Add-On Acquisition Sourcing: Finding Bolt-On Targets

Why add-on sourcing is a different problem

An add-on acquisition is a company bought to extend an existing platform's operations, customers, capabilities or geography. Sourcing one differs from sourcing a platform because the buyer is already known. The question is no longer "which business is worth owning?" but "which businesses can this particular company absorb, and in what order?"

Bain & Company's 2019 buy-and-build chapter makes the operating point explicit. It defines buy-and-build as using a well-positioned platform company "to make at least four sequential add-on acquisitions of smaller companies", and it notes that "the platform company usually makes the add-on acquisitions—not the PE fund", which is why it stresses the platform's free cash flow ^[1]. Two consequences follow for research. First, add-on criteria should be written from the platform's facts (its branches, licences, customers and integration capacity), not from a generic sector screen. Second, the people who will act on the list are often platform managers, so each record has to be legible to someone outside the deal team.

This article covers the front end of that work: criteria, market map, entity resolution, knockouts and prioritisation. Stage definitions, ownership of records and pipeline metrics are covered in origination pipeline stages and are not repeated here.

What the Bain figures measure, and what they do not

Add-on activity is often quoted as a single share, but published figures use different bases and should not be compared with each other.

• Share of deal value. Bain's 2025 outlook states that "still-high interest rates continued to take a chunk out of add-on transactions, which represented 11% of buyout deal value, vs. a peak of 40% in 2015", adding that "bolt-on acquisitions were less affordable" as financing costs rose ^[2]. This is a value share. The sentence does not name a region or a year for the 11%; it sits in the report's review of 2024 (the report was published in March 2025), so read it as the latest year covered, not as a confirmed calendar-2024 figure.
• Headline totals exclude add-ons. The same report gives 2024 buyout investment value of $602 billion "excluding add-on deals", and its chart notes repeat that exclusion. Bain's 2026 outlook does the same, and its notes add that deal value counts only deals with disclosed value while deal count includes deals with undisclosed value ^[3].
• Shares within add-on counts. The 2019 chapter measures something else: in 2003, 21% of add-on deals were at least the fourth acquisition by a single platform, a figure "closer to 30% in recent years", and in 10% of cases the add-on was at least the tenth ^[1]. That is a share of add-on deals, not a share of all buyouts.

Because add-ons are by design smaller than their platforms, and many small deals have no disclosed price, a value share and a count share can differ widely. This article does not cite a count-based share of buyouts, because none was read on a primary page today. For a sourcing team, the practical message is narrower: the figures describe market conditions, not the size of any one platform's opportunity. That has to be counted bottom up.

Step 1: Derive add-on criteria from the platform

Bain's 2019 chapter defines a company's core as the clients, products and services that drive most of its profits and give it its competitive advantage, and frames each acquisition by its overlap in "customers, costs, channels, capabilities and competition" ^[1]. Those five overlaps are a usable template. Each criterion should cite a platform fact, state whether it is a knockout or a score, and name the evidence that would settle it.

Fictional platform. Northfold Mechanical (fictional) is a commercial heating, ventilation and air-conditioning (HVAC) service contractor with four branches in California. It holds California Contractors State License Board (CSLB) classifications C-20 (Warm-Air Heating, Ventilating and Air-Conditioning) and C-36 (Plumbing), and wants to add C-38 (Refrigeration) capability ^[4]. Its integration team can onboard roughly one add-on per quarter.

Add-on criteria template

The size floor answers Bain's two diligence questions: whether add-ons will carry lower valuations than the platform, and whether they will be "meaningfully accretive", that is, smaller than the platform "but not so small that acquiring them doesn't add value" ^[1]. The radius should be drawn from the territory the platform actually serves; the geographic infill thesis note sets out how to define a coverage gap and what evidence disproves one.

Step 2: Build the market map from registries and licence records

Start from the record that defines the activity rather than from a company database. In a licensed trade, the licence register is the closest thing to a complete list of firms entitled to do the work.

