The weaker first-meeting posture is still the default:

Tell me your biggest problems and I’ll suggest some AI ideas.

It sounds humble. It is also a request that the organisation perform the one job it is systematically bad at: naming its own structural friction. The stronger posture is not louder confidence. It is a different contract.

I have formed an evidence-backed theory about where your business model burns customer attention, senior labour and margin. Here is why I believe it, what would disprove it, and the product I think becomes possible if it is true.

That second sentence can sound like arrogance until two constraints sit beside it. Outside evidence may only nominate the problem. Only internal evidence may earn the word “biggest.” Between those constraints is the machinery this book owns: a Friction Thesis Compiler — an extension of the Proposal Compiler’s Marketplace-of-One motion — that turns industry structure and public company expression into a falsifiable artefact, not a claimed diagnosis.

Why the organisation is a poor first witness

Structural friction rarely announces itself as “the system.” It becomes how the work normally feels. It is distributed across sales, architecture, delivery and finance. It is visible only in fragments to each participant. It is protected by past decisions and incentives. And it is described in the vocabulary of the current process — proposals, gates, day rates, SOW revisions — which makes the process sound like the natural order of things rather than a design choice.

Ask a salesperson and you hear about slow approvals or weak relationships. Ask an architect and you hear about poor requirements. Ask delivery and you hear about sales overselling. Ask finance and you hear about margin variance.

Each account can be correct while none names the system manufacturing all four outcomes.

That is not a claim that people inside are stupid or dishonest. It is a claim about vantage point. People who live inside a commercial model experience its frictions as local problems with local owners. The system that joins those locals into one causal chain is easier to see from a few steps back — from industry economics, from how buyers are invited to enter, from the artefacts the firm publishes as “how we work.”

There is a usability analogue that makes the architecture clearer. Classic interface heuristics insist that systems minimise memory load — people should not have to remember information from one part of an interaction to another.¹ A commercial engagement that forces the customer to re-explain intent at every handoff, and forces the provider’s seniors to re-join fragments for every pursuit, is the same failure class at organisational scale. It is not “how professional services works by nature.” It is a design that uses human memory as the integration layer.

The outside observer’s question

The Friction Attack Surface already gave outside observers a sharper diagnostic question than “what keeps you awake at night?”: where are customers or staff repeatedly translating stable intent into the interactions demanded by the provider’s systems and business model?

That friction may be accidental architecture. It may also be wrong friction — friction that exists because of the commercial model itself, not because the customer’s outcome requires it. This book will use that classification; it will not re-teach the full taxonomy. The taxonomy is inherited. The new object is the compiler that turns industry priors and public evidence into a killable thesis about which friction is load-bearing for a specific firm.

What this book will make you able to do

After this book you should be able to:

• build an industry friction seed for a service category you sell into;
• compile a target company’s public evidence against that seed;
• generate and attack competing structural diagnoses, keeping the losers visible;
• produce a 15-field Friction Thesis with assumptions, known absences, first-signal tests and a kill condition;
• run a first meeting as falsification, not discovery theatre.

The scalable asset that falls out of that motion is not a longer list of targets. It is a library of industry seeds and observed company deviations — each engagement improving the prior for the next firm in that industry.

Where this book stops

What the diagnosed opportunity becomes as a commercial unit — the AI-native successor offer — is owned elsewhere. One product family worked in depth — preparedness as the product — is owned elsewhere. The generic Proposal Compiler loop (kernel → research → synthesis) is inherited, not re-taught. Paid evidence engagements after entry are out of scope. This book stops at the falsification conversation: the first meeting that tests the thesis.

The reader question

How do you walk into a first meeting with an evidence-backed theory of a prospect’s biggest structural friction — without interviewing anyone and without being arrogantly wrong?

The answer is not a better questionnaire. It is a different epistemic object: a Friction Thesis that states the equation from outside and leaves the coefficients to internal evidence. Chapter 2 draws that boundary without hedging it into uselessness.

The useful distinction is not “outside knowledge versus inside knowledge” as a vague hierarchy of access. It is a hard boundary on what each kind of evidence is allowed to do.

That is the title of this book for a reason. The equation is the structural relationship — which commercial design creates which translation burdens, which handoffs, which pricing-before-evidence moment. The coefficients are magnitude, incidence, ranking against other pains, and local exceptions. Confusing the two is how competent research becomes arrogant diagnosis.

What the equation looks like from outside

Take the shape of a mid-sized data consultancy that sells bespoke analytics and transformation work. From outside you can often see enough of the commercial structure to form a structural model:

• broad, bespoke consulting work rather than a narrow catalogue of buyable products;
• clients entering with imperfectly framed data or reporting needs;
• several disciplines contributing to scope;
• uncertain delivery effort before the estate is known;
• substantial proposals and statements of work as the first major transaction;
• high-value downstream projects as the economic prize;
• a market increasingly hostile to undifferentiated advisory.

That structure almost necessarily creates pressure around unpaid pre-sales cognition, pricing before sufficient shared understanding, incomparability between proposed scopes, contingency padding or underquoting, and rediscovery during delivery. You do not need a confidential margin pack to see the shape of that equation. You need industry literacy and public expression of how the firm invites buyers in.

What the coefficients require

What you cannot yet know externally — and must not pretend to know — includes:

• how many unpaid hours, by role, actually enter an average proposal;
• whether one practice has already solved the problem locally;
• how much gross margin the issue destroys;
• which client segments experience it most;
• how often proposal quality is truly decisive in wins and losses;
• whether another friction currently has greater economic consequence.

Those are coefficients. They decide whether the nominated problem is “a real pain,” “a material commercial leak,” or “the biggest structural friction this firm should reconstitute first.” Outside evidence is not allowed to skip that step by rhetorical force.

When the outside reading is wrong

A serious method must survive its own failure. Suppose you arrive with a sharp thesis about pricing-before-evidence, and internal observation shows that proposals are short, win rates are healthy, estimate-to-delivery variance is tight, and the firm already sells a bounded readiness product that most opportunities enter through. Then the thesis weakens or dies.

That is not embarrassment. That is the system working. You killed an attractive but incorrect proposal before you built a product around it. The kill condition is not a footnote for legal safety; it is part of the diagnostic product. Without it, you have only performed discovery theatre in reverse — still certain, only earlier.

What this boundary buys you commercially

The boundary is not only epistemic hygiene. It redesigns the first meeting. The client is no longer hired as unpaid chief diagnostician of their own business model. They are hired as the only party who can supply residual evidence: confirm, modify, exception, or kill. That is more respectful of their time and more persuasive about your capability — provided you actually document alternatives and tests rather than performing humility as a verbal garnish.

It also tells you what the first paid access is for, when that day comes: coefficient-solving against a named equation, not blank-page interviewing. Designing that paid engagement is out of scope here. Knowing what it must measure is not.

A practical test for your own draft thesis

Before you walk into a room, run this self-check on the document you plan to share:

1. Can you point to public artefacts for every load-bearing structural claim?
2. Have you written at least two alternative explanations you genuinely considered?
3. Have you listed what you cannot know from outside — in the document, not in your head?
4. Is there an observable kill condition that would force you to abandon the intervention you prefer?
5. Does any sentence award “biggest,” “always,” or a hard percentage without an internal measure behind it?

If item 5 fails, you have crossed from nomination into coefficient fiction. Delete the superlative. Keep the equation.

Language that stays honest

Prefer: “exposed structural friction,” “nominated primary diagnosis,” “public shape suggests,” “coefficients required.” Avoid: “your biggest problem is,” “obviously,” “always,” “industry average of X% for firms like yours” when you have no measure. The vocabulary of nomination is not weaker than the vocabulary of crowning. It is more precise — and precision is the credibility move when everyone else is performing certainty.

