The CDD Toolkit + Voice-of-Customer

Branch: Business Evaluation & Commercial Due Diligence (Consulting galaxy) · Node CN5.03 · Purple belt (Engagement Manager) · ~12 focused hours · Mastery gate ≥ 85% + mapped case

A private-equity partner calls on a Tuesday and says the words that start every engagement in this branch: "We have four weeks of exclusivity on a deal we like. Tell us if we're right." Nothing else is specified. Not which four weeks matter most, not which twenty people to call, not which of the seller's numbers to trust. That gap (between "we like this deal" and a defensible, evidenced, answer-first verdict) is commercial due diligence (CDD), and it is one of the highest-leverage things a strategy consultant is ever paid to do, because the fee is a rounding error against the size of the check the client is about to write. Get it right and you save a client from a bad acquisition or give them the conviction to move fast on a good one. Get it wrong, miss the one customer segment that is quietly churning, accept the seller's growth story without testing it against a single real buyer, size a market top-down and never check it bottom-up, and the client finds out the hard way, eighteen months and one large wire transfer later.

This module builds the whole toolkit end to end, threaded through one continuous, fully worked case so you see how the six pieces actually fit together. You turn a vague thesis into hypotheses that can be killed, then build the issue tree that turns "is this a good business" into a workplan. You design a Voice-of-Customer (VoC) and expert-call program that surfaces what the seller's deck cannot. You build and reconcile a market model from two independent directions, and run a competitive teardown that finds the actual mechanism of advantage, not a logo count. Last, you land a weighted go/no-go with a named risk and a named mitigation for everything that could still go wrong. Every number in every worked example below is invented for teaching and stated as such (synthetic, internally consistent, recomputed in Python), and the companies do not exist. The method is the one you would run on a real one.

One honest note before you start, in the spirit the very first module opened the whole galaxy with: the judgment to know which red flag is a dealbreaker and which is just noise compounds from having sat through dozens of real diligences and watched which calls turned out right. That judgment cannot be fully installed by a module. What can be installed, and what the work below drills to reflex, is the structure underneath that judgment: the habit of falsifiable hypotheses, a MECE workplan, a bias-checked research design, a reconciled market model, an evidenced teardown, and a synthesis that never lets a good composite score overrule a killed hypothesis. That structure is the teachable floor. It is also, not coincidentally, most of what separates a diligence a partner trusts from one they quietly redo.

Learning objectives

By the end you can:

1. State what a Commercial Due Diligence is and is not (distinguishing it from financial, tax, and legal due diligence) and identify who commissions one (a PE sponsor pre-investment, a corporate acquirer pre-deal, a lender pre-financing) and why the timeline is always short and the stakes always large.
2. Convert a vague investment thesis ("we like this space") into a small set of falsifiable hypotheses, each carrying an explicit kill criterion, the specific finding that would make you recommend walking away.
3. Build a MECE CDD issue tree and workplan across the five standard lenses (market, competitive position, customer, financial quality, and execution/organizational risk) and map every hypothesis to the lens(es) that will test it.
4. Design a Voice-of-Customer and expert-call research program: build a sampling frame across genuinely different segments, write a discussion guide free of leading questions, and name the standing biases (selection, sponsor, small-n, social-desirability) that quietly corrupt primary research if left unchecked.
5. Build a top-down and a bottom-up market model and reconcile them, treating the gap between the two approaches, not just their average, as a source of insight.
6. Run a competitive teardown that isolates the actual mechanism of advantage (or its absence) rather than a feature checklist, and connect it explicitly to a moat check.
7. Read a target's financial quality commercially (revenue quality, customer concentration, cohort/retention behavior, a same-store growth bridge) without re-performing financial due diligence.
8. Synthesize every lens's evidence into a weighted go/no-go, build a risk register with a named mitigation for every material risk, and apply the veto rule that a killed hypothesis overrides any composite score.
9. Hand the verdict off as an answer-first CDD memo (governing thought → key line → support), and know precisely which part of that hand-off belongs to this galaxy and which belongs to deck craft.
10. (Productivity objective: R10.) Use AI to accelerate VoC-transcript synthesis, market-sizing desk research, competitor benchmarking, and risk-register/memo drafting, while keeping every figure, quote, and risk traced to evidence you generated or verified yourself.

Prerequisites & connections

Builds on. CN5.01 (Business-Model Teardown & Unit Economics. Advisory Lens) supplies the contribution/CAC-LTV/cohort vocabulary that the financial-quality lens here leans on directly. CN5.02 (Market Attractiveness & Competitive Dynamics) supplies the market-structure and profit-pool judgment that the market model and the teardown apply. From further back, CN1.01 (Issue Trees & MECE) is the tree-building method behind every CDD workplan, and CN1.02 (Hypothesis-Driven Problem Solving) is the same machinery, now pointed at an investment thesis. CN2.02 (Market Sizing & Guesstimates) is where you first met the top-down/bottom-up mechanics. Here you apply them to a live deal without re-deriving them. CN3.01 (Competitive & Industry Frameworks) supplies Five Forces, 3C, and the value chain for the teardown. CN3.03 (the meta-skill: when frameworks mislead) is the standing warning against mistaking scale for a moat, and the density example below makes it concrete with real arithmetic.

Feeds forward. A "go" verdict here is the direct input to CN6.01/CN6.02, the deal-rationale, synergy-build, portfolio, and PMI work that only starts once commercial diligence has cleared. The risk register you build in the final stage is the seed of the 100-day risk log that CN7.01/CN7.02 formalize during implementation. The answer-first memo hands off to CN1.03 (The "So What": Synthesis & Answer-First Storyline). Outside the consulting galaxy it hands off to M10.03 ("The Communication Toolkit: Making the Analysis Change a Decision," the Minto Pyramid in full), and eventually to the Presentation galaxy for deck craft proper. What you get here is only the consulting-specific answer-first logic of the verdict. Storylining and slide design are taught elsewhere, and Minto's mechanics aren't re-taught.

In the Finance galaxy, M4.01 (Business Model Anatomy) and M4.02 (Unit Economics) are the mechanics the financial-quality lens applies here without re-teaching them. M4.03 (Industry Structure: Porter & Value Chains), M4.04 (Moats I: The Five Sources & the Numbers Test), and M4.05 (Moats II: Greenwald, 7 Powers & Measuring the Moat) are the frameworks the competitive teardown leans on. Go there for the underlying mechanics and come here for the advisory application. M3.09 (Modeling II: DCF, Comps, LBO & M&A Basics), M3.10 (Special Situations + Phase Capstone), and E11.02 (M&A, Deals & Special Situations Lab) are where the deal math that follows a "go" gets done. You own the advisory-seat judgment of whether and why, and the modeling galaxy owns the spreadsheet. M5.01 to M5.10, the seventeen sector playbooks, give you KPI-level industry depth once you know which sector you're diligencing. The generic CDD method you learn here decides which playbook to open. And if this is your first CDD, welcome back. E11.03 (The Consulting Case-Interview Lab) and its case CL-11.3 ("Kite Foods Profitability") were the branch's Cycle-2 taster, and this cluster is the deep treatment that taster was built to lead into.