• State licence boards. The CSLB lookup searches by licence number, business name and personnel name. A business-name search "will return up to 50 names", with further pages behind a link, and a separate tool creates "a list of licensed contractors by City or ZIP code" ^[5]. City and ZIP lists overlap, so expect duplicates when several areas are pulled.
• Healthcare provider registries. For clinical platforms, the National Plan and Provider Enumeration System (NPPES) API searches by taxonomy, organisation name, city, state and postal code. It returns at most 200 results per request and can skip up to 1,000 records, so one query yields "a maximum of 1,200 records over six requests"; broad searches must be split, for example by postal code. CMS also warns that issuance of a National Provider Identifier (NPI) "does not ensure or validate" that the provider is licensed ^[6].
• Company registries. The UK Companies House advanced search filters by company status, Standard Industrial Classification (SIC) code, incorporation and dissolution dates, company type and registered office address ^[7]. A registered office is a legal address and is often an accountant's office, so it is a weak signal for an infill radius.
• A denominator. The Census Bureau's County Business Patterns reports establishments with paid employees by industry and employment size; the most current data on the page is for 2023 ^[8]. It counts establishments, not firms, and names no one, so use it to sanity-check coverage ("does our list look plausible against the establishment count?"), not as a target list.

Geocode the operating address of every record and compute the distance to the nearest platform branch. Where a source holds more than one address, keep each with its type: the NPPES output separates the primary practice location from the mailing address ^[6].

Step 3: Resolve brands to legal entities

A raw pull counts records, not companies. One business can appear under several licences, several city lists, a trade name and a former name. The NPPES documentation illustrates the point: an organisation-name search examines the legal business name, "doing business as" (DBA) names, former legal names and other names, "therefore, the results might contain an organization name different from the one entered" ^[6].

The unit of a market map should be the legal entity that owns the operations, with its brands, licences and sites attached beneath it. Resolution has three outcomes that matter here: merge (a duplicate of an entity already on the map), match (a new entity) and hold (two plausible entities, or none). Held records stay on the map with a stated reason; they are not knocked out, because an unresolved identity is a research gap, not evidence of a poor fit. The method for tracing a trading name to its filing entity is set out in brand to legal entity.

Resolution is also where many consolidator-owned businesses surface. A local brand whose licence lists the same officers as a regional group, or whose website footer names a parent, is usually already someone else's add-on.

Step 4: Apply integration-fit knockouts before scoring

Knockouts come before scoring because they are cheaper to establish and they stop analysts spending time on businesses the platform cannot buy or absorb. Each knockout should be a single question with a recorded source:

1. Is the required licence classification active on the licence record?
2. Is the operating address within the radius of a platform branch?
3. Does the company's own material show work in the platform's customer segment?
4. Is the company already part of a group, a franchise system or a sponsor-backed consolidator?
5. Does observable evidence put the company below the size floor?

Write down the evidence for every knockout. A knockout made on stale evidence (an old website, a lapsed licence later renewed) is the most common reason a good candidate is lost, and the record lets a reviewer reverse it.

Step 5: Prioritise, with owner-readiness as evidence, not assumption

Bain asks of a sector's "white space" whether "a lot of founders [are] reaching retirement age" ^[1]. Public records rarely answer that directly, and owner age should not be guessed. What a researcher can record are observable proxies, each labelled as a proxy:

• Tenure. Licence issue date or entity incorporation date, which show how long the current structure has existed.
• Key-person dependence. A licence whose qualifying individual is also the sole officer, found through a personnel search on the licence board.
• Officer changes. Recent appointments or resignations in the entity filings, which can indicate succession activity in either direction.
• Named authorised officials. NPPES records can name an authorised official ^[6], which identifies whom to approach, not whether they will sell.

None of these establishes an intent to sell. They order outreach. A simple, transparent score works better than a weighted model nobody can audit:

Tier 1 is 70 points or more, tier 2 is 50 to 69, and tier 3 is below 50.

Worked example: a fictional market-map funnel

All figures below are fictional and illustrate the arithmetic only.

Three points come out of the arithmetic. First, the 600 raw records became 372 companies; a team that reported "600 targets" would have overstated its market by more than half. Second, the largest single knockout was the size floor, which suggests that criterion deserves a second look with the platform before it is fixed. Third, the 31 held records are unfinished research, not rejections; a few may belong in tier 1.