Company-specific workflows, politics, systems, exceptions and approval chains ultimately decide AI fit. That Vertical-of-One instinct remains correct. It does not require you to begin with an empty page.

The better hierarchy is three layers, in order:

INDUSTRY STRUCTURE
What friction normally follows from this market and business model?
        ↓
COMPANY EXPRESSION
How does this firm’s website, catalogue, organisation and public activity
appear to instantiate or resist that structure?
        ↓
INTERNAL RESIDUAL
What is genuinely different here — scale, politics, exceptions,
unit economics, hidden systems and strategic priorities?

Ship the skeleton; ingest the difference

This is the Seed Wiki move applied to commercial diagnosis. In knowledge systems, the point of shipping a skeleton is not to pretend the map is complete. It is to make the difference legible. When the skeleton is explicit, every contradiction between prior and local reality becomes information rather than noise.

Apply that to a prospect. Instead of asking them to explain their business from a blank page — which invites process vocabulary and departmental fragments — arrive with an explicit structural model of how firms like theirs usually burn attention and margin. Then hunt deviations.

What counts as a deviation

Deviations are not “interesting facts.” They are differences that change the coefficients or the equation itself. Examples in data consulting:

• The firm already sells a fixed-price readiness or diagnostic product that most opportunities enter through — the first commercial unit may already have changed.
• One practice has productised scoping and the others have not — the firm is not one system; the residual is internal variance.
• Public language is bespoke, but procurement materials show packaged SKUs — expression and transaction grammar disagree; investigate which is real.
• Hiring is thick with delivery engineers and thin with pre-sales architects — the unpaid cognition may sit somewhere unexpected.
• Case studies emphasise multi-year retainers rather than large SOW transformations — different wrong-friction map.

Each deviation either weakens a candidate thesis, strengthens a rival, or rewrites a seed stub for the next firm in the industry. That is how the library compounds — developed fully later, introduced here as the reason the hierarchy is worth the discipline.

Company expression is evidence, not proof of operations

A website is not a process audit. Job ads are not an org chart with truth stamps. Case studies are marketing. Treat them as what they are: compressed statements of what the firm believes it sells, whom it hires to sell and deliver it, and how it expects a buyer to enter. That is enough to nominate. It is not enough to award “biggest.”

The common failure is either to dismiss public material as “just marketing” — and thus start empty — or to treat it as operational truth — and thus arrive arrogantly certain. The hierarchy forbids both. Public expression fills the middle layer. Internal residual finishes the job.

Discovery redesigns without disappearing

This is not “never interview.” It is “do not use the first conversation to invent the structural model.” Interviews become instruments for residual hunting: which coefficient is wrong, which exception is load-bearing, which alternative explanation wins. The client still speaks. They no longer have to perform the entire theory for you.

Worked micro-example of residual hunting

Suppose the industry seed says data consultancies typically lack buyable entry products. Your target’s website shows a “Data Readiness Assessment” with a fixed price and a three-week clock. That is not a minor marketing detail. It is a deviation that may kill the lead thesis for that firm — or reframe it as “the product exists but is not the default first unit.” Your next questions change: what share of pursuits enter through it? Is it sales theatre or a real gate? Those are coefficient questions. The skeleton made them visible in one glance.

Without the skeleton, the same page might have been filed under “they have an assessment offer — nice” and forgotten. With the skeleton, it is either a kill signal or a modification signal. That is deviation hunting earning its keep.

How the three layers fail when inverted

Start with residual and you get anecdotes: “our partner said margin is fine.” Start with company expression alone and you get homepage psychology: “they use the word transformation a lot.” Start with industry alone and never descend, and you ship generic industry advice with a company name pasted on. The hierarchy is a sequence, not a menu. Each layer constrains the next.

Industry without expression produces a true prior that may not fit this firm’s commercial architecture. Expression without residual produces a nominated thesis that awards coefficients it has not earned. Residual without industry produces client-led fragments with no system model — the discovery theatre of Chapter 1 wearing better clothes.

What you write down at each layer

Notice that residual is not “more discovery interviews until something interesting appears.” Residual is structured measurement against a named equation. The interview may still happen; its job is now coefficient-solving and exception-finding, not theory invention.

When company expression resists the industry seed

Resistance is not a bug. A firm that productised entry units, or that sells only retained capacity, or that is really a software product company with a services veneer, should break the default data-consultancy seed. Your job is to record the break, not to force the seed. The library gets richer when resistance is typed: “resists because readiness is default first unit,” “resists because multi-practice variance,” “resists because buyer is procurement-led panel, not SOW dance.”

Those typed resistances become the next seed’s conditioning rules. That is how industry structure stays a prior rather than a prejudice.

An industry friction seed is not a market report. It is a structured prior for the transaction grammar of a category: who acts, what artefacts move, where handoffs fail, how price is formed, and which wrong frictions the commercial model tends to manufacture.

Without a seed, every target company forces you to rediscover the industry. With a seed, each target is a compile pass — and each compile pass can improve the seed.

What the seed must contain

At minimum, a usable seed holds the following stubs. Chapter 7 fills every one for data consultancies. Here is the schema you will reuse across industries.

Transaction grammar as the spine

The seed’s spine is a transaction grammar: the sequence of states through which a buyer and provider usually move. For many professional services categories it rhymes even when the nouns change — outcome problem, translation into provider language, reconstruction of context, pricing under uncertainty, freezing assumptions, rediscovery during delivery. The data-consultancy grammar is spelled out fully in Chapter 7. The point here is architectural: if you cannot write the grammar, you do not yet have a seed. You have vibes.

How public evidence fills the stubs

Marketplace of One already treats public research as a first-class surface: websites, job postings, organisation signals, recent changes, case language. The Friction Thesis Compiler uses the same surface for a sharper purpose — not only “understand the company,” but “instantiate or challenge the industry seed.”

• Website and service catalogue — entry offers, language of bespoke vs productised, workshop-led vs product-led paths.
• Case studies — what “success” looks like commercially (transformation programmes vs bounded outcomes).
• Job advertisements and role mix — where labour is thick; whether product ownership of offers exists.
• Partnerships and technology badges — platform-led sales motion vs pure advisory.
• Executive posts and talks — what leaders frame as the value unit.
• Procurement / tender material when public — how buyers are forced to request work.
• Reviews and community chatter when available — client-side translation burden.

A company that describes everything as “tailored transformation,” advertises a large multidisciplinary bench, has no narrow entry offers and leads with workshops or advisory has supplied evidence about its commercial architecture — even without disclosing its internal proposal process.

Seed quality tests

A seed is good enough to compile against when:

1. a practitioner in the industry recognises the transaction grammar without needing your target’s name;
2. wrong friction is named as business-model distortion, not as “people should try harder”;
3. at least two alternative structural diagnoses could be generated from the same seed (so you are not monomaniacal);
4. each stub can be challenged by a public signal (otherwise it is unfalsifiable dogma).

Versioning the seed

Treat the seed as a living prior, not a manifesto. When three company compiles show the same deviation — for example, readiness products that exist on the website but never appear in case studies — update the seed with a conditioning note: “public readiness SKU is weak evidence of default first unit unless case mix confirms it.” That is library hygiene. Seeds that never change are either perfect (unlikely) or frozen dogma (common).