Core teaching content

5. The market model: top-down, bottom-up, and the gap between them

A CDD market model reuses the market-sizing mechanics you already know. There's a top-down estimate: population or unit count × incidence or penetration × spend or price per unit. And there's a bottom-up estimate: the target's own footprint, unit economics, and stated growth plan, rolled forward. CDD adds one step of its own. Build both independently, then reconcile them against each other and against the seller's own plan. Treat any large gap as the most important finding in the whole exhibit, and resist the urge to average it away as an embarrassing rounding error.

The seller's business plan is, in effect, a third bottom-up estimate, built by someone with every incentive to make the number look good. Their arithmetic is usually correct, so don't spend your four weeks re-checking their sums. You're testing whether their assumptions survive contact with an independently built top-down ceiling.

Three moves keep this reconciliation from turning into a box-ticking exercise.

First, size the segment the target can actually serve. A national total-addressable-market figure is close to meaningless for a target that operates in one region or one price tier. Build the regional or segment-level serviceable-addressable-market (SAM) figure that the target's own footprint implies, and compare the bottom-up plan against that figure instead of the national headline.

Second, check the bottom-up plan's implied share of the SAM as well as its absolute growth rate. A plan with healthy double-digit annual growth can still imply capturing an implausible fraction of the addressable market once you do the division. A diligence team is paid to run that division. A sell-side deck is never structured to invite it.

Third, where the deal spans two markets at different stages of maturity (a domestic base plus an international expansion thesis, say), build a comparator market model for the second geography. Don't assume the domestic multiple simply transplants. The mechanics travel. The maturity curve, penetration rate, and even the unit of measurement often don't, and Worked Example 4 below walks through that contrast.

{
  "skill": "Accelerating market-sizing desk research for a CDD market model",
  "use": "Have an AI research tool draft the market-sizing structure and point you to named public sources for penetration rates, population or unit counts, and spend-per-unit benchmarks — then you build and defend the actual model yourself.",
  "tools": ["Research/agentic tools — a browsing-enabled assistant", "Chat assistants for structuring the model's logic"],
  "prompt": "I'm sizing the <MARKET> market top-down for a commercial due diligence. Do not state a market-size figure you cannot source. Instead, list the exact named sources (industry association, government statistical office, a named research firm) where I could find (a) the unit or population count, (b) the penetration or incidence rate, and (c) average spend per unit — and show me the formula to combine them.",
  "verify": "Open every named source yourself and pull the actual figure — an AI's market-size number is a plausible-sounding guess dressed as research unless every input traces to a dated, named source (Guardrail #1 Trace, #4 Date-check). Recompute the combination yourself; never paste an AI-stated total straight into an exhibit.",
  "diy": "You must already know how to build and reconcile a top-down and a bottom-up estimate by hand (CN2.02) — AI only accelerates finding the input numbers, never the arithmetic, and never the judgment on which range to trust.",
  "market": "IN"
}

6. The competitive teardown: finding the mechanism, not the logo count

A competitive teardown that lists competitors' headcounts, funding, and marketing taglines is desk research. Diligence starts when you isolate the mechanism that produces the target's current performance relative to peers. Then you test whether that mechanism is durable, which would make it a moat, or merely a current-period artifact of scale, timing, or a market not yet fully competed away. Build the comparator table around the two or three metrics that would move your recommendation. Those are the ones the thesis depends on, so skip the generic list. Then put the same question to every large gap between the target and its peers: what, specifically, produces this gap, and could a well-funded competitor close it?

The most common trap is treating raw scale (more locations, more customers, more revenue) as self-evidently a moat. The meta-skill of spotting when a framework misleads exists to inoculate you against this one. Scale only becomes a moat when it converts into a mechanism a smaller rival can't replicate just by raising capital: shared fixed costs across a dense local footprint, a data or network effect, a switching cost, a regulatory position. Worked Example 5 below shows this with real arithmetic. Two clusters of the same target's own locations, nearly identical in every visible respect, produce meaningfully different margins purely because one is geographically dense and the other is scattered. Where a moat exists at all, it comes from the density and not from the total location count. A roll-up thesis that keeps adding scattered locations because "bigger is better" is wearing away the very mechanism that made the first cluster work.

Win-loss evidence from the VoC program belongs in the teardown. Don't file it separately. When a customer explains in their own words why they chose the target over a named competitor, that's direct evidence of the mechanism a spreadsheet comparison can only infer. Even more valuable is a lapsed customer explaining why they left for a named competitor. Put the quantitative benchmarking table next to a short synthesis of what VoC heard about why customers choose one way or the other, and the teardown is worth far more than either piece alone.

{
  "skill": "Building a first-pass competitor benchmarking table for the teardown",
  "use": "Ask an AI research tool to assemble a first-draft comparison table of named competitors' public metrics as a starting scaffold you then verify line by line — never as the finished exhibit.",
  "tools": ["Research/agentic tools", "Chat assistants for drafting the comparison structure"],
  "prompt": "Build a comparison table for <TARGET> against these named competitors: <LIST>. For each, include only figures you can attribute to a specific named public source (a filing, a press release, a reputable trade-press article), and mark anything you are inferring rather than sourcing as 'estimated — unverified.'",
  "verify": "Every cell must be traced to its named source and date-checked before it enters an exhibit (Guardrail 1-5) — a competitor's private, non-public financials are not knowable to a bare AI tool, and a plausible-looking number in that cell is exactly the fabrication risk this program has drilled you to catch. Recompute any derived metric (revenue per location, growth rate) yourself from the sourced raw figures.",
  "diy": "You must already know which dimensions matter for THIS teardown — density, mix, growth quality (this Core section) — before asking AI to fill a table; otherwise you get an impressive-looking table answering the wrong question.",
  "market": "US"
}

7. The financial-quality bridge: commercial reading, not financial re-audit

CDD touches the financials, but only through a commercial lens, and the boundary matters. Financial due diligence normalizes EBITDA, tests working-capital swings, and audits accounting policy. Commercial due diligence asks whether the revenue itself is the quality the growth rate suggests. Three checks do most of the work.

A same-store (like-for-like) bridge separates growth from comparable, mature operations from growth contributed by new locations, new customers, or recent acquisitions. You need it because a blended growth rate can look healthy for years while the mature base slows underneath a stream of fresh additions that refresh the average. Worked Example 6 below is built entirely around catching this.

A cohort or retention read asks whether customers acquired eighteen months ago spend more, the same, or less than they did in month one. It tells you whether growth compounds or churns, using the cohort vocabulary from the business-model teardown.

A customer-concentration check asks what share of revenue comes from the single largest customer, and from the top five. It flags a fragile revenue base that financial due diligence's normalized-EBITDA number will never surface on its own. Concentration is a commercial fact about who pays, and the accounting numbers only tell you how much.

One question keeps this lens grounded. Ask it of every striking number: what would VoC need to confirm or contradict this? A revenue-quality question is very often a customer-behavior question in a financial disguise. The fastest way to answer it is the right ten phone calls. A longer spreadsheet is the slow way round.

8. Synthesizing to a go/no-go: weighing evidence, not averaging opinions

By the final week every lens has produced findings, and the job shifts to synthesis: weighing the evidence into a single, defensible recommendation. A weighted scorecard scores each lens against the evidence gathered and weights it by how much it matters to this specific thesis. It turns five separate work products into one number. It also forces the team to say which lens the deal depends on most. A roll-up thesis should weight execution and financial quality heavily, while a market-timing thesis should weight the market lens heavily.

