exactly where they always were, at the same speed, by the same rules, because nothing that actually governs behavior moved with the chart.

P11 (independent). Design a governance cadence. A multi-region company needs a standing cadence for quarterly capital-expenditure approval, with a mix of routine small requests and rare large ones. Propose a cadence of forums and name what each one covers.

Solution. A monthly regional operating review clears routine, small capex requests within the regional DOA tier, keeping the vast majority of requests off any escalation path. A quarterly capital steering committee reviews the rarer, larger requests above the regional threshold, plus a rolled-up view of what regions approved locally, preserving central visibility without central bottlenecking. An annual policy-refresh session, owned by Finance in a Consulted role, resets the DOA thresholds themselves against the company's current scale, the cadence that prevents Common Mistake 9's stale-DOA-table failure.

P12 (independent). RACI with no Responsible. A decision matrix assigns Finance = A, Regional Head = C, CEO = I, with no role assigned R at all. Explain what goes wrong and fix it.

Solution. With no Responsible, nobody's job is to actually draft the analysis or the first-cut recommendation, the decision stalls because everyone is waiting for someone else to start the real work, even though an Accountable exists to eventually approve it. Fix: assign R explicitly to whichever role has the closest working knowledge of the specific decision (commonly the Regional Head, moved from C to R, with Finance remaining the sole A), a decision needs someone who does the work before it needs someone who signs off on it.

P13 (independent). The incentives gap. A company redesigns structure, governance/RACI, and metrics correctly, delegating real regional pricing authority with a clean DOA table. Six months later, regional managers are still deferring every decision to HQ. Investigation finds the annual bonus plan still rewards HQ merchandising staff (not regional managers) for "national promotion count." Name the missed lever and the fix.

Solution. The missed lever is incentives/rewards: the boxes-and-lines trap's quieter cousin: structure, governance, and metrics all moved, but what people are actually paid to do did not. Regional managers rationally hesitate to use authority that carries risk when the reward system still credits someone else for the outcome. Fix: rewrite the bonus plan so regional managers are measured and paid on the regional-level metrics (sell-through, margin-protected promotion capture) the redesign just gave them accountability for, retiring the now-obsolete "national promotion count" credit to HQ staff.

P14 (independent). Cost of delayering, fresh numbers. A bank branch network has 5 Area Managers, each a fully-loaded cost of $95,000/year, currently overseeing 20 Team Leads (span of 4 each). Leadership proposes delayering so Team Leads report directly to 2 Regional Directors instead. Compute (a) the annual cost saved by eliminating the Area Manager layer, and (b) the new span per Regional Director.

Solution. (a) 5 × $95,000 = $475,000/year saved. (b) 20 ÷ 2 = 10 Team Leads per Regional Director, worth checking, before approving, whether Team Lead work is routine enough to support a span that wide; the arithmetic only states the consequence, not the verdict.

P15 (independent). Sequencing, explained. In two or three sentences, explain why a redesign should diagnose the actual cause of a symptom before touching structure, and why incentives are the lever most often forgotten even in an otherwise careful redesign.

Solution. Diagnosing first matters because structure is the easiest lever to change reflexively, so skipping diagnosis risks "fixing" a layer that was never the actual cause, exactly Worked Example 1's cosmetic reorg, which redecorated a chart while the real cause, a frozen five-step approval chain, sat untouched underneath it. Incentives are forgotten most often because they are the least visible lever, nobody draws "the bonus plan" as a box, so a team can believe a redesign is complete after moving the other four levers, without checking whether what people are paid to do still matches what the redesign just asked of them.

P16 (timed, 30 minutes: integrative). Konark Logistics. Konark Logistics (synthetic, India) runs 360 drivers through Route Supervisors (span of 12 drivers each) and Depot Managers (span of 6 Route Supervisors each), all reporting to a single Operations Director. (a) Compute the number of Route Supervisors and Depot Managers, and the Operations Director's resulting span; state whether this layer needs delayering. (b) The real problem: approving a new customer route-and-pricing contract takes four weeks because the Route Supervisor, the Depot Manager, and the Operations Director each independently re-review the same margin calculation: three sequential Accountables. Redesign the RACI with a two-tier DOA (contract value ≤₹10 lakh/year; >₹10 lakh/year). (c) Depot Managers have never held contract-approval authority before. A structured 2-week program costs ₹1.5 lakh per manager (5 managers); replacing the two weakest with experienced hires costs ₹12 lakh each. Compute both costs and recommend. (d) Propose one speed metric and one guard-metric for the new approval process. (e) Six months ago, Konark renamed "Depot Supervisor" to "Depot Manager" and added a dotted line to the Operations Director, promising "greater regional empowerment," but never removed the three-way margin re-review. Was this the boxes-and-lines trap? Justify in one sentence.