How to draft a seed without inventing a novel

Practitioners freeze when “write an industry seed” sounds like a research project. Use this drafting order instead:

1. Write the transaction grammar as a chain of verbs — no adjectives, no solutions. If you cannot finish the chain, you do not yet understand the category’s commercial path.
2. Name the join — who currently reconciles fragments for the provider and for the customer. If you cannot name a role, the seed is still a market overview.
3. Classify one wrong friction — business-model distortion forced onto attention, not “people should communicate better.”
4. List three structural rivals that could also be primary — this seeds Chapter 5’s protocol.
5. List the public signals that would support or contradict the wrong-friction claim — this seeds the compile surface.
6. Name one replacement direction at unit altitude — not a feature list. “Evidence-first commercial unit” is altitude; “use GPT to draft SOWs” is not.

That draft is enough to compile the first three firms. After three, version the seed with what broke. Do not wait for academic completeness. A thin explicit seed beats a thick implicit intuition you cannot share with a colleague or a compiler.

Seed versus persona versus ICP

Ideal customer profiles and personas describe who buys. A friction seed describes how the commercial system burns attention and margin when those buyers engage. You can have a perfect ICP and still have no seed. You can also have a sharp seed that applies across several ICPs in the same industry. Do not confuse marketing segmentation with structural diagnosis. Segmentation answers “who.” The seed answers “what structural friction does this market’s transaction grammar manufacture.”

When not to write a new seed

If you are entering a category for one opportunistic meeting and will never return, a full seed may be overkill — write a one-page grammar and three candidates, then archive or discard. The seed becomes load-bearing when you intend to run multiple compiles in the same industry. That is when the library economics justify the prior. Marketplace of One already invested in frameworks once and recompiled per company; the industry seed is the diagnostic half of that same amortisation.

Chapter 7 is the complete worked seed for one service industry. Treat this chapter as the mould; that chapter as the casting.

The compiler should not search broadly for “problems at Company X” and summarise whatever the web returns. That is search-guided reasoning: the accident of available pages becomes the structure of the diagnosis. The correct order is the reverse.

The protocol

1. Load the industry seed — transaction grammar, typical wrong friction, known failure shapes.
2. Generate three to six structural candidates before opening a browser tab about the target. Candidates must be structural (“the first commercial unit prices unobserved transformation”), not symptomatic (“proposals take too long”).
3. Write the attack questions for each candidate — what public signal would support, contradict, or condition it?
4. Research per candidate — separate queries, separate notes. Do not run one mash search and allocate leftovers.
5. Score survivors against the win criteria below.
6. Keep losers in the document with the specific reason each lost.

Win criteria for the surviving diagnosis

A candidate survives as the lead thesis when it has:

• Structural inevitability — given the commercial model, this friction is hard to avoid without changing the model.
• Multiple independent public signals — not one blog post, but catalogue language + hiring + case shape + entry path, for example.
• Substantial plausible economic consequence — stated as shape until measured (pre-sales cost, cycle time, margin variance, win friction, churn of trust).
• A named buyer — someone who can own the pain and the first test (managing partner, head of delivery, sales lead — not “the organisation”).
• An accessible first test — internal measures that could confirm or kill within a short horizon without boiling the ocean.
• A credible AI-constituted replacement hypothesis — a different first commercial unit or transaction, not “use AI to write proposals faster” alone.

Why losers stay

The Proposal Compiler already treats visible rejection as proof of judgment: the fish you throw back is part of why the fish you keep is trusted. In a Friction Thesis, deleted alternatives look like tunnel vision. Retained alternatives look like work. They also protect you in the room: when the client names a rival diagnosis you already considered, you are not improvising — you are updating coefficients.

Symptom candidates versus structural candidates

Train yourself to rewrite symptom candidates upward. “Unpaid proposal production” may be true and still be a symptom of “pricing an unobserved transformation as the first major transaction.” “Poor knowledge reuse” may be true and still be a join-algorithm failure secondary to the commercial unit. Chapter 9 walks five candidates for the data-consultancy case and shows exactly how a true symptom loses as the primary structural diagnosis.

If you cannot generate genuine rivals, your seed is too thin or your attachment to a favourite story is too thick. Either way, stop and fix that before you research.

What “attack research” looks like in practice

For each candidate, write three columns before you search: support signals, contradict signals, condition signals. Example for “unpaid proposal production is primary”:

• Support: job ads that emphasise bid support; leadership posts about proposal burden; case studies that brag about custom responses without naming a product entry.
• Contradict: public fixed-price diagnostic SKUs; partner interviews that already reframe the first unit; reviews that praise speed-to-value products.
• Condition: unpaid labour is real but only on enterprise pursuits; mid-market already productised — thesis becomes segment-conditional.

Then search those columns separately. When support and contradict both return strong hits, do not average them into mush. Write the condition. Conditional theses are often more honest — and more useful in the room — than a forced single story.

A note on AI assistance

AI can help generate candidate lists and draft attack questions. It must not be allowed to collapse rivals into a single “most likely” paragraph before you have attacked them. The John West step is a human integrity gate even when the drafting is machine-assisted. If your toolchain hides losers, your thesis will look like every other confident vendor brief.

Common generation failures

• Symptom soup — five candidates that are all rewordings of “process is slow.” Force structural altitude: commercial unit, authority design, evidence timing, join placement.
• One true way — generating four straw men so the favourite wins. Ask a colleague to add a candidate you dislike.
• Search leakage — peeking at the target’s site before candidates exist, then “generating” what the homepage already suggested. That is reverse rationalisation.
• Tool fetish — candidates about models and platforms rather than transaction grammar. Tools can appear in replacement hypotheses; they rarely are the structural diagnosis of a services firm’s first unit.

If your candidate list fails these checks, delete it and regenerate from the seed alone. Speed is not a virtue when it reintroduces blank-page discovery under a different name.

From candidates to the fifteen fields

The surviving lead thesis does not jump straight into a meeting script. It feeds field 7 (structural cause), field 11 (alternatives = losers), field 8 (receipts gathered during attack research), and fields 14–15 (tests and kill derived from what would have to be true). The protocol in this chapter is therefore not a brainstorming icebreaker. It is the manufacturing step that makes the Friction Thesis schema in Chapter 6 fillable without invention.

When attack research returns only weak public signals for every candidate, you have two honest options: widen the seed (maybe you are in the wrong industry model), or admit that public expression is too thin to nominate and that the first conversation must be framed as seed-building with the client rather than falsification of a sharp thesis. Thin public signal is not a licence to bluff. It is a different epistemic stage — still better than empty discovery if you say so explicitly.

The Proposal Compiler broadly did: your kernel plus company research yields a company-specific proposal. The Friction Thesis Compiler does a sharper compile:

industry friction seed
+ company public evidence
+ your capability kernel
+ comparable failure shapes
→ falsifiable friction thesis
→ AI-constituted product hypothesis
→ fixed-price verification path (pointed at, not designed here)

That is an explicit extension of Marketplace of One, not a replacement of it. The proposal may still be the leave-behind. The thesis is the diagnostic spine that keeps the proposal from performing generic understanding theatre.

The fifteen fields

These fields are the definitive schema for this book. Chapter 10 fills every row for the worked mid-sized data consultancy. Here each field’s job is stated so the later specimen is assembly, not invention.

Why the uncomfortable fields are non-negotiable

Fields 9–11 and 14–15 are where arrogance dies. A thesis that only states the clever structural claim (fields 1–8 and 12–13) is a pitch deck. A thesis that also states assumptions, absences, alternatives, tests and a kill condition is a diagnostic instrument. If those fields are empty, you are not ready for the meeting — no matter how sharp the lead sentence is.

This is the book refusing to do to the reader what it warns against doing to a prospect: perform certainty without a test.

What “good” looks like before the specimen

• Transaction grammar is specific enough that a stranger in the industry can walk it.
• Join algorithm names a role, not “the team.”
• Classification says wrong / accidental / necessary with a one-line defence.
• Public receipts point at observable artefacts.
• Economic consequence is directional without fabricated precision.
• Replacement is a commercial unit hypothesis, not a tool list.
• Kill condition is observable, not “if they don’t like us.”

