Two objects that share one acronym
“TAM” describes two entirely different work products, and confusing them is the source of most bad market maps in acquisition work. This guide covers how to build the one that supports decisions: a census of participants, built from the open web.
Market-research TAM
A dollar figure from survey averages, analyst estimates, and growth assumptions. Sizes an opportunity for a slide — no names, no rows, no auditability.
Acquisition TAM (census)
An enumerated list of actual companies with names, because you intend to evaluate and contact them. Every row is inspectable; every error is individual and fixable.
Compounding errorsExtrapolated TAMs carry invisible errors — skewed surveys, adjacent markets, mismatched geographies — and nobody can audit the final number.
Row-level auditabilityCompany #341 is misclassified? Look at it and fix it. An IC asking “how do you know?” gets walked through the answer one row at a time.
Who needs thisPlatforms mapping add-on runway, corp dev sizing whitespace, funds testing a thesis before committing a searcher-year.
Why the web makes a census tractable
Essentially every operating B2B company maintains a website describing what it does, for whom, and with what credentials — making the population of category websites a close proxy for the population of category participants.
24.7M
business & finance sites
367,478
domains in one industrial category
How many independent calibration labs in the US?254, after triage and independence screening — a real count, not an estimate.
What share hold ISO/IEC 17025?Answerable — because the census reads accreditation claims from each company's site.
How many were acquired recently?Answerable — ownership disclosures are extracted and logged as group-membership evidence.
None of these questions are answerable from a spend-based TAM — and all of them are the questions an acquirer actually needs.
The mechanics: triage, then deep extraction
A census run has two passes, and the separation is what makes it affordable at any scale.
Answer cheap questions at volume across every classified domain:
- Is this a company, or a directory / association / job board?
- Is it alive — dated content, active hiring, current copyright?
- Is it in scope geographically?
Specimen result: 25,000 domains → ~17,300 live operating companies. Nearly a third gone — the first real fact most market maps never learn.
Full-site reading of triage survivors against a 15-signal framework:
- Ownership & independence
- Service model, recurring-offering language
- Certifications as exact claim text
- Footprint, leadership, hiring, activity
Every extracted claim carries a verbatim quote and a source URL.
Works at any scale. If your census is 400 domains from registries, pass one is a two-minute liveness check per site, pass two is a structured read against a written rubric. The discipline — cheap questions first, evidence verbatim, every domain accounted for — is identical.
Segment on the splits buyers price, not the codes statisticians use
Industry codes tell you “industrial machinery services” — not which participants are service-led, independent, contract-based, or certified. Those splits determine multiples, integration difficulty, and add-on runway — and they live in website language a census can measure.
- Service-led vs product-led. Field-service language, maintenance programs, and technician hiring versus catalog pages and dealer locators. The same code contains both; they trade very differently.
- Consolidation rate. What share of the vertical already discloses group ownership? This single number is your competition gauge, your multiple context, and your remaining-runway estimate at once. In our industrial runs, roughly one in ten keyword-perfect candidates was already group-owned — and the rate varies sharply by subvertical.
- Recurring-offering density. The share of participants showing contract, program, or consumables language. Filtration's consumables recurrence looks nothing like machining's project work, though both sit inside one “industrial” code.
- Credential density. The share holding the certifications that gate real work: ISO/IEC 17025 in calibration, AS9100 and NADCAP in aerospace supply chains, ASME stamps in pressure equipment, UL 508A in control panels. Credential density is a proxy for both quality floor and barrier height.
- Founder association. Roughly half of confirmed fits in our industrial specimens carried explicit founder or family evidence — a segmentation with obvious meaning for anyone whose thesis involves ownership transitions.
Each of these is a percentage you can compute over an enumerated population, footnote with quoted evidence, and defend in committee. None appears in a market report.
What the numbers look like in practice
Specimen numbers from one full run — global category population of 367,478 domains, US-focused triage of 25,000 reducing to ~17,300 live operating companies:
| Subvertical | Eligible independent US companies |
|---|
| Precision machining & fabrication | 702 |
| Industrial equipment repair | 545 |
| Automation & controls integration | 534 |
| Material handling services | 513 |
| Compressed air systems | 276 |
| Calibration & testing | 254 |
| Water treatment | 221 |
| Boiler & steam services | 172 |
| Filtration & separation | 109 |
| Surface finishing | 93 |
Numbers at this grain change acquisition strategy directly.
Platform add-on laneCompressed air holds 276 independents — enough for years of M&A, small enough for complete coverage this quarter.
Search fund thesis testSurface finishing is 93 companies — decide with open eyes whether that survives realistic response rates.
Board-deck certainty“We evaluated 100% of the universe” — literally true, not an approximation.
Note what the table does not contain: dollars. An enumerated TAM can be dollarized later from the bottom up — and that estimate inherits the census's auditability. Starting from dollars and working backwards produces neither.
The consolidation ledger: measuring who already owns the market
Group membership surfaces from footers, about pages, and news announcements during extraction — logged systematically, these disclosures assemble into a ledger of active consolidators, their acquisitions, and their velocity.
PE platforms
Competitive intelligence that normally costs a retained buy-side mandate — the add-on radar shows your runway and your rivals' recent tracks.
