The first page of "add-on acquisition strategy" is mostly finance training and accountant content. Wall Street Prep walks through the multiple-arbitrage math, Financial Edge repeats the same arithmetic with different examples, and CohnReznick explains why sponsors like buy-and-build. All of it is correct. Almost none of it is about sourcing.
From the data layer, add-on sourcing and platform sourcing look like two different jobs. They share the word "sourcing" and almost nothing else.
Cadence and process
A platform deal runs once. The same thesis, run across the hold, runs many add-ons against that platform — buy-and-build is the dominant North American PE structure now, and CohnReznick's piece on the strategy walks through why add-on volume is where the multiple-arbitrage math comes from. The cadence is not "twice as fast." It runs faster, at a smaller average check size, against a target universe that is also denser inside the platform's lane. Set up an add-on process the way a platform process runs (kickoff memo, screen, LOI, nine months of diligence) and the firm closes one add-on a year and the thesis dies of math.
The firms that close add-ons at volume run a smaller, faster process. Four-week diligence. A standing integration team that does not stand down between deals. A CRM that holds every target inside the platform's geography, scored against the integration plan, refreshed weekly. The constraint is the sourcing engine, not the capital. The multiple-arbitrage spread works on paper. It only works in practice if the sourcing layer can keep feeding.
Targeting inputs
In a platform deal, the targeting universe is "every operator in the thesis vertical that hits the revenue band." In an add-on motion the universe is narrower and weirder: every operator close enough to the platform's footprint that integration becomes cheap. The radius can be a short drive from an existing branch, one MSA over, or inside a customer-referral graph the platform already runs.
This flips the data input. The most valuable input for add-on flow is not a vendor's industry filter. It is the platform's own operating data: branch locations, customer postal codes, technician routing density, gaps in the existing service map. None of that lives in PitchBook or any other industry-filtered vendor product, because the vendor doesn't know which platform the firm just bought. Cross the platform's own data against an operator universe pulled from secretary-of-state registries, license boards, and the regulatory filings the vertical actually runs on, and the target list collapses from "every operator in the vertical" to a much smaller set where the integration savings are real.
A vendor like PitchBook is built for the platform deal. The add-on motion needs a parallel layer, owned by the firm, that takes the platform's operating data as the primary signal.
What an add-on candidate actually looks like
Strip away the finance-training framing and the on-the-ground profile is specific. Sub-scale operator relative to the platform, owner near retirement, sitting inside the platform's geography. A handful of trucks where the platform runs a fleet. A couple of locations where the platform has a dozen. The owner is past normal retirement age, has been on the corporate filing for a long stretch, and has not changed their LinkedIn role since the LLC was formed. Morgan & Westfield's glossary describes the add-on as adding "a few specific skills or capabilities" to the platform, which is the polite version of the same thing.
The signals that surface this operator invert the platform-sourcing signals. Platform sourcing leans on growth indicators and hiring velocity. Add