How the fields prevent the two failure postures

Discovery theatre skips fields 1–8 and jumps to asking the client to invent the system. Website arrogance fills fields 1–8 with swagger and leaves 9–11 and 14–15 blank. The schema makes both failures visible as missing rows. If you cannot show the table, you are not ready to claim a diagnosis — you are ready to research further or to ask better residual questions, which is a different honesty.

Field pairs that must stay coupled

Some fields only make sense together. Public receipts (8) without assumptions (9) pretends observation was pure. Assumptions without known absences (10) pretends you know the edge of your knowledge. Economic consequence (12) without first-signal tests (14) is a scary story. Replacement (13) without kill condition (15) is a product pitch. When you draft, write the pairs in the same sitting. Split drafting is how the uncomfortable half goes missing.

Also couple alternative explanations (11) to the competing-diagnosis protocol in Chapter 5. Field 11 is not a creative-writing box for “other ideas.” It is the home of attacked rivals with reasons. If field 11 is empty or vague, you skipped John West.

Drafting order that prevents clever emptiness

Do not write the lead sentence first and reverse-fill the table. Use this order:

1. Transaction grammar (2) from the industry seed, adjusted for public expression.
2. Stable intent (3) and translation burden (4) — what is constant, what unpaid work the grammar demands.
3. Join algorithm (5) and classification (6) — who joins, whether the friction is wrong/accidental/necessary.
4. Structural cause (7) in one hard sentence.
5. Public receipts (8) listed as artefacts, not adjectives.
6. Assumptions (9) and known absences (10) written before you allow yourself a lead claim.
7. Alternative explanations (11) from attacked rivals.
8. Only then crystallise desired outcome (1) and the economic consequence hypothesis (12) as shape.
9. Replacement (13) at unit altitude.
10. First-signal tests (14) and kill condition (15) last — and refuse to ship without them.

That order feels slower than drafting a brilliant paragraph. It is faster than recovering from a meeting where the brilliant paragraph had no kill condition and no rivals when the client pushed back.

What the thesis is not

• Not a SWOT.
• Not a persona journey map (though journey steps may appear in the grammar).
• Not a full proposal (the proposal may contain it).
• Not a product requirements document (replacement is a unit hypothesis).
• Not a claim of “biggest” until residual measures earn the word.

If a stakeholder asks you to “just put the insight in a slide,” you can still lead with the structural sentence — but the leave-behind must carry the full schema or you have performed insight theatre. The fifteen fields are the product of this chapter’s doctrine. Chapter 10 is the full resolution specimen. Chapters 7–9 build the seed, public compile and competing diagnoses that feed it.

This chapter is not illustration for colour. It is the first half of the book’s proof burden: a seed complete enough that a reader could compile a real firm against it next week. The category is mid-market data, analytics and digital-transformation consulting — the industry that sells meaning from data under bespoke commercial models.

Transaction grammar

Client experiences an outcome problem
  (reporting lag, trust gap, migration fear, board metric, ops blindness)
        ↓
Client translates it into a consulting request
  (“we need a dashboard / data platform / AI strategy / modernisation”)
        ↓
Salesperson qualifies budget, timing, stakeholders
        ↓
Consultants reconstruct context from workshops, samples, tribal knowledge
        ↓
Architects imagine a target state under incomplete observation
        ↓
Delivery estimates effort against a partially known estate
        ↓
Price is negotiated (often before shared evidence of the estate)
        ↓
SOW freezes a provisional model of work, risk and outcomes
        ↓
Delivery discovers the actual systems, data quality, politics and scope
        ↓
Changes, contingency burn, dispute or overbuild follow
        ↓
Margin and trust consequences land on both sides

Walk that chain slowly and you can already feel where unpaid cognition sits: before the SOW freezes, and again when delivery reopens what the SOW pretended was settled.

Actors

• Client economic buyer — often a CIO, CDO, Head of Analytics, or business unit lead who owns the outcome but not the full estate map.
• Client technical influencers — data engineers, architects, report owners who know fragments.
• Client procurement / finance — force the relationship into a priced SOW early.
• Provider salesperson / account lead — qualifies and shepherds price.
• Provider consultants — reconstruct problem language into analysis language.
• Provider architects / principals — invent target states and write or bless the SOW.
• Provider delivery leads — inherit the freeze and meet reality.
• Provider finance / partner — owns margin variance after the fact.

The seed should also note missing actors that signal productisation: offer owners with P&L on a bounded product, product marketers of entry units, success managers who sell the next unit from evidence rather than from a new pursuit workshop. Their absence is a public signal about commercial architecture when you compile a target firm.

Touchpoints

Customer-side: website enquiry or referral; discovery calls; workshop days; security questionnaires; proposal review meetings; SOW redlines; kickoff; weekly steers; change-request debates; QBR if retained.

Staff-side: pursuit meetings; solution reviews; peer estimation; partner price coaching; late-night SOW drafting; delivery recovery meetings; write-downs.

Notice the asymmetry. The customer experiences a few high-stakes meetings. The provider’s senior labour may re-enter the same reconstruction work across many pursuits that never become revenue.

Evidence artefacts

• Initial email brief or RFP (often outcome-shaped, estate-light).
• Discovery notes and workshop whiteboards (fragmented, rarely durable).
• Proposal deck and narrative (promise-heavy).
• Statement of work (assumption-heavy freeze).
• RAID / risk registers (often opened properly only after signature).
• Change requests and variation logs.
• Timesheets and write-off reports (the economic truth, late).

Handoffs and decision rights

Critical handoffs: sales → solution; solution → estimation; estimation → partner pricing; signed SOW → delivery; delivery → finance when variance appears. Decision rights are often misaligned: sales can create expectation; architects can expand scope language; partners can cut price; delivery owns the pain; finance sees the wreckage. No single role is villain. The join is the problem.

Pricing mechanisms

Typical mechanisms in this industry:

• time-and-materials day rates with a “not to exceed” that everyone knows is soft;
• fixed-price projects with contingency buried in days or risk line items;
• phased SOWs that still front-load a large design/build commitment;
• retainers that begin only after a bespoke transformation sale.

What is rare as a default first unit: a modest fixed-price purchase of shared evidence about the estate before either side prices the transformation. That rarity is itself a seed signal about wrong friction.

Sources of uncertainty

• unknown data quality and lineage;
• unknown integration surface and shadow systems;
• multi-stakeholder intent that was never joined;
• regulatory and security constraints discovered mid-flight;
• political ownership of metrics and reports;
• unclear definition of “done” for analytic outcomes.

Metrics the industry tracks vs metrics that would reveal the friction

Usually tracked: utilisation, billable mix, pipeline, win rate, average deal size, NPS after delivery.

Would reveal structural friction if tracked honestly: unpaid pre-sales hours by role; estimate-to-actual variance by engagement type; discount frequency after first price; time from serious discussion to paid start; change-request recurrence themes; margin variance by “large SOW” vs “bounded diagnostic” motions; percentage of pursuits that could have entered through a readiness product.

Firms often optimise the first list while the second list contains the structural story.

Typical wrong friction

Using Friction Attack Surface classes:

• Necessary: informed consent on price bands; security assurance; real capacity limits; a human authority to accept residual risk.
• Accidental: rediscovering the same client history because tools do not join; reformatting the same content across proposal templates; calendar archaeology for stakeholders.
• Wrong: forcing the first major commercial unit to be a priced bespoke transformation before either side holds a shared, evidence-backed model of what must be transformed. That is not physics. It is commercial model design. It manufactures translation burden for clients (“make my vague reporting pain look like a project”) and unpaid join labour for seniors (“turn fragments into an SOW that can be sold”).