Corporate strategy
Consolidation velocity is the difference between a market with a five-year window and one with eighteen months.
Investment committee
“11% group-owned today, up from 8% at last refresh” — a thesis anchored in observable motion, not narrative.
“In December 2024, the company was acquired by a global industrial group” — one dated sentence from one news page, worthless to a keyword filter, structural information for a market map.Material handling specimen exclusion
Honest boundaries: what a web census cannot claim
Every method should publish its error model — an IC trusts a map more, not less, when its edges are drawn.
Coverage gaps
Sole operators with only social-media presence, businesses inside closed supplier networks, and the newest entrants may be under-represented. Quantify this per vertical; do not hand-wave.
Size estimation
Websites support size bands from proxies — headcounts, facility footage, branch counts — never revenue figures. Publish the proxy and the band, never a fake-precise dollar estimate.
Insufficient evidence
A few percent of any population reads as too thin to classify. Report them as a category rather than guessing or silently dropping — the count of unknowns is itself a market fact.
Snapshot decay
A map more than two quarters old misleads in proportion to deal velocity. Build monitoring deltas into the plan and date-stamp every number the memo cites.
From census to committee: packaging the map
A census earns its keep when it survives the room. Structure the deliverable in three layers.
1
Population arithmetic
Raw domains → live operating companies → eligible independents. State the reduction at each step — the reductions are findings.
2
Segmentation cuts
Consolidation rate, service-model split, credential density, recurring-offering share — each as a percentage with specimen quotes.
3
The ranked list
Per-company evidence for outreach. When the deal team returns with a sharpened thesis, re-screening the universe is fast — enumerated populations compose.
Write the methodology paragraph first: source of domain population, triage criteria, extraction rubric, evidence standard, date of run. It converts the map from an artifact into a process the institution can re-run.
Our productized version is TAM mapping for industries — full universe plus deep shortlist from €9,900, custom ICP re-runs included. The packaging advice stands whatever machinery you use: enumerate, segment on priced splits, publish boundaries, date the snapshot.
Dollarizing a census without wrecking its credibility
Committees sometimes still want a dollar figure. There is a defensible way: bottom-up banding from the enumerated population.
1
Sort into size bands
Use visible proxies — stated headcounts, named staff, facility footage, branch counts.
2
Assign revenue ranges
From ground truth: portfolio comparables, completed diligences, verified broker teasers.
3
Present as a range
Assumptions on the same slide. A skeptic can attack them without the count collapsing.
Rule 1: No revenue in company files
A record should say “65 employees, 44,000 sq ft (stated)” — never a revenue figure someone will later mistake for data.
Rule 2: No per-company precision
The census's value is that population-level numbers are hard. Spending that credibility on soft company-level dollar guesses is a bad trade.
The by-product is often more useful than the total: knowing that 545 eligible independents skew heavily toward the 10–50-employee band, with a thin tail above 100, tells an acquirer more about deal availability, integration load, and realistic multiples than any aggregate dollar figure. It is a distribution no survey-based report can draw.
Frequently asked questions
A market report estimates spend in a category using surveys and analyst models; it names a handful of large players and extrapolates the rest. A census enumerates the participants — every company, with evidence — and computes market structure from the enumeration. Reports answer “how big is the spend?”; a census answers “who exists, who is independent, who is credentialed, who has been bought?” Acquirers need the second set. The two are complements: a report for the macro paragraph of the memo, the census for every paragraph that follows.
A commissioned market report typically runs five figures and returns no names. Our proof project starts at €4,900, and a full universe with deep shortlist from €9,900 — with custom ICP re-runs included, because theses sharpen. A manual census of a narrow vertical costs mostly time: a few weeks of disciplined registry work and site reading for a population in the low hundreds. The comparison worth making is per decision supported, and a census supports outreach, diligence framing, and competitive mapping from the same artifact.
Deduplication is a real workload and worth doing explicitly. Multi-brand operators disclose themselves through shared footers, addresses, phone numbers, and cross-links; the census should collapse them to one operating company with brands noted, since the acquirable unit is the company. The reverse case — one site covering multiple legal entities — is rarer and usually resolvable from imprint and about pages. Expect low single-digit percentages of both in industrial verticals, and log every merge so the arithmetic stays auditable.
Yes — that is one of its main economic properties. The expensive step is enumerating and reading the population; a thesis is a scoring function applied on top. Re-running a new ICP against an already-mapped universe reprices every company against the new criteria without re-doing the census. We include custom re-runs in engagements for this reason, and manual mappers get the same benefit by keeping extraction notes structured: your rubric can change without your fieldwork repeating.
Match the refresh to the vertical's deal velocity. Quietly fragmenting spaces drift slowly; actively consolidating ones change measurably in a quarter — new group ownership disclosures, rebrands, dead sites. A practical default is quarterly deltas (what entered, what exited, what changed ownership) with an annual full re-read. Deltas are cheap relative to the original census because triage state is already known. Whatever the cadence, date-stamp the numbers a memo cites; an undated census count is a rumor with a spreadsheet.
Keep reading
What we refuse to sell: no “ready to sell” flags, no revenue or EBITDA guesses,
no owner-age profiling, no distress detection — and no engagements in consumer-captive verticals.
Read our
standards; serious buyers tell us this page is why they trusted the rest.