Every fund works the same visible universe; the companies nobody calls are the ones nobody sees. A 15-minute guide to add-on sourcing as coverage: census the vertical, screen against the written thesis, and run it as a radar.
Add-on M&A is the center of gravity of middle-market PE, and every fund's sourcing stack draws on the same databases, filters similar codes, and works the same visible universe. The genuinely uncontacted companies sit outside every stack — invisible for structural reasons, not lack of effort.
Owners of certified shops and regional operators receive near-identical letters from competing funds — same opening claims, same profile-data provenance. Response rates decline industry-wide.
Small operators with unpolished sites carry thin database profiles. The controls integrator coded as software, the water-treatment specialist coded as plumbing — miscategorized and unreachable.
In our screening runs, a fifth or more of confirmed thesis fits never used the category's obvious keywords on their homepage. Nobody calls them because nobody's tooling can see them.
In a mature sourcing market, the remaining edge is coverage — seeing companies competitors cannot. A platform holding the full 545-company universe of its vertical works 245 conversations nobody else is having.
The census runs as two passes — a cheap triage across the full category, then a deep thesis-specific extraction on survivors. Group ownership zeroes a company out entirely, because a subsidiary is not a smaller opportunity — it is no opportunity.
In one specimen run, 25K domains triaged to 17.3K live operators — 30% of BD time saved before any thesis logic ran.
Thesis is re-runnable — sharpen your ICP after two closings and re-screen without re-doing the census.
Of the fifteen signals in the framework, five do disproportionate work in add-on screening, and each rewards platform-specific tuning.
Runs both directions: negatively, it screens out subsidiaries via footer disclosures, “proud to join” announcements, and parent-branded career portals — roughly one in ten keyword-perfect candidates fail. Read positively, the same disclosures map every active consolidator in the vertical — competitive intelligence most funds buy separately.
The 70% score weight, and why generic category lists underperform. A platform's thesis is never “equipment repair” — it is “rotating equipment services for process industries, field-service-led, EASA preferred.” Each clause is textual, not firmographic, which is why database filters cannot compose it and full-site reading can.
Contract language is the multiple-driver in most industrial theses, and it is pure text: planned maintenance agreements, calibration recall programs, chemical service contracts, rental fleets with service attached. One specimen fit paired “family owned and operated” with a 65-employee headcount and explicit maintenance-program language — visible in text, invisible in firmographics.
Certifications gate thesis fit: AS9100 for aerospace, ITAR for defense, ASME stamps for pressure equipment, EASA for motor repair. We capture credentials as exact claim text because “ISO 9001:2015 certified” and “working toward ISO certification” are different companies — and credential density tells the platform how deep the qualified pool runs.
Add-on logic is density logic: fill the service-territory map, acquire the adjacent metro, avoid cannibalizing existing branches. Websites state HQ and branch locations directly, with the crucial distinction between owned locations and partner-network mentions that firmographic data routinely conflates.
Coverage determines who you can call; evidence determines what happens when you do. Owners pattern-match for seriousness, and verbatim familiarity with their published story is the cheapest credible signal of it.
“We track your industry and are impressed by your business…”
“I read that you have been EASA-accredited motor repair since 1987, family-owned, with planned-maintenance programs across Michigan…”
A universe mapped once is a photograph of a moving object. Platforms that compound sourcing advantage operate coverage as a cadence — monthly or quarterly deltas that keep the map permanently current.
Companies that form, pivot into the vertical, or newly reach screening threshold — surfaced automatically in each delta cycle.
Each “acquired by” disclosure removes a target from the runway and maps a rival consolidator — two products from one stream.
Sites going dark or companies closing — flagged before a BD associate wastes outreach on a dead lead.
Honest boundaries, because a sourcing method that oversells is just another convergent vendor.
No revenue, EBITDA, or valuation estimates. Size arrives as proxy bands — headcount, facility footage, branch counts — and hardens on the first call.
No “likely seller” flags, no distress detection, no owner-age profiling. We consider those claims unsupportable from website evidence.
A census tells you who exists. Then the fund's craft — the call, the visit, years of patient contact — takes over exactly where it always has.
One vertical from our published specimen set — industrial equipment repair — illustrates the full arc: 545 eligible independent US operators, every one scored and evidence-quoted.
Michigan, EASA-accredited, 65 employees, explicit family-ownership language, maintenance-program offerings. Mandate fit 94.8, every claim quoted.
Ohio rotating-equipment house, founded 1991, 44,000 sq ft facility, strong recurring language — invisible to keyword screens because homepage vocabulary never matched.
“A subsidiary of a larger industrial group” — one in ten keyword-perfect candidates, documented with the sentence that disqualified them.
For the platform holding this deliverable: 545 knowable conversations, a rival map as a by-product, and an outreach file where every letter opens with the target's own words.
One email. We send the specimen report the same day — every company scored and ranked with full signal transcripts, plus the exclusions we documented and why.
Request the specimen report