Solution. (a) Route Supervisors = 360 ÷ 12 = 30. Depot Managers = 30 ÷ 6 = 5. Operations Director's span = 5 Depot Managers, comfortably inside a defensible range for a role with genuinely complex, judgment-heavy coordination across depots, so no delayering is needed here; not every diagnosis finds a broken layer. (b) Route Supervisor = R, always, and drafts the contract terms. Depot Manager = sole A for contracts ≤₹10 lakh/year. Operations Director = sole A for contracts >₹10 lakh/year. Route Supervisor becomes C, not a repeated A, on the larger tier, since they already contributed as R. (c) Training: 5 × ₹1.5 lakh = ₹7.5 lakh. Replacing two: 2 × ₹12 lakh = ₹24 lakh, training all five costs less than replacing even two, and retains existing depot relationships; train, don't replace. (d) Speed metric: contract-approval cycle time (target ≤5 business days). Guard metric: contract margin-compliance rate (% of approved contracts meeting the minimum margin policy, target ≥95%), without it, a Depot Manager chasing the speed target could rubber-stamp margin-poor contracts to hit the number. (e) Yes, a title and a dotted line changed, but the three-way redundant margin re-review (the actual decision-rights problem) was never touched, so cycle time could not have moved regardless of the new box on the chart, the exact pattern Worked Example 1 names.

Applied mini-project: the operating-model redesign memo

The assignment. Produce a one- to two-page, answer-first memo (Minto-style: cross-link M10.03, do not re-derive its craft here) that runs a full redesign, across all five levers, on an organization of your choosing.

Choose one path.

• Path A: extend the synthetic dataset. Nandavan Retail (Worked Examples 1–4) opens a new fifth region. Nandavan East Coast, 42 stores, structured identically to the other four regions at launch, but inheriting the old, pre-redesign five-step approval chain because the new region was stood up by a different, recently-acquired team who never received the RACI/DOA fix. Run the redesign on this fifth region from scratch, reusing Worked Examples 2–4's method (not their exact numbers).
• Path B, a real organization. Choose a real organization whose structure is public record (a company's investor-relations organization chart, a public-sector body's published structure, a nonprofit's annual report) and, using only public information, diagnose one real, specific symptom of a decision-rights, process, or metrics problem, citing your source (R3: real, cited, in-session).

Do this, in order. (1) Diagnose: state the specific symptom and trace it to the lever(s) actually causing it, the way Worked Example 1 did. (2) Structure: compute at least one span-of-control or layer calculation, and state whether delayering is warranted, with the arithmetic shown either way. (3) Governance: build a RACI with exactly one Accountable per tier and a stated DOA table with real thresholds. (4) Capabilities: run a small heatmap for whichever role just gained new authority, and make a train-versus-hire-versus-redeploy call with a real cost comparison. (5) Metrics: propose one cascade with at least one paired guard-metric, stating how someone could game the primary metric if the guard were absent. (6) State plainly whether any part of your own redesign risks becoming a boxes-and-lines trap, and how you guarded against it. (7) Write it answer-first: the governing recommendation in the first line, then the supporting evidence lever by lever, then the risks and the next concrete step.

Grading rubric (score each component 0/1/2: 0 = absent or wrong, 1 = present with gaps, 2 = professional). (a) Diagnosis correctly traces the symptom to a specific lever, not a generic "the org is broken" claim. (b) Structure calculation shown and its delayering verdict justified by the arithmetic, in either direction. (c) RACI has exactly one Accountable per tier and a real, numbered DOA table. (d) Capability call backed by an actual cost comparison, correctly distinguished from a deal-scale build-buy-partner decision. (e) Metrics cascade includes a genuine guard-metric with a stated gaming scenario it prevents. (f) Answer-first memo structure, recommendation stated in the opening line, and an explicit self-check against the boxes-and-lines trap. Pass at 9/12 with no zero on any item.