The scorecard is still a synthesis aid and never the final word, because of the veto rule from stage 2. A hypothesis that met its own stated kill criterion is dead, full stop, however favorable the composite score reads. A team that lets a 4-out-of-5 average talk it past an explicitly killed hypothesis has thrown away the reason for writing the kill criterion down in the first place.

Real CDD conclusions are rarely a clean go or a clean no-go. The most common and most valuable output is a conditional go: proceed, but only with specific, named conditions. That might be a price adjustment reflecting a financial-quality finding, or a retention mechanism (earnout, equity rollover, employment agreement) addressing a key-person risk. It might be a growth plan re-underwritten to reflect a market-model correction.

Building that conditional recommendation takes a risk register. For every material finding, name the risk and assess its likelihood and impact. Then name a specific mitigation, which is the part teams skip most often under time pressure. Vague aspirations don't count. "Customer concentration is a risk" is an observation. "Customer concentration is a risk; mitigate with a price adjuster tied to top-five-customer retention over the first 12 months post-close" is a risk-register row a partner can act on.

Keep dealbreakers and priceable risks explicitly apart. A dealbreaker is a finding that killed a hypothesis, and no mitigation rescues it. A priceable risk lowers value or needs a specific contractual protection, but it doesn't kill the deal on its own. Confusing the two in either direction, by pricing a dealbreaker or by killing a deal over a priceable risk, is the most consequential synthesis error a CDD team can make.

{
  "skill": "Drafting a first-pass risk register and the skeleton of the go/no-go memo",
  "use": "Use AI to draft a starter risk register (risk / likelihood / impact / mitigation rows) from your own findings notes, and a first skeleton of the answer-first memo — both as edited drafts you rewrite in your own judgment, never as the submitted output.",
  "tools": ["Chat assistants — Claude, ChatGPT", "Word-processor AI drafting features"],
  "prompt": "From these due-diligence findings notes <PASTE YOUR OWN NOTES>, draft a risk register with columns risk / likelihood / impact / mitigation, and a one-paragraph answer-first governing thought in the Minto style (answer, then three supporting points, then the key risk). Do not add any risk, figure, or finding that is not already in my notes.",
  "verify": "Check every row against your own underlying evidence — the VoC finding, the market model, the teardown — before it survives into a deliverable. An AI draft can quietly invent a plausible-sounding risk that was never actually in your findings, or soften a real red flag into vague language. You, not the AI, own the final go/no-go call and every number in the memo (Guardrail #3, Recompute, applies to every derived figure in the register).",
  "diy": "You must already be able to weigh the evidence and make the go/no-go call yourself — the scorecard discipline this section teaches. AI only removes the drafting friction of turning your own conclusions into clean rows and clean prose.",
  "market": "IN"
}

9. From verdict to memo: the hand-off

The CDD's final deliverable is an answer-first memo. Its first line is the governing thought, and that line is the recommendation itself. Three or four supporting points follow, each mapping directly onto the lenses' evidence, then the material risks with their mitigations and a recommended next step. That's the Minto shape you used for answer-first synthesis, and the communication toolkit later extends it into a full document and deck. Here you work only on the logic of the answer: which sentence goes first, what counts as sufficient support, how to decide whether a risk belongs in the top line or in an appendix. Slide layout, visual storylining and presentation delivery belong to the Presentation galaxy, and that hand-off is deliberate.

A CDD team that nails every lens's analysis and then buries the recommendation on page nine of the appendix has wasted four weeks of good work on the one page that mattered.

Common mistakes & how experts think differently

1. Treating VoC as confirmation shopping. Interviewing only customers the seller introduces, or only the happiest accounts, produces a sample built to agree with the thesis. The expert insists on building at least half the sample independently and compares the two halves' answers explicitly. Any gap between them is itself the finding.
2. Leading questions that manufacture the answer. "You've been happy since the ownership change, right?" invites agreement regardless of the truth. The expert anchors every question to recalled behavior: what did you actually buy, from whom, and why? Behavior is far harder to bias than opinion.
3. Small-n overconfidence. Drawing a go/no-go conclusion from four interviews that happened to agree. The expert treats anything under roughly eight to ten interviews per segment as a hypothesis that still needs corroboration before it can anchor a recommendation.
4. Confusing an expert's opinion with fact. An expert call is triangulation and isn't evidence on its own. One confident expert is still one data point. A good analyst always asks how the expert knows what they claim to know, and corroborates it before it enters the memo.
5. A market model with no reconciliation. Building only a bottom-up plan (or only a top-down ceiling) and never checking one against the other lets an aggressive growth assumption sail through unchallenged. The expert always builds both, independently, and treats the gap between them as the headline finding instead of a rounding error.
6. Ignoring share-of-wallet. Assuming 100% of a customer's category spend flows to the target when real customers split their basket with competitors or informal channels. The expert uses VoC to ask directly what share of the relevant spend the target captures, then rebuilds the market model around the answer, as Worked Example 8 does.
7. Letting a roll-up's blended growth rate hide a decelerating base. Freshly acquired locations post high growth off a small, ramping base and mathematically flatter the headline number while the mature core stalls. The expert always splits growth by cohort age before trusting a blended figure.
8. Treating commercial DD as financial DD, or ignoring financials entirely. Re-auditing the books duplicates a workstream that already exists. Ignoring financial-quality signals entirely means missing the customer-concentration or same-store story the numbers are trying to tell. The expert reads the financials commercially, asking what each number says about customer behavior, and hands anything forensic to the financial-DD team.
9. A thesis with no kill criterion. "The market is attractive" can't be falsified by anything, so it's a hope dressed up as a hypothesis. The expert writes down the specific number or finding that would kill each hypothesis before the research starts, while the answer is still unknown.
10. Confusing scale with a moat. Assuming "we have the most locations" is automatically defensible without testing the actual mechanism. The expert tests whether scale converts into something a well-funded rival can't replicate (density, switching costs, a data effect), and treats bare scale, on its own, as no moat at all.
11. A risk register with no mitigation, or no owner. A list of worries with a likelihood column and an impact column isn't a risk register yet. It becomes one only when every row also names a specific action and, ideally, who is accountable for it. The expert refuses to close a diligence with any row still reading "monitor."
12. Letting a favorable composite score overrule a killed hypothesis. A weighted scorecard is a synthesis aid. It can't hand down a verdict, and if a hypothesis met its stated kill criterion, no average rescues the recommendation. The expert states the veto rule out loud before the scoring begins, so nobody can talk their way around an inconvenient result once the numbers are in front of them.

Worked examples

Every company, fund, and figure below is invented for teaching (synthetic) and stated as such. Every calculation was recomputed and is shown so you can check it yourself. Worked Examples 1 to 7 follow one target all the way through, and Worked Example 8 is a separate India vignette.

Worked example 1: The CDD thesis and its kill criteria (US, $)

Cordillera Capital (a synthetic mid-market private-equity fund) has four weeks of exclusivity to diligence BrightPath Veterinary Partners (synthetic). BrightPath is a 42-clinic veterinary-services platform across the US Mountain West and Desert Southwest, currently owned by an independent sponsor exploring a majority recapitalization. Its last-twelve-months (LTM) revenue is $210 million on $31.5 million of EBITDA, a 15.0% margin. The sponsor's asking price implies roughly 11× EBITDA, an entry enterprise value of $346.5 million.