Cadence and integration capacity

The list is only useful if the platform can absorb what it finds. Bain's 2019 chapter says "value creation depends on a steady cadence of acquisitions", calls for "foundational infrastructure" such as robust IT systems and repeatable financial and operational models, and advises identifying bottlenecks such as "IT systems, integration team" early ^[1]. Time also matters: Bain's 2026 outlook reports buyout holding periods at exit "hovering around seven years", with almost 40% of companies held more than five years, on data that excludes add-ons ^[3]. Longer holds give a programme more time, but each year of delay still shortens the period over which an add-on's contribution can be demonstrated.

In the fictional example, a platform that integrates one add-on per quarter has more tier 1 candidates (14) than it can close in two years. That is the right shape: the list should exceed capacity, because many owners will not engage. Refresh the map on a fixed schedule, since licences lapse, new firms register and competitors buy candidates.

Integration outcomes are not uniform either. Kim (2025) studies serial private equity acquisitions in U.S. hospitals, classifying targets as platforms or add-ons; the working paper's abstract (not yet peer reviewed) reports profitability gains at platform hospitals and contractions in patient volume and profitability at local add-ons, which the author attributes to deliberate reallocation ^[9]. It is one study of one regulated industry, but it is a reminder that an add-on's standalone figures may not survive integration. Borell and Heger ^[10] point the same way from a different angle: their abstract reports that investors pair platforms with lower capacity utilization and returns with add-ons that have higher utilization and returns, and that profitability improves only for firms with increasing industry-adjusted utilization.

Checklist before outreach

• Criteria table written from platform facts, each with a rule type and evidence source.
• Market map built from the activity-defining register, with an establishment-count sanity check.
• Every record resolved to a legal entity, or held with a stated reason.
• Knockouts recorded with source and date.
• Scores and tiers reproducible by a second reviewer.
• Target list larger than the platform's integration capacity, with a refresh date.

Docket's Triage agent covers the identity, de-duplication and disqualifier steps above, before Scout answers the remaining criteria from retained pages.

Sources

1. Buy-and-Build: A Powerful PE Strategy, but Hard to Pull Off (Global Private Equity Report 2019) — Bain & Company, 2019; buy-and-build definition, platform makes the add-ons, add-on count shares (21%, about 30%, 10%), diligence questions, core and overlap framework, white space, infrastructure, cadence and bottlenecks. Accessed 2026-10-28.
2. Private Equity Outlook 2025: Is a Recovery Starting to Take Shape? — Bain & Company, 3 March 2025; add-ons at 11% of buyout deal value vs. a 40% peak in 2015, interest-rate explanation, $602 billion 2024 buyout value excluding add-ons. Accessed 2026-10-28.
3. Private Equity Outlook 2026: Gaining Traction — Bain & Company, 22 February 2026; chart notes excluding add-ons, disclosed-value vs. count bases, holding periods at exit around seven years, almost 40% held more than five years. Accessed 2026-10-28.
4. When Buy-and-Build Breaks: Serial Private Equity Acquisitions in the Hospital Industry — Sungil Kim, SSRN working paper, 2025; abstract (SSRN page blocks automated access; abstract read from the Crossref record of the SSRN DOI). Accessed 2026-10-28.
5. Sources of Value Creation Through Private Equity-Backed Mergers and Acquisitions: The Case of Buy-and-Build Strategies — Mariela Borell and Diana Heger, ZEW Discussion Paper 13-094 / SSRN, 2013; abstract read on RePEc (https://econpapers.repec.org/RePEc:zbw:zewdip:13094). Accessed 2026-10-28.
6. License Classifications — California Contractors State License Board; C-20, C-36 and C-38 classification names. Accessed 2026-10-28.
7. Check a License or Home Improvement Salesperson Registration — California Contractors State License Board; search options, 50-name result pages, list by city or ZIP code. Accessed 2026-10-28.
8. NPPES API Help — Centers for Medicare & Medicaid Services; search fields, 200-result and 1,200-record limits, organisation-name matching across name types, address types, NPI not a licensure check. Accessed 2026-10-28.
9. Advanced company search — Companies House (UK); available search filters. Accessed 2026-10-28.
10. County Business Patterns (CBP) — U.S. Census Bureau; establishments with paid employees by industry and employment size; 2023 the most current data year listed. Accessed 2026-10-28.