Candidate AI-constituted replacements (hypotheses only)

The seed should name replacement directions without designing the full offer (that is sibling territory):

• a fixed-price evidence / readiness / preparedness first unit that produces line-item findings and dispositions before transformation pricing;
• a successor commercial unit that replaces day-rate transformation as the default entry — the bounded promise object owned by the successor-offer doctrine;
• compiled firm memory that reduces re-reconstruction of history (join support, not a chatbot trinket).

The seed’s job is to make those candidates available as structural responses to wrong friction — not to qualify them through product gates. Qualification is another book.

How the seed is used in a compile pass

When you open a target company’s public materials, you do not “research the company” in the vague sense. You walk the seed stubs:

1. Does their entry language match “consulting request after outcome problem,” or do they already sell a bounded first unit?
2. Which actors appear in case studies and job ads — and which are missing?
3. What artefacts do they show (playbooks, assessments, SOW templates), and what do they hide?
4. Where would handoffs break if the grammar is standard?
5. Is wrong friction still the pricing-before-evidence unit, or has something else taken its place?

Each answer either fills a stub, marks a deviation, or weakens a candidate diagnosis before you invent a favourite story. That discipline is what separates a Friction Thesis compile from a marketing brief with footnotes.

Seed summary card

Chapter 8 takes this seed and compiles public expression for a generalised mid-sized data consultancy — still without internal coefficients.

Public signals that fill the seed stubs

Service catalogue and homepage language. Offerings framed as tailored transformation, end-to-end data capability, strategy-through-delivery. Few or no SKU-like entry products with fixed scope and price on the public surface. Buyers are invited into conversation, workshops or “let’s discuss your challenges.”

Case studies. Stories of platforms stood up, estates modernised, reporting estates rebuilt, multi-month programmes. Success is narrated as large implementation outcomes, not as repeated sale of a bounded evidence product.

Hiring and roles. Vacancies and team pages thick with consultants, engineers, architects and client principals. Thin or absent: product managers of packaged offers, offer owners measured on product P&L, readiness specialists as a named product line.

Entry path. Lead with advisory workshops, assessments described as sales motions rather than buyable products, or “free discovery” that still terminates in a large SOW. The public path does not show a default first commercial unit of purchased shared evidence.

Partnerships and tech badges. Cloud and platform partnerships present — useful delivery capability, still compatible with bespoke transformation as the economic unit.

None of these signals is a process audit. Together they instantiate the Chapter 7 seed: the firm appears to sell multidisciplinary bespoke work entered through high-touch advisory, with the first major priced artefact still a transformation-shaped agreement.

The lead structural thesis (nomination only)

The firm’s exposed commercial friction is that the first major transaction asks the client and consultancy to agree on the price of a bespoke transformation before either possesses a shared, evidence-backed model of what must be transformed.

That is not “they spend too long writing proposals.” Proposal labour is a symptom surface. The structural claim is about the commercial unit: pricing the unobserved.

Causal chain

Client has a partly understood reporting / data / AI need
        ↓
Client translates it into consulting language
        ↓
Sales and consultants reconstruct different parts of the world
        ↓
Budget and price are negotiated before scope is well grounded
        ↓
A large SOW freezes provisional assumptions
        ↓
Neither side can easily measure the quality of the proposed answer
        ↓
Reality is rediscovered during delivery
        ↓
Margin variance, changes, dispute or unnecessary overbuild

The join algorithm

In this grammar the senior SOW author — principal, solution architect, senior consultant — becomes the manual join across client request, account history, prior delivery, architecture, capability, staffing and commercial risk. AI-Constituted Services already named that suppression case: the senior SOW author is the join algorithm. The point for this book is diagnostic: if public expression shows heavy principal-led bespoke pursuits and no evidence-first product, you should expect join labour to concentrate in SOW production. Whether that concentration is the firm’s largest economic leak remains a coefficient question.

What the replacement direction is (route, do not design)

If the thesis holds, the first commercial unit should change:

Uncertain transformation
→ fixed-price purchase of shared evidence
→ line-item findings and dispositions
→ evidence-backed implementation options
→ informed budget decision

What that object becomes as a successor offer, and how a preparedness product family is built, are owned by sibling work. This book only needs the directional claim: the thesis implies a different first transaction, not merely faster drafting of the old one.

How strong is “public only” as evidence?

Strong enough to nominate. Weak enough to forbid crowning. Public materials are performative — firms show what they want buyers to believe. That does not make them useless. Performance still reveals commercial architecture: what unit they invite the buyer into, what labour they hire, what success looks like in their storytelling. A firm that never shows a productised entry unit on any surface is making a statement even when the statement is incomplete.

The ethical use of that statement is to form a hypothesis with teeth — alternatives, absences, kill condition — not to bully a prospect with their own homepage. Chapter 10 packages the teeth. This chapter only establishes that the seed and the public shape can lock together without a single internal interview.

Compile notes a practitioner should keep

While reading public materials, keep a running note with four columns: seed stub, public signal, support/contradict/condition, confidence. “Confidence” here is not a fake percentage. It is a plain label — strong multi-surface agreement, single-surface hint, or marketing-only. That habit prevents a flashy homepage line from outweighing three quieter signals that point the other way.

Also timestamp sources. Public sites change. A compile without dates cannot be audited later when the library asks whether the seed still holds. Marketplace of One research was always supposed to be specific; the Friction Thesis Compiler makes specificity structural rather than stylistic.

Worked stub fill (illustrative, generalised)

That table is the public compile’s real work product — not a mood board of screenshots. Chapter 9 will attack rivals using these signals; Chapter 10 will freeze them into receipts, assumptions and absences.

What this chapter deliberately does not claim

• That this is the firm’s biggest problem (coefficients unknown).
• Any invented hours, win rates or margin percentages for the target.
• That every data consultancy matches this public shape (deviations are expected and valuable).
• That the replacement product is already qualified through product gates.

Chapter 9 attacks five rival structural diagnoses and keeps the losers. Chapter 10 freezes the full fifteen fields and the validation pack that would earn or deny “biggest.”

For the mid-sized data consultancy specimen, five structural candidates were generated before settling the lead thesis. Each can be partly true. Only one is nominated as the primary structural diagnosis of the first commercial unit. The others remain in the document with the specific reason they lost as primary.

Candidate A — Unpaid proposal and SOW production

Claim: The major friction is the unpaid labour that goes into proposals and statements of work.

Why it is plausible: Public shape shows bespoke multi-discipline pursuits; principals must join fragments; Boutique Consulting Club commentary treats low proposal-to-win rates (around 20–30% for many consultants) as a sign of broken pre-proposal economics, not merely weak prose.²

Attack: If the firm merely automated drafting while still pricing unobserved transformations as the first unit, would the structural burn continue? Yes — faster wrong freezes are still wrong freezes. Unpaid hours are a cost surface of a deeper unit-of-sale problem.

Reason it lost as primary: True as a cost centre and first-signal metric; insufficient as the structural root. It names the symptom labour, not the commercial equation that demands that labour.

Candidate B — Poor history reuse

Claim: The major friction is failure to reuse client and project history, forcing rediscovery every pursuit and engagement.

Why it is plausible: Knowledge-intensive firms routinely re-ask what they already knew; tools are document cemeteries; join labour restarts.

Attack: History reuse can improve without changing the first commercial unit. A firm with perfect retrieval can still ask clients to buy a large SOW before shared evidence of the estate. Reuse is often a subordinate join failure, not the generator of pricing-before-evidence.

Reason it lost as primary: Real and worth fixing; better framed as a contributor to translation burden and join cost than as the structural diagnosis of how the firm sells.