Cordillera's deal team likes the space. Veterinary services is large, non-cyclical, and fragmented, with room to keep buying small independent clinics ("tuck-ins") at a lower multiple than the platform trades for. That's the classic multiple-arbitrage roll-up thesis. Once a "go" is reached, the deal-rationale work will stress-test it as a synergy case. Your job here is only to test whether the commercial premise holds.

Solution. Step 1, write the thesis as claims that can be wrong. "We like veterinary services" becomes four falsifiable hypotheses:

• H1 (durable growth): Same-store revenue growth, stripped of the effect of recent acquisitions, is durably at or above the low-single digits. Kill criterion: legacy-clinic same-store growth below 2% once tuck-ins are excluded.
• H2 (client loyalty survives ownership change): Clients' loyalty is to the clinic/brand, not solely to an individual veterinarian who could leave and take the relationship. Kill criterion: VoC shows a large share of clients say they would follow their vet to a new employer if the vet left.
• H3 (sourcing discipline holds): Tuck-in acquisitions can be sourced at a sustained multiple gap against the platform multiple. Kill criterion: the realistic near-term pipeline shows tuck-in multiples have already drifted to within roughly two turns of the platform's own entry multiple.
• H4 (white space remains): The regional market is not already substantially consolidated by well-capitalized rivals who will out-bid Cordillera for future tuck-ins. Kill criterion: the two largest national consolidators already hold the majority of the organized (multi-site) segment in BrightPath's operating footprint.

Step 2, read what the kill criteria buy you. Each one is a specific number or finding, decided before a single interview takes place. Deciding them first is what separates diligence from motivated reasoning. Notice, too, that none of the four hypotheses is really about the market in the abstract. Every one asks whether this specific roll-up mechanism (buy cheap, integrate, keep client loyalty, avoid a bidding war) actually works. That's the bespoke tree from Core section 3 in action. A generic "is veterinary services attractive?" issue tree would miss H2 and H3 entirely, and those two are where the serious risk is.

Worked example 2: The key-questions issue tree and workplan (US, $)

Solution. Step 1, build the MECE tree across the five lenses, mapping each hypothesis to where it will be tested:

• Market (H1, H4): size the US veterinary-services market top-down and bottom-up, and assess how consolidated BrightPath's specific footprint already is.
• Competitive position (H4): benchmark BrightPath against national consolidators and fragmented independents, and test whether the mechanism of margin is density or raw scale.
• Customer (H1, H2): run VoC across legacy clients, recently acquired (tuck-in) clients, and referring veterinarians to test client loyalty directly.
• Financial quality (H1): build a same-store bridge splitting legacy from tuck-in growth and margin, to check whether the headline blended growth rate is hiding a decelerating base.
• Execution/organizational risk (H2, H3): assess the sourcing pipeline and realistic tuck-in pricing, and flag key-veterinarian dependency at specific clinics. Also flag state-level regulatory structuring. Several US states restrict or closely regulate non-veterinarian corporate ownership of veterinary practices, typically requiring a licensed-veterinarian-owned professional entity paired with a management-services agreement. Legal diligence confirms that structuring fact state by state. It isn't itself a commercial red flag.

Step 2: staff and sequence it. With four weeks, the team front-loads VoC and the market model, because both take calendar time to set up: sourcing interviewees, reaching experts. Desk research on competitors and the financial-quality bridge run in parallel from data-room material already in hand. The tree is the one page the partner reviews before week one starts, and everything in Worked Examples 3 to 7 is that tree being carried out.

Worked example 3: Designing the Voice-of-Customer and expert-call program (US, $)

Solution. Step 1, build the sampling frame across segments that could tell different stories. The team plans 64 conversations in total:

• 25 legacy-clinic clients (owned 5+ years)
• 25 tuck-in-clinic clients, acquired within the last 24 months, to test H2 directly by comparing pre- and post-acquisition sentiment
• 8 referring/emergency-vet partners
• 3 departed BrightPath or competitor veterinarians or managers, as expert calls
• 2 independent-clinic owners approached as potential future tuck-in targets, to sanity-check H3's sourcing pipeline and real-world pricing
• 1 state veterinary-board licensing expert, for the regulatory-structuring question from Worked Example 2

Step 2, catch a sponsor-bias problem before it corrupts the finding. The first 25 legacy-clinic client calls, all with names the BrightPath regional manager suggested, come back at 92% satisfaction. For any consumer-facing service business, that's an almost implausibly clean number. So the team demands an independent supplemental sample of 40 clients drawn directly from the CRM export, stratified by visit recency instead of by manager recommendation. That random-dial sample comes back at 78% satisfaction, fourteen points lower. It's also the sample that first surfaces the number that matters most. 34% of respondents said they would follow their treating veterinarian to a new employer if that veterinarian left. The gap between 92% and 78% is itself a finding. Built entirely from management referrals, the first sample had been painting a rosier picture than the client base actually holds.

Step 3, read the loyalty number against H2's kill criterion. Thirty-four percent choosing the vet over the brand is a real, quantifiable risk. The legacy clinics carry $180 million of revenue. One mechanical way to frame the exposure, deliberately illustrative and an upper bound, is 34% × $180m ≈ $61.2 million of revenue with clients whose primary loyalty runs to an individual instead of the clinic. That's a lot of revenue resting on individual relationships. Still, it falls short of H2's stated kill criterion. That criterion was written in advance to require "a large share", meaning loyalty that is not clinic-based at all. At 34% you're looking at a real, material, but mitigatable risk, one that retention agreements for high-risk veterinarians can address. It isn't evidence that the whole roll-up mechanism is broken.

This is the veto rule from Core section 8 working correctly in reverse. The hypothesis survives, but the finding gets a specific, named row in the risk register built in Worked Example 7. It doesn't get buried just because it didn't quite kill anything.

Worked example 4: Building and reconciling the market model, with an India comparator (US $, India ₹)

Solution. Step 1: size the US market top-down. The figures here are illustrative and synthetic. Before you use any real number, verify current pet-population and spend data against a named source such as an industry association or a paid research firm. Take a US pet population of about 175 million. About 65% of them get at least one veterinary visit a year, so 175m × 0.65 = 113.75 million pets seen. Average annual veterinary spend per treated pet is about $350. Top-down US TAM = 113.75m × $350 = $39.8 billion, consistent with the "$40+ billion" figure in the original thesis (Worked Example 1).

Step 2, narrow to the segment BrightPath can actually serve. BrightPath's real addressable category is only the organized share of the market: multi-site, non-independent operators. Illustratively that's ≈25% of the total, giving a national organized-segment SAM of $39.8bn × 0.25 = $9.95 billion. But BrightPath operates in six states holding roughly 8% of the US population. Scale the national organized SAM down to that footprint and you get a regional organized SAM of $9.95bn × 0.08 ≈ $796 million. Call it "roughly $800 million." The bottom-up plan has to be checked against this regional figure. The $40 billion national headline is the wrong yardstick.