Reading & resources

Everything below is an anchor to learn from and verify against, never a source to cite as if it were your client's own data. Free unless marked; level tags run Beginner → Advanced.

Operating-model design, the core spine:

• Jay Galbraith, Designing Organizations (and Designing Matrix Organizations That Actually Work), the Star Model (strategy, structure, processes, rewards, people) this module's five-lever frame descends from, and the deepest treatment of when a matrix actually works versus collapses into paralysis. [Paid] [Advanced]
• McKinsey Quarterly and BCG's published articles on operating-model design and organizational health, free, current, practitioner-level treatments of structure, governance, and capability redesign; verify any statistic before quoting it. [Free] [Intermediate]
• Widely-published project- and program-management guides to the RACI matrix (and its RASCI/DACI variants), free templates exist across most program-management bodies of knowledge; treat any numeric threshold in a template as illustrative, never a default to copy into a real DOA table. [Free] [Beginner–Intermediate]
• Goodhart's law ("when a measure becomes a target, it ceases to be a good measure," commonly attributed to economist Charles Goodhart), search any reputable overview for the original context and modern illustrations of metric gaming; treat it as a durable warning, not a single citable formula. [Free] [Intermediate]

Cross-links inside the program (do not re-derive; read there for the depth):

• CN3.02: the McKinsey 7-S framework, the diagnostic lens this module's construction toolkit builds on without re-teaching.
• CN0.02: individual task delegation, the smaller-scale cousin of this module's organizational decision rights.
• CN6.02: the Integration Management Office and 100-day PMI plan, a time-boxed application of this module's governance and capability toolkit under a deal clock; also the deal-scale build-buy-partner framework this module's capability lever is deliberately distinguished from.
• CN7.02: the transformation roadmap, program governance, and change-management journey (Kotter, stakeholder management, steering committees) that carries this module's target-state design from memo to lived reality.
• M5.01–M5.10: the finance galaxy's sector playbooks, for the specific operating rhythms (a bank's three-lines-of-defense, a hospital's clinical governance) that this module's generic method adapts to.
• M10.03: Minto's Pyramid Principle, the answer-first memo craft this module's deliverables use but never re-teach.

Practitioner-oriented, free unless marked:

• Harvard Business Review, search its archive for "organizational design," "span of control," and "RACI" for practitioner case studies; treat every specific figure as illustrative until traced to the named organization's own public disclosure. [Mixed free/paid] [Intermediate–Advanced]
• John Kotter, Leading Change: cross-link CN7.02, which owns the change-sequencing application; cited here only for the reader who wants the full change-management text now rather than at CN7.02. [Paid] [Intermediate]

The canon shared with the rest of this galaxy (Minto, Conn & McLean, Rasiel, Porter, Helmer, Greenwald, Koller) is indexed once, with full free/paid and level tags, in CN0.01.

Teach it back & journal

Feynman prompt. Explain to a colleague who has just been told "we're reorganizing, here's the new chart" why that sentence, on its own, tells you almost nothing about whether anything will actually change. Walk them through the five levers in order: structure (the chart itself), process (how work actually flows), governance and decision rights (who can actually say yes, made explicit as a RACI and a DOA table), capabilities (whether the people about to be empowered can actually use that authority), and metrics (what will actually be measured and rewarded, and whether it can be gamed). If your colleague can, unprompted, ask "okay, but who's actually accountable for what now, and did the bonus plan change too?", you have installed this module's real payoff: an org chart is a drawing, not a redesign.

Journal prompt. Think of one real reorganization you have witnessed, read about, or experienced, at a company, a school, a club, anywhere structure was announced to change. Using only what you can honestly recall or find in public record, write four short paragraphs: what the org chart actually changed; whether decision rights, process, capabilities, and metrics/incentives moved with it, lever by lever, or stayed exactly as they were; whether the redesign, in retrospect, was a boxes-and-lines trap, a genuine five-lever redesign, or something in between; and, honestly, what you still cannot know from the outside, where only someone who watched the real decisions get made day to day would be able to tell you whether the redesign actually worked. That last paragraph is not a hedge; naming clearly what a floor of structure and mechanics cannot tell you from the outside is itself the advisory instinct this whole galaxy is trying to build.

This module's flashcards and mastery quiz are wired into the app: see the node's Quiz and Reviews.