Candidate C — No buyable entry products

Claim: The major friction is the absence of narrow, buyable entry products.

Why it is plausible: Public catalogue shows tailored transformation and workshop entry; no SKU-like path. Productisation weakness is a strong public signal.

Attack: “No entry products” is close to the winner — sometimes the same photo from a different angle. It can lose as the name of the root if it describes the missing response rather than the structural friction. The structural problem is the demand that the first major transaction price an unobserved transformation. Entry products are a candidate reconstitution of that unit. Naming the absence of the cure is not the same as naming the disease — though both belong in the artefact.

Reason it lost as primary wording: Kept as tightly coupled sibling. Lead thesis states the friction in the current transaction; entry products appear under AI-constituted replacement and under alternative explanations as the commercial response path.

Candidate D — Margin variance from weak current-state evidence

Claim: The major friction is downstream margin variance caused by weak evidence at sale time.

Why it is plausible: Causal chain ends in variance, change, dispute, overbuild. Finance will recognise this description immediately.

Attack: Margin variance is typically a consequence node. If you start diagnosis at the consequence, you invent recovery programmes and estimation training while leaving the first commercial unit intact. The structural lead must sit upstream: what forced pricing before evidence?

Reason it lost as primary: High value as economic consequence hypothesis and as validation metrics (estimate-to-actual, GM by engagement type); not the first structural name.

Candidate E — Unmatched installed base

Claim: The major friction is an installed client base without systematic next-offer matching — expansion left to heroics.

Why it is plausible: Service firms leave money on the table post-delivery; no productised next unit; account growth is relationship weather.

Attack: This is often a real growth friction at a different moment in the customer lifecycle. Public signals for the specimen point hardest at acquisition and first SOW, not at installed-base matching. A firm can have severe unmatched-base problems and still have the first-unit problem — or not. For this compile, independent public signals for first-unit wrong friction are thicker.

Reason it lost as primary for this specimen: Different buyer moment; weaker public instantiation in the generalised mid-sized data consultancy shape used here; retained as a future seed-library branch for expansion motions.

Why the lead thesis won

Retention table (do not delete)

John West is not a flourish. If a client walks into the falsification meeting and says “our real problem is knowledge management,” you already have Candidate B written, attacked and demoted — ready to re-promote if their residual evidence supports it.

Scoring discipline

Do not rank candidates by how interesting they sound in a blog post. Rank them by the win criteria from Chapter 5, written as a short scorecard you can show a colleague. When two candidates tie, prefer the one higher in the causal chain — causes over consequences — unless public signals for the consequence are overwhelmingly stronger and the cause is invisible. For this specimen, Candidate D (margin variance) lost that tie-break to the lead thesis even though finance will feel D first. Feeling is not structure.

Also separate “true and important” from “primary structural diagnosis of the first commercial unit.” A firm can have a serious unmatched installed-base problem (E) and still have the lead thesis correct for acquisition. The library can hold both; the first meeting usually cannot lead with five primaries. Choose, retain, and date the choice.

The fifteen fields — filled

Validation pack — nine bounded internal questions

You do not need months of interviewing. You need a short set of observations a partner can authorise:

How to read the pack

• Confirm: material unpaid senior hours, soft estimate quality, margin variance on large SOWs, client hesitation, and a non-trivial pool of opportunities that could enter through evidence-first units.
• Modify: thesis holds for one segment or practice only; installed-base matching ranks higher for growth; history reuse is the dominant unpaid labour mode.
• Kill: see field 15. Celebrate the kill. That is the method earning its humility.

Worked “outside wrong” scenario

Imagine the nine questions return cleanly against the thesis: average unpaid proposal labour is modest and mostly junior; win rates on qualified pursuits are healthy; estimate-to-actual variance is tight; large SOWs are rare because most work is retained capacity; a readiness product already captures the majority of first meetings. The public website still says “tailored transformation” because marketing never updated. Your equation, built from industry prior plus homepage language, was a false positive on commercial architecture.

The correct response is not to argue with the coefficients. Write the kill receipt: “public transformation language over-predicted wrong friction when retained capacity + readiness entry already dominated.” Promote that into the industry seed as a conditioning rule. Then ask whether a retained alternative — unmatched installed base, for example — now deserves primary status given the residual. The method survives because it planned for this outcome in field 15 before pride got involved.

What “earning biggest” sounds like after validation

Only after coefficients move can language upgrade. Before validation: “We nominate this as the exposed structural friction of the first commercial unit.” After confirmation patterns: “On the measures we agreed, this friction is material — and among the issues we compared, it is the one with the largest addressable economic consequence for the pursuits we examined.” That second sentence is allowed. Inventing it in week zero is not.

Chapter 11 turns this artefact into a meeting — without collapsing into either discovery theatre or website arrogance.

The first meeting’s purpose changes when you hold a Friction Thesis. The client is no longer hired to invent the insight. They are hired to supply residual evidence: confirm, modify, exception, or kill. That is more respectful of their calendar and more revealing of your method — if you can hold conviction and falsifiability in the same breath.

Failure posture A — Discovery theatre

What keeps you awake at night?

What it costs:

• You outsource structural diagnosis to people who experience friction as local symptoms.
• You invite process vocabulary (“we need better requirements,” “sales and delivery are misaligned”) without a system model.
• You signal that your value starts when they educate you — the opposite of Marketplace-of-One meta-credibility.
• You may still hear useful residual data, but you have spent the highest-attention minutes of the relationship on blank-page interviewing.

High-effort service interactions train disloyalty in consumer contexts; the consulting analogue is forcing busy leaders to perform unpaid diagnostic labour for a vendor who arrived empty.⁴ Effort is not always the same as care.

Failure posture B — Website arrogance

I diagnosed your business from your website.

What it costs:

• You award coefficients you do not have.
• You delete alternatives in the room even if you wrote them at your desk.
• You make disagreement feel like insult rather than residual update.
• You teach the client that your method cannot survive contact with their internals.

Public expression nominates. It does not crown.

The falsification script

How the minutes actually run

1. State the equation in one minute — structural thesis, not a tour of your CV.
2. Show two or three public receipts — catalogue language, entry path, case shape — so the claim is visibly grounded.
3. Surface the top alternatives — “we also considered unpaid proposal labour as primary, history reuse, missing entry products, margin variance as root, installed-base matching — here is why they lost as primary, and what would re-promote them.”
4. Name the known absences — explicitly what you could not know from outside. This is the humility that is operational, not decorative.
5. Walk the validation pack — the nine questions. Ask which they can answer now, which need a week, which are sensitive.
6. State the kill condition aloud — if these observations fail, we stop treating this as the primary diagnosis.
7. Agree the residual work — who pulls which measure; what “confirm / modify / kill” would look like; whether a follow-on conversation is warranted.

Claims about them carry a test

Strategy Engine doctrine is exact on the commercial ethics of pre-meeting diagnosis: claims about you carry receipts; claims about them carry a test. Convert “your problem is X” into “we believe your constraint is X — and here is how we would verify it.” The falsification meeting is that rule embodied as a conversation design.

What the leave-behind is

Do not leave a vague “great to meet you” email. Leave the Friction Thesis spine:

• lead structural claim;
• causal chain;
• public receipts;
• alternatives with rejection reasons;
• assumptions and known absences;
• nine questions;
• kill condition;
• replacement direction at one paragraph of altitude (with sibling destinations if useful).

If they forward only one artefact internally, that pack should be able to travel without you in the room — the same board-presentation test the Proposal Compiler already demanded of proposals.

When they push back

“That’s not our biggest problem.” Perfect. That is a coefficient claim. Ask which measures would show what ranks higher. Offer to re-score alternatives.