Step 3, build the bottom-up plan and reconcile. BrightPath's own five-year plan grows the existing 42 clinics at 4% annually, matching the blended same-store rate. It also adds 15 new tuck-ins over five years, each averaging $4 million of revenue at acquisition and maturing roughly 2.5 years by year five. Existing clinics: $210m × 1.04⁵ ≈ $255.5 million. New tuck-ins: 15 × $4m × ~1.10 (growth factor for an average 2.5 years' ownership at 4%/yr) ≈ $66.0 million. Total year-5 bottom-up platform revenue ≈ $255.5m + $66.0m = $321.5 million.

Step 4, the reconciliation is the finding. Against the national organized SAM of $9.95bn, $321.5m is a modest 3.2% share. It's unremarkable and easy to wave through. Now divide by the regional organized SAM of ~$796m, which is the number that matters because BrightPath's tuck-in pipeline is geographically concentrated. On that base, $321.5m is roughly 40% of the entire regional organized segment. An aggressive assumption was hiding inside a plan that looked conservative at the national level. If your market model only ever checks the flattering denominator, it will never surface a gap like that. The corrected read, carried into Worked Example 7, is that the five-year buy-and-build pace, as planned, likely assumes away more competition for tuck-ins than the regional market can absorb.

Step 5, the India comparator, as a cross-market sanity check, not a second deal. The team also sizes the analogous opportunity in India, where organized multi-city veterinary chains are only just emerging. Illustratively, India has ~40 million pets, of which only ~20% receive any veterinary care in a year (8 million pets seen), at an average annual spend of ₹4,000 per treated pet. India TAM = 8m × ₹4,000 = ₹32,00,00,00,000, i.e., ₹3,200 crore (≈$385 million at ~₹83/$). Cordillera has no plans to enter this market, so treat the figure as a directional reminder that links back to the market-attractiveness lens. The same mechanics (population × incidence × spend) produce wildly different conclusions depending on how mature and how organized the underlying category already is. Every CDD market model has to make that market-maturity judgment explicitly. Never leave the choice of country to imply it.

Worked example 5: The competitive teardown and the density moat (US, $)

Solution. Step 1: build the comparator table. BrightPath runs 42 clinics on $210m revenue, or $5.0m revenue/clinic, at a 15.0% EBITDA margin. The team benchmarks it against two named-synthetic rivals and the independents:

• VetCore Holdings, a larger, more mature national consolidator: 220 clinics, $1.32bn revenue, $6.0m revenue/clinic, 19.0% EBITDA margin ($1.32bn × 0.19 = $250.8m EBITDA).
• PawPoint Health, a newer, aggressively PE-backed roll-up still absorbing integration costs: 90 clinics, $360m revenue, $4.0m revenue/clinic, 12.0% EBITDA margin ($360m × 0.12 = $43.2m EBITDA).
• Fragmented independents, whose owner-operators typically post lower revenue per clinic but higher margin. They carry no corporate overhead, but they face a hard ceiling on growth and a real succession-planning problem, and that problem is where Cordillera's tuck-in supply comes from.

Step 2: resist the scale-equals-moat reflex. BrightPath's 42 clinics run at a 15% margin, below VetCore's 19% at more than five times the clinic count. It's tempting to conclude "we just need more scale." Instead, split BrightPath's own legacy clinics (30 of them, $180m revenue, 16% margin) by geography. The Denver-metro cluster has 7 clinics, $42m revenue and a 19% margin, i.e., $7.98m EBITDA. The remaining 23 scattered, single-clinic-metro legacy clinics have $138m revenue. Total legacy EBITDA is $180m × 0.16 = $28.8m. Take away the Denver cluster's $7.98m and you're left with $20.82m on $138m, a 15.1% margin.

Look at what's held constant: same brand, same ownership, same corporate overhead allocation. Yet there's a full ~4 percentage-point margin gap between the dense cluster and the scattered clinics, and geographic density produces all of it. Clinics within driving distance of each other share on-call coverage, specialist referral relationships, and back-office reach. On this evidence, VetCore's 19% margin advantage over BrightPath's blended 15% is much more plausibly a density story than a scale story. VetCore likely has more Denver-style clusters, and not simply more total clinics.

Step 3: connect to the thesis. This feeds straight into H4 and into the buy-and-build plan corrected in Worked Example 4. The highest-value use of tuck-in capital is tuck-ins within driving distance of BrightPath's existing clusters, because they deepen density. The geographically broadest expansion would only chase a scattered logo count. That insight came from this specific data set. Reciting "scale is a moat" on faith would never have produced it.

Worked example 6: The financial-quality bridge: what VoC caught that the P&L alone would not (US, $)

Solution. Step 1, split the headline growth rate by cohort age. BrightPath's reported LTM same-store growth is a healthy-looking 4.0%. Split revenue into the 30 legacy clinics ($180m, owned 5+ years) and the 12 tuck-in clinics acquired within the last 24 months ($30m), and compute each cohort's own same-store growth. Legacy clinics grew +1.0%. Tuck-in clinics grew +22.0%, a post-acquisition "honeymoon" ramp: new marketing, extended hours, referral-network integration. Weighted check: (180 × 0.01 + 30 × 0.22) ÷ 210 = (1.8 + 6.6) ÷ 210 = 4.0%, which ties to the reported headline.

Step 2, read what the blend was hiding. The 4.0% headline is entirely an artifact of mix. The mature, representative 86% of the business, the legacy clinics, is slowing toward flat. Meanwhile a small, temporarily fast-growing 14%, the tuck-ins still riding their acquisition ramp, is propping up the average. This is the "acquisition sugar high" that Core section 7 and Common Mistake 7 describe. It's the most consequential financial-quality finding in the whole diligence. A roll-up that keeps adding fresh tuck-ins can keep the blended growth rate looking fine for years while the core business stalls, as long as new acquisitions refresh the mix faster than the legacy base decelerates.

Step 3, corroborate with VoC as well as arithmetic. The spreadsheet wasn't where this finding started. Two results together prompted it: the 34% "would follow my vet" result from Worked Example 3, and the observed legacy deceleration. They led the team to ask a direct follow-up question in the remaining legacy-clinic interviews. Several long-tenured clients volunteered, unprompted, that "the clinic feels more corporate now". They also said visit frequency had dropped since two well-liked, longer-tenured veterinarians left over the prior 18 months. The financial-quality number (legacy same-store at 1%) and the VoC evidence (a client-perceived service change, correlated with the departures) reinforce each other. Cross-lens corroboration like this is what a CDD is built to produce. Working from normalized EBITDA alone, a financial-DD team would have had no natural way to surface it.

Worked example 7: The go/no-go scorecard and risk register (US, $)

Solution. Step 1, score each lens against its evidence, weighted by what this specific thesis depends on most. Weights: market attractiveness 25%, customer loyalty 20%, competitive position 20%, financial quality 20%, execution/sourcing 15% (sums to 100%). Scores run from 1 to 5, with 5 the strongest:

• Market scores 4. It's large and structurally growing, per Worked Example 4, but the regional-SAM check flags an aggressive assumption, so it doesn't get a 5.
• Customer loyalty scores 3. The corrected 34% "follow the vet" finding from Worked Example 3 is real and material, but it's short of H2's kill criterion.
• Competitive position scores 4. The density mechanism in Worked Example 5 is real and defensible where clusters exist, and materially weaker where clinics are scattered.
• Financial quality scores 2. Worked Example 6's legacy deceleration is the most serious finding in the diligence.
• Execution/sourcing scores 3. Expert calls suggest tuck-in multiples have crept from roughly 5 to 6× historically toward 7×. It stays below H3's 8× kill threshold, but it's a real erosion of the arbitrage and worth watching.

Composite = 0.25(4) + 0.20(3) + 0.20(4) + 0.20(2) + 0.15(3) = 1.00 + 0.60 + 0.80 + 0.40 + 0.45 = 3.25 / 5 (65%).

Step 2, apply the veto rule before reading the composite as a verdict. None of the explicit kill criteria for H1 to H4 was actually met. Legacy growth is weak but not below the 2% kill line. The loyalty figure is concerning but below the "large share" kill bar. Tuck-in multiples have crept but remain under the 8× kill threshold, and no single consolidator dominates the regional footprint. So no hypothesis is dead. The 65% composite falls below a typical 70% straight-"go" threshold, and it correctly signals a conditional go, short of a clean pass.

Step 3, build the risk register: risk, likelihood, impact, mitigation.

Step 4, the recommendation. "Conditional go: proceed, but only with a purchase-price adjustment reflecting the legacy same-store deceleration, retention agreements for the highest-risk veterinarians, a tuck-in underwriting cap of 7× rather than 6×, and a re-cut five-year plan built against the regional organized-market SAM rather than the national total." Every clause in that sentence traces to a specific worked example above, and that's what Worked Example 7 exists to show. A synthesis is the evidence, weighed and dated, with no fresh opinion added.

Worked example 8: Designing Voice-of-Customer for an India commercial due diligence (India, ₹)

Kavery Growth Partners (a synthetic India-focused growth-equity fund) is evaluating a minority growth investment in Saffron Grocers (synthetic). Saffron is an organized grocery retailer running a hybrid of quick-commerce dark stores in Tier-1 metros and small-format supermarkets in Tier-2 cities, and it wants capital to fund an eight-city Tier-2 expansion.

Solution. Step 1, segment before dialing a single number. Three different customer stories could be hiding in one blended sample. Tier-1 dark-store users are younger and convenience-driven, and less sensitive to price. Tier-2 supermarket-format customers are more price-sensitive, and often still loyal to a neighborhood kirana relationship built over years. Lapsed users have churned back to a kirana store or to a competing app. The team plans 40 Tier-1 interviews, 30 Tier-2 interviews, 15 lapsed-user interviews, and 5 expert calls, 90 conversations in total. The five experts are a former regional category manager, two FMCG distributors supplying both Saffron and local kirana stores, a quick-commerce logistics specialist, and a Tier-2 retail-real-estate expert on dark-store site economics.

Step 2, adapt method to reach, not habit. In Tier-2 cities, in-app survey response rates run far lower than in the Tier-1 metros where the app-native customer base is concentrated. So the team runs structured WhatsApp-based surveys alongside phone interviews. WhatsApp is the dominant reach channel for a broad, cost-effective consumer sample across Indian cities of every tier. An online-panel-first approach would systematically under-reach the Tier-2 segment the expansion thesis depends on. The team also fields the discussion guide in each customer's language of choice instead of defaulting to English.

Step 3: fix the leading question. The draft guide opens with "Saffron's ten-minute delivery has made your life so much easier, hasn't it?" That's a textbook leading question (Core section 4, Common Mistake 2). It invites agreement and tells the interviewer nothing about actual behavior. The fixed, behavior-anchored version asks "Thinking about your last five grocery purchases, which store or app did you use for each, and why?" You get a revealed-preference account of where spend actually goes, and it becomes the most important input to the market model.

Step 4, the finding that corrects the market model. Take the "active" Tier-2 sample, defined as customers with at least one Saffron order in the measurement month. 60% had also shopped at their local kirana store within the same month, so only 40% had fully switched their grocery basket to Saffron. And these were customers the company's own dashboard would classify as loyal, active users. Saffron's internal business plan implicitly assumed something closer to full wallet capture among active users.

Here's the corrected market model. Target-city households ≈ 6 million. The share in Saffron's addressable socio-economic segment ≈ 35%, i.e., 6,000,000 × 0.35 = 2,100,000 addressable households. Average annual grocery spend per household ≈ ₹8,000/month × 12 = ₹96,000. The naive (100%-wallet) serviceable market = 2,100,000 × ₹96,000 = ₹2,01,60,00,00,000, i.e., ₹20,160 crore. Now apply the VoC-observed wallet-share ceiling of roughly 42%, the midpoint of the 40 to 45% range the interviews suggested. Corrected SAM = ₹20,160 crore × 0.42 = ₹8,467 crore. That's well under half the naive figure. It's the number Kavery's team should underwrite the Tier-2 expansion against, in place of the company's own optimistic plan.

Read across Worked Examples 4 and 8 together. In both markets the mechanics of "reconcile the model against real customer behavior" are identical. Only the specific correction differs: a geographic-concentration check in one case, a share-of-wallet check in the other. The toolkit travels, even though the numbers never do.

Practice set

Work each problem fully before you read its solution. Every company and figure is synthetic, and numeric answers are exact given the stated inputs.

P1 (guided). CDD versus financial DD. In one or two sentences, state the question a commercial due diligence answers and contrast it with financial due diligence's question.

Solution. CDD asks, from the buyer's chair, whether the business will perform commercially the way the deal model assumes: is the market real, is the customer base loyal, does the growth hold up? Financial DD asks whether the reported numbers themselves are accurate and sustainable (normalized EBITDA, working-capital normalization, accounting policy). That's a books-and-records question, where CDD's is a market-and-customer question.

P2 (guided). Writing a kill criterion. A sponsor's thesis is "this SaaS company's expansion revenue is strong, so net retention is healthy." Write one falsifiable hypothesis with an explicit kill criterion.

Solution. Hypothesis: "Net revenue retention is durably above 100% because existing customers expand, not merely because a shrinking base of survivors happens to expand while many others silently churn." Kill criterion: gross logo churn among the cohort acquired 12 to 24 months ago exceeds, say, 15% annually. That finding would mean the healthy blended NRR figure is hiding serious churn beneath expansion revenue from the customers who remain.

P3. Building a bespoke issue tree. A CDD thesis is "this coffee-chain roll-up can double store count in three years by replicating the flagship city's unit economics nationally." Name the one lens this thesis most depends on beyond the standard five, and why a generic tree would under-weight it.

Solution. Execution/organizational risk carries unusually high weight here. The entire thesis rests on a replication claim: that the flagship city's unit economics (site selection, staffing, supply chain) can be executed consistently in new, unfamiliar markets. A generic tree gives execution its standard weight. A bespoke tree recognizes that if replication capability is weak, every other lens's favorable finding (attractive market, strong brand) is moot, because the company can't deliver the plan.

P4. Same-store bridge, solve for the missing rate. A retailer reports $150 million total revenue and 5.0% blended same-store growth. Legacy stores ($120 million) grew 2.0%. What growth rate did the newer stores ($30 million) post?

Solution. Blended growth × total = Σ(segment revenue × segment growth): 0.05 × 150m = $7.5m. Legacy contribution: 120m × 0.02 = $2.4m. Newer-store contribution required: $7.5m − $2.4m = $5.1m. Newer-store growth = $5.1m ÷ $30m = 17.0%. The gap against legacy's 2% is large and worth investigating as a possible acquisition-ramp effect like the one in Worked Example 6.