“You couldn’t know that without talking to us.” Agree on coefficients. Hold the line on equation: industry structure and public expression are allowed to nominate.

“We already know our problems.” Ask them to state the system, not the fragments. If they can name the structural unit cleanly, your residual work shortens. If they list four departmental pains, you have just demonstrated Chapter 1 live.

Facilitation details that keep the posture

Print or screen-share the alternatives table. People argue less with a ghost they cannot see. When a partner attacks Candidate C, point to the written demotion reason and ask what residual would re-promote it. That converts status contests into evidence contests.

Time-box the equation to five minutes. The meeting’s scarce resource is residual access, not your monologue. If you spend thirty minutes proving how clever the seed is, you have performed capability deck theatre with better vocabulary.

End with a written residual request: three measures you most need, owners, and a date. Vague “let’s keep talking” is how falsification dissolves back into relationship maintenance. The validation pack is a work order, not a vibe.

After the meeting

Same day, update thesis status in the library: confirmed direction, modified, or killed. Capture exact phrases the client used for their system — those phrases are residual gold for the next compile in the industry. If they promised numbers, diary the chase. A Friction Thesis that dies of neglect after a good meeting is still a method failure.

Side-by-side: three openings in the same room

Print that contrast for your own team if you train others. The third posture is teachable only if people feel the first two in their bones as failures, not as “styles.”

Marketplace of One already said to perform company-specific research before the engagement, apply your kernel, synthesise a distinctive recommendation and make the artefact the first proof. The upgrade this book adds is that the proposal no longer merely demonstrates “I understand your company.”

It demonstrates: I can detect a structural business problem you had normalised, distinguish it from nearby symptoms, propose a different commercial mechanism and make my diagnosis easy to falsify.

That is almost the entire capability many AI advisors intend to sell. The proposal therefore performs the method:

outside evidence
→ structured friction model
→ competing diagnoses
→ visible rejection
→ proposed recomposition
→ testable commercial offer

Where the Friction Thesis sits

The thesis is not always a separate PDF. It can be:

• Section 1 of a Mo1 proposal — the diagnostic spine before recommendations;
• a leave-behind from the falsification meeting that later becomes the proposal’s core;
• the research object a compiler produces before narrative prose is generated.

What matters is schema fidelity: the fifteen fields, the retained losers, the kill condition. Without those, you are back to eloquent company research.

Recomposition without product design

The proposal may point at a recomposed first commercial unit — shared evidence before transformation pricing — without qualifying the full successor offer. Route readers who need the commercial design object to the successor-offer work; route readers who need a product family specimen to preparedness. This book’s integrity rule holds: stop at the falsification conversation and the hypothesis of the unit change. Do not smuggle a full product blueprint into a diagnosis chapter and call it “necessary.”

Receipts for you, tests for them

Keep the two citation ethics distinct inside the same document. Claims about your capability carry receipts (prior work, method, frameworks). Claims about their constraint carry tests (validation pack). Mixing the two — using your brand to bully a diagnosis, or using their public website as if it were an audit — is how meta-credibility collapses.

Why this beats a capability deck

A polished capability deck says: we are the sort of firm that could help. A Friction Thesis inside a proposal says: we already began the work of helping, and we built a machine that can be wrong in public. Prospects drowning in “tell us your AI problems” workshops recognise the difference immediately — not because the prose is prettier, but because the epistemic posture is rarer.

Three receipts, upgraded

The Proposal Compiler’s receipts — research, framework application, visible rejection — still apply. The Friction Thesis Compiler sharpens each:

• Research receipt: not only “we read your site,” but “here is how your public expression instantiates or resists the industry seed.”
• Framework receipt: not only “we have a method,” but “here is the fifteen-field object and the classification of wrong friction.”
• Rejection receipt: not only “we considered options,” but “here are five structural rivals and the specific reason each lost as primary.”

Add a fourth, which this book insists on: the test receipt — claims about them carry the validation pack and kill condition in the same artefact as the claim. Without that fourth receipt you have performed intelligent arrogance.

What not to put in the proposal

• Fabricated ROI for the target firm.
• A full product blueprint smuggled as “the only answer.”
• Deleted alternatives that made you uncomfortable.
• The word “biggest” without a path to coefficients.

A skeleton proposal map

If you need a shape for the document without inventing a template religion:

1. Opening contract — we nominate; we do not crown; here is the kill condition.
2. Industry seed in one page — transaction grammar the reader can recognise.
3. Company expression — public receipts mapped to seed stubs; deviations called out.
4. Lead Friction Thesis — structural claim and causal chain.
5. Alternatives ledger — John West table with reasons.
6. Assumptions and absences — the honesty layer.
7. Validation pack — questions that earn coefficients.
8. Replacement direction — one altitude of unit change; route to sibling doctrines for design depth.
9. What we propose next — usually the falsification work itself, not a multi-year transformation.

That map is enough to perform the method. It is also enough to fail honestly when residual evidence refuses the nomination. A proposal that can only succeed by remaining untested is not a Friction Thesis — it is a brochure.

How meta-credibility shows up in the buyer’s internal forward

The Proposal Compiler already insisted that a good artefact arms the buyer to sell internally when you are not in the room. The Friction Thesis sharpens what they can defend:

• Not “the consultant seemed smart,” but “here is the structural claim and the public receipts.”
• Not “they have a framework,” but “here are the alternatives they rejected and why.”
• Not “we should hire them,” but “here are the nine measures that would confirm or kill the diagnosis before we buy a large programme.”

That last bullet is commercially powerful. It gives a sceptical CFO or delivery lead a way to engage without accepting a multi-year transformation on charisma. Pre-meeting evidence work and qualification already associate with healthier win economics in consulting and RFP practice generally;² the thesis makes that evidence work visible as an object, not a vibe.

Relationship to the full Mo1 proposal

You still may include capability, team, commercial terms and an implementation outline. Those sections must not bury the diagnostic spine. If the reader only skims, they should still hit: nominated equation, rivals, tests, kill. Burying the honesty layer at the end of page 40 is how organisations reintroduce arrogance while performing thoroughness.

Conversely, do not ship a naked thesis with no path to work together. The path, in this book’s scope, is often “help us complete the validation pack” or “run the falsification measures” — not “sign the transformation SOW.” Paid engagement design after entry is another doctrine. The proposal’s next step should match the epistemic stage you are actually in.

Quality bar before send

1. Can a hostile reader find the kill condition in under two minutes?
2. Can they find at least two rejected structural alternatives with reasons?
3. Is every percentage or superlative either sourced externally or marked as unknown coefficient?
4. Does the replacement section stop at unit altitude rather than inventing a full product?
5. Would you still respect the document if the thesis dies next week?

If item 5 is no, you wrote a pitch, not a Friction Thesis. Rewrite until a clean kill would still leave the artefact looking like good judgment.

After you investigate several data consultancies, the reusable layer should know more than any single thesis:

• the common proposal and SOW transaction grammar;
• where senior labour usually enters the join;
• which business-model patterns create which wrong friction;
• what public signals predict productisation weakness;
• which metrics validate or kill the diagnosis;
• which replacement units have credible buyers;
• what conditions caused prior theses to fail.

Each new target starts from a stronger prior. That is Marketplace of One with a compounding diagnostic substrate — not a CRM full of logos.

The Seed Wiki lesson applies again: the skeleton is valuable because difference becomes legible. A library that only stores “wins” is a highlight reel. A library that stores kills and modifications is a learning system.