P5. Top-down market sizing. A category has 50 million potential buying units, 40% annual incidence, and $200 average annual spend per active unit. What is the top-down TAM?

Solution. Active units = 50,000,000 × 0.40 = 20,000,000. TAM = 20,000,000 × $200 = $4.0 billion.

P6. Reconciling bottom-up against two denominators. A target's five-year bottom-up plan reaches $360 million of revenue. The national SAM for its category is $9 billion; the target operates in a region holding 5% of national population. Compute the plan's implied share of (a) the national SAM and (b) the regional SAM, and state which one should actually worry the diligence team.

Solution. (a) National share = $360m ÷ $9,000m = 4.0%, which looks entirely unremarkable. (b) Regional SAM = $9,000m × 0.05 = $450m, so regional share = $360m ÷ $450m = 80.0%. The implied capture of the regional market the target can actually address is extremely aggressive. The regional figure is the one that should worry the team, as it did in Worked Example 4. A plan can look conservative against the wrong (too-large) denominator and aggressive against the right one.

P7. Spotting the bias. A diligence team's customer sample is built entirely from a list the target's VP of Sales personally emails over. What bias is this, and what is the fix?

Solution. Sponsor/referral bias. The party being evaluated controls who gets interviewed, which structurally selects for the target's best relationships. Fix: build at least half the sample independently (a CRM export, public customer list, or cold outreach) and compare the referred half's answers against the independent half's explicitly. Any gap between them is itself a finding, as it was in Worked Example 3.

P8. Fixing a leading question. Rewrite this question to remove the leading bias: "You'd say the new pricing has been fair, wouldn't you?"

Solution. A behavior-anchored fix: "Walk me through the last time you compared this product's price to an alternative, what did you compare it to, and what did you decide?" It asks for a recalled decision instead of inviting agreement with a stated opinion, and it's far harder to bias toward a pleasant answer.

P9. Weighted scorecard. Weights: market 30%, customer 25%, competitive 20%, financial quality 25%. Scores: market 5, customer 3, competitive 4, financial quality 2. Compute the composite (out of 5) and as a percentage.

Solution. Composite = 0.30(5) + 0.25(3) + 0.20(4) + 0.25(2) = 1.50 + 0.75 + 0.80 + 0.50 = 3.55 / 5 = 71.0%.

P10. The veto rule. A CDD scorecard reads a healthy 4.1 out of 5, but one hypothesis's stated kill criterion was, in fact, triggered by the evidence. What is the correct recommendation, and why?

Solution. No-go, or at best a go only after the killed hypothesis's underlying problem is structurally fixed pre-close. Pricing it in isn't enough. A kill criterion exists to stop a favorable composite from overruling a specific, decisive piece of contrary evidence. Treat the 4.1 as dispositive and the kill criterion was never binding, which defeats the purpose of writing one down in advance (Core section 8, Common Mistake 12).

P11. Wallet-share market-model correction. A company's naive (100%-wallet) serviceable market is ₹5,000 crore. VoC interviews find that even loyal, "active" customers only route 55% of relevant category spend to the company. What is the corrected SAM?

Solution. Corrected SAM = ₹5,000 crore × 0.55 = ₹2,750 crore. It's the same kind of share-of-wallet correction that Worked Example 8 applied to Saffron Grocers.

P12. Building a risk-register row. Given the finding "the target's largest customer accounts for 28% of revenue, and its contract renews in six months," write one complete risk-register row (risk / likelihood / impact / mitigation).

Solution. Risk: customer-concentration exposure on a near-term renewal. Likelihood: Medium (no specific evidence of dissatisfaction, but a large single-customer dependency is fragile by nature). Impact: High (28% of revenue at risk in a single event). Mitigation: negotiate a price-adjuster or earnout tied to confirmed renewal of the top customer before final consideration is released. Also verify the relationship's health independently, with a direct (non-seller-arranged) call to that customer during diligence, before close.

P13. Density margin premium. Two clusters of the same chain post identical $80 million revenue each. Cluster A (dense, clustered locations) runs a 22% EBITDA margin; Cluster B (scattered) runs 14%. Compute the EBITDA gap in both percentage points and dollars.

Solution. Margin gap = 22% − 14% = 8.0 percentage points. Cluster A EBITDA = $80m × 0.22 = $17.6m. Cluster B EBITDA = $80m × 0.14 = $11.2m. Dollar gap = $17.6m − $11.2m = $6.4 million on identical revenue, all of it attributable to density. It's the same mechanism Worked Example 5 isolated.

P14. Expert-network ethics. During an expert call, a currently-employed competitor manager offers, unprompted, to share their employer's unreleased pricing roadmap. What should the interviewer do, and why?

Solution. Decline and redirect the conversation back to general market knowledge and the expert's own past experience. Soliciting or accepting a current employee's disclosure of their employer's confidential, unreleased plans is an ethical (and typically expert-network compliance) violation, however useful the information would be. The expert-network ethics in Core section 4 exist so that a diligence team doesn't trade a short-term information edge for a serious professional-conduct breach.

P15. Quantifying a deceleration gap. A $100 million mature revenue base has historically grown at 4% annually but has recently decelerated to 1%. Compute the cumulative three-year revenue gap between the two growth paths.

Solution. At 1%/year for three years: $100m × 1.01³ ≈ $103.03m. At 4%/year for three years: $100m × 1.04³ ≈ $112.49m. Cumulative gap ≈ $112.49m − $103.03m = $9.46 million, or roughly 9.5% of the original base. That gives you a concrete dollar size for what a same-store deceleration like Worked Example 6's costs over the hold period if nobody addresses it.

P16. Dealbreaker versus priceable risk. Classify each as a dealbreaker or a priceable risk, and justify briefly: (a) the target's core patented process turns out to be unenforceable and easily replicated; (b) the top customer contract renews in six months and has historically always renewed.

Solution. (a) Dealbreaker, if the investment thesis specifically depended on IP-based defensibility. No price adjustment fixes a moat that doesn't exist. A finding like this should kill the relevant hypothesis outright instead of just lowering the offer. (b) Priceable risk. It's a real but manageable, quantifiable exposure (Practice P12's row) that a price adjuster, earnout, or closing condition can reasonably address without abandoning the deal.

P17. The answer-first opening line. Given the BrightPath findings across Worked Examples 1–7, write the single opening sentence of the CDD memo (the governing thought, Minto-style: answer first).

Solution. "Recommendation: proceed with a conditional go on BrightPath, the roll-up thesis holds, but only with a price adjustment for the legacy same-store deceleration, veterinarian-retention agreements at the highest-risk clinics, a tighter tuck-in pricing cap, and a growth plan re-underwritten against the regional, not the national, market." Compare it with Worked Example 7's Step 4. The memo's opening line should read as a direct compression of that synthesis, and it shouldn't restate the whole analysis.

Applied mini-project

Run a mini commercial due diligence, start to finish. Choose a business category you find interesting. Any category works: a services roll-up, a consumer brand, a B2B software vendor. Invent a synthetic target and a synthetic acquirer, the way the worked examples do. Never use a real company's real, unpublished financials, and never claim a real diagnosis of a named real business. Produce, in order:

1. The thesis. State the investment thesis in 2 to 3 sentences, then convert it into 2 to 4 falsifiable hypotheses, each with an explicit, numeric kill criterion written before you do any further work.
2. The issue tree. Build a one-page MECE tree across the five lenses (market, competitive position, customer, financial quality, execution/organizational risk), mapping every hypothesis to the lens(es) that will test it.
3. The VoC/expert-call design. Name at least three distinct customer or stakeholder segments and state a sample size per segment. Write one deliberately bad leading question and its behavior-anchored fix, and name which bias (selection, sponsor, small-n, social-desirability) each segment choice is guarding against.
4. The market model. Build a top-down estimate and a bottom-up estimate with invented, internally consistent numbers, and show every step of the arithmetic. Then reconcile them and state explicitly what the gap between them implies.
5. The competitive teardown. Build a comparator table against at least two invented rivals, and identify one specific mechanism (not "scale") that would make the target's position durable or fragile.
6. The go/no-go. Build a weighted scorecard, stating your weights (they must sum to 100%) and your scores, and compute the composite. Apply the veto rule explicitly by stating whether any hypothesis was killed. Then write a risk register of at least five rows (risk / likelihood / impact / mitigation), plus the one-sentence answer-first governing thought.

Self-check rubric. Score each dimension 0/1/2, for 12 points in total. You pass at ≥9 with no zero on any dimension:

• Thesis (0/1/2): hypotheses are falsifiable, with numeric kill criteria stated before analysis (2); stated but vague or not really falsifiable (1); no kill criteria (0).
• Issue tree (0/1/2): MECE, bespoke to the thesis, every hypothesis mapped to a lens (2); reasonable but generic or with a gap (1); a checklist with no connection to the hypotheses (0).
• VoC/expert design (0/1/2): ≥3 real segments, a sample-size logic, a bias correctly matched to each segment choice, and a real leading-question fix (2); most elements present but thin (1); no bias discipline (0).
• Market model (0/1/2): both approaches built, reconciled, and the gap's implication stated explicitly (2); only one approach, or reconciled without interpretation (1); no market model (0).
• Competitive teardown (0/1/2): a specific mechanism identified and tested, not asserted (2); a comparison table with no mechanism (1); a feature list only (0).
• Go/no-go synthesis (0/1/2): weights sum to 100%, veto rule explicitly applied, ≥5 risk-register rows each with a real mitigation, and an answer-first opening line (2); scorecard present but the veto rule or mitigations are missing (1); an opinion with no scorecard or register (0).

Case Lab (when available): this module's mapped Case-Engine case is CC-CDD-01, "Diligencing Bloom", a candidate-led commercial due diligence on a synthetic US D2C brand, run for a private-equity evaluator. Complete this mini-project first; the case then exercises the identical toolkit, cold, on a target you have never seen.

Reading & resources

Use these resources as anchors to learn from and verify against. Never quote one as if it were the target's own data. Each is free unless marked, and level tags run from beginner to advanced.

The thesis and the answer-first synthesis:

• Richard Rumelt, Good Strategy / Bad Strategy. Its diagnose-before-prescribe spine is the logic behind writing a thesis as falsifiable claims instead of a story you already believe. [Paid] [Advanced]
• Barbara Minto, The Pyramid Principle. The memo hand-off in Core section 9 borrows its structure directly: governing thought first, then the key line, then the support. Go here, or to M10.03, for the full mechanics. [Paid] [Advanced]
• Ethan M. Rasiel, The McKinsey Way: a practical, beginner-friendly walk through structured problem-solving under a deal clock. [Paid] [Beginner]

Voice-of-Customer and the customer lens:

• Clayton Christensen and collaborators on the Jobs-to-be-Done framework, widely available via HBR and Christensen's own writing. It's the behavior-first interviewing logic behind Core section 4's habit of asking about the last purchase instead of the opinion. [Mixed free/paid] [Beginner–Intermediate]
• Harvard Business Review, ongoing coverage of qualitative research method and interview bias, a useful general anchor for the bias taxonomy in Core section 4. [Mixed free/paid] [Intermediate]

Competitive teardown and the moat check:

• Michael Porter, Competitive Strategy and Competitive Advantage, with Joan Magretta's Understanding Michael Porter as the accessible on-ramp. Porter supplies the value-chain and industry-structure logic behind Worked Example 5's teardown. [Paid] [Beginner–Advanced]
• Hamilton Helmer, 7 Powers: a sharp, modern vocabulary for naming which mechanism (scale economies, network effects, switching costs, and others) actually produces a durable position, directly useful for distinguishing density from bare scale. [Paid] [Intermediate]
• Bruce Greenwald and Judd Kahn, Competition Demystified, the numbers-first test of whether a claimed advantage survives scrutiny, cross-linked from M4.05. [Paid] [Advanced]

Market sizing (mechanics live in CN2.02, revisit its resources, Cosentino's Case in Point and Victor Cheng's Case Interview Secrets/caseinterview.com, if the top-down/bottom-up arithmetic itself feels unfamiliar):

• Aswath Damodaran's data library (pages.stern.nyu.edu): free, regularly updated sector benchmarks for sanity-checking any market-model input against a live, dated source instead of a memorized figure. [Free] [Beginner–Advanced]

Business evaluation and the financial-quality lens:

• Tim Koller, Marc Goedhart, and David Wessels (McKinsey), Valuation, selected chapters on assessing the quality and sustainability of reported performance, the deeper toolkit behind Core section 7's commercial-quality checks. [Paid] [Advanced]

Tools of the trade (named as a category, not an endorsement of any specific vendor): two kinds of service are standard commercial-diligence infrastructure at most sponsors and advisory firms. Private-market data platforms are what analysts use to pull comparable transaction and company data. Expert-network services are paid platforms that arrange compliant calls with industry specialists and former employees, within the ethical bounds Core section 4 describes. Access and vendor choice vary by employer.

General: Harvard Business Review across every section above (mixed free/paid) remains the standing anchor for case-level treatments of due diligence, customer research, and competitive strategy.

Teach it back & journal

Feynman prompt. Explain to a colleague who has only ever seen a deal's financial model why a favorable market-sizing slide and a clean set of audited financials aren't, on their own, enough to say "go." Use the BrightPath thread. Show how a healthy blended 4% same-store growth rate hid a decelerating legacy base at 1%, and how a national market-size figure hid an aggressive regional-capture assumption. Then show how a sponsor-referred customer sample hid a real loyalty risk that only came out once the team built an independent sample. If your colleague can state the veto rule unprompted (a killed hypothesis overrides any favorable composite score), you've passed on the toolkit's central rule.

Journal prompt. Think of a recent purchase decision that a company's own marketing or sales data would have completely mischaracterized. It might be a subscription you kept for the wrong reason, a store you split loyalty with, a brand you almost dropped for a reason no dashboard would show. Write two paragraphs. In the first, describe what a company relying only on its own transaction data would have concluded about you. In the second, describe what a single well-designed, behavior-anchored interview would have surfaced instead, and why that gap is the one Voice-of-Customer research exists to close. Then add a sentence naming the habit you're most likely to skip under real time pressure: the kill criterion, the independent sample, the same-store bridge, or the veto rule. Say what you'll do differently the next time a deadline tempts you to skip it.

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