The loop that improves the prior

Industry friction seed
→ target-company public compile
→ friction thesis
→ falsification conversation
→ (later) paid product proof / residual learning
→ outcome receipt
→ seed improvement

This book stops at the falsification conversation. The later steps still matter to the library: when a thesis is killed, write the kill into the seed (“this public pattern over-predicted wrong friction when X was present”). When a deviation is stable across three firms, promote it into the industry prior. When a replacement unit repeatedly fails buyability, demote it as a default response path.

What to store as library objects

Second-industry seed sketch — specialised industrial equipment service

To show the library compounds across industries without designing a successor product, here is a sketch only — not a full seed and not an offer blueprint.

Category shape: firms that sell and service capital equipment into industrial customers, with knowledge-intensive continuity needs (uptime, parts, technician judgment, customer-specific configurations).

Candidate transaction grammar (sketch): equipment in field → failure or planned service → customer translates urgency into tickets/calls → dispatcher and technician reconstruct machine + site context → parts and knowledge may not join → visit happens → history poorly captured → next failure restarts reconstruction.

Typical wrong-friction nomination (sketch): the commercial model may price reactive labour and parts while the structural burn sits in repeated reconstruction of machine-and-customer context that never becomes a durable, sellable continuity unit. Public signals might include: service brochure language, job ads for technicians vs knowledge roles, absence of continuity products, case stories of heroics.

What this sketch is not: a full fifteen-field thesis, a validation pack, or the industrial successor offer developed in adjacent conversations. Those product objects are out of scope here. The sketch exists only to prove the method ports: new industry → new seed stubs → same compiler discipline → later library entry.

Why the library is the moat

Anyone can ask better discovery questions for a week. A library of seeds and deviations is cumulative industrial knowledge about where attention and margin burn in the categories you serve. It makes the next falsification meeting cheaper, sharper and harder to fake. It also makes your compiler improve with use — the definition of a compounding system rather than a hero consultant’s notebook.

Anti-patterns that destroy the library

• Logo collecting — storing targets without theses or kill receipts.
• Silent kills — abandoning a thesis in a meeting without writing why; the next agent (or future you) re-learns the same error.
• Seed freeze — treating the first industry seed as scripture for years after deviations have piled up.
• Confidentiality laundering — writing internal coefficients from one engagement into a public seed without generalising and anonymising. The library compounds; it must not leak.

Protect the library the way you would protect a kernel: version it, cite sources, separate public priors from engagement-confidential residuals.

Minimum viable library after five compiles

After five firms in one industry you should be able to open the library and answer, without rediscovery:

• What transaction grammar is default, and what three deviations have appeared?
• Which public signals are strong predictors versus marketing noise?
• Which candidate diagnoses win most often, and which kills recur?
• Which validation questions clients can actually answer quickly?
• Which replacement directions get air-time versus which die in residual testing?

If you cannot answer those after five compiles, you have been doing one-off cleverness, not building a substrate. The Friction Thesis Compiler is only an extension of Marketplace of One if the prior compounds.

How a kill rewrites the seed — worked shape

Suppose three mid-sized data consultancies kill the “price before shared evidence” thesis because each already routes most pursuits through a paid readiness product that marketing under-describes. The library write-back is not “thesis failed.” It is a conditioning rule:

When case studies and sales hiring still look bespoke, check utilisation of any readiness SKU before nominating first-unit wrong friction. Homepage transformation language is weak evidence of default entry path if a paid diagnostic captures the majority of pursuits.

That sentence belongs in the industry seed’s version notes. The next compile starts smarter. Without write-back, the fourth firm costs the same as the first. With write-back, outside nomination gets sharper and false positives fall — which is the only legitimate way “outside inference” becomes more confident over time without becoming more arrogant.

Cross-industry transfer without false analogy

The industrial equipment sketch earlier shows transfer of method, not transfer of the data-consultancy thesis. Wrong friction in capital equipment service may centre on continuity and reconstruction of machine-and-site context; wrong friction in data consulting centres on pricing unobserved transformation. What transfers is: seed schema, competing diagnoses, fifteen fields, falsification meeting, library objects. What must not transfer is the lead sentence from one industry pasted onto another because it sounded good.

When two industries share a rhyme — both price labour before shared evidence of state — you may note the rhyme in a meta-note. You still compile each industry’s seed separately. Rhymes are hypotheses for research, not licences to skip the seed.

Operating ownership

Someone must own the library: version seeds, refuse silent kills, anonymise residuals, and stop logo collecting. In a solo practice that owner is you. In a firm it should be explicit — not “marketing’s content folder” and not “whoever ran the last pursuit.” A compiler without an owner reverts to hero notes. The seed library is the diagnostic kernel of Marketplace of One for commercial friction; treat it with the same seriousness you would treat a code kernel you recompile from.

Doctrine without a Monday list becomes a mood. Here is the operating sequence this book earns. It is not a multi-month transformation programme. It is the work that turns outside nomination into a meeting you can defend.

Field checklist

Integrity rules (non-negotiable)

• Do not award “biggest” without internal coefficients.
• Do not delete rival diagnoses; demote them with reasons.
• Do not invent statistics to sound precise — write the shape.
• Do not design the full successor product inside the diagnosis motion; route when needed.
• Do not skip the kill condition to protect your favourite intervention.
• Do name the target only as your ethics allow; this book’s specimen stays a mid-sized data consultancy.

Where this book stops — and why that is a feature

Paid evidence engagements, offer foundries, and sales storytelling around a single visible defect are other jobs. If you continue past the falsification conversation without noticing, you will smuggle product design and engagement design into a diagnosis compiler and dilute all three. Stop when the thesis has been tested enough to know whether the car belongs on their road.

A one-week operating cadence (shape, not a timeline fetish)

If you want a bounded first run without inventing a programme office:

• Days 1–2: write or refresh the industry seed; generate candidates; freeze attack questions.
• Days 2–3: public compile; score candidates; draft fifteen fields including kill condition.
• Day 4: peer-review the uncomfortable fields — assumptions, absences, losers. If a colleague cannot attack your thesis, it is not ready.
• Day 5: falsification meeting or asynchronous send of the pack with a clear request for residual measures.
• After: write confirm / modify / kill into the library the same day. Stale kills teach nothing.

The cadence is a shape. Some compiles take a morning; some take longer because the industry seed is new. What must not stretch is the integrity sequence — candidates before search, losers retained, kill condition written before the room.

The reformulation to keep

Do not ask prospects to diagnose themselves. Bring them a falsifiable theory of where their business burns attention and margin. Let their internal evidence supply the delta.

You do not need the client to see the car before you do. You need enough humility and machinery to prove whether the car belongs on their road.

Infer the equation from outside. Solve for the coefficients inside. That is the whole method — and the whole ethic.

If you only do three things

If the full checklist is too much for this week, do not abandon the boundary. Do these three:

1. Write the industry transaction grammar on one page.
2. Write five structural candidates and keep all five after you pick a lead.
3. Walk into the next first meeting with a kill condition you are willing to say out loud.

Those three alone already separate you from discovery theatre and from website arrogance. The rest of the compiler makes the separation systematic, reusable and library-grade. But the ethic fits in three lines — which is how you know it is real doctrine rather than a content outline.

Closing map

Part I taught why the client is an unreliable first witness and drew the outside/inside boundary. Part II gave the compiler: seed, reasoning-guided search, fifteen fields. Part III proved it with a complete industry seed, a public compile, five attacked rivals, and a fully filled thesis with validation pack. Part IV turned the artefact into a meeting, a proposal posture, a compounding library and a Monday checklist.

What remains is practice. Pick an industry you actually sell into. Ship a skeleton. Compile one firm. Let a kill condition be allowed to fire. Write what you learned back into the seed. That is the Friction Thesis Compiler at work — an extension of Marketplace of One that compiles diagnosis before it compiles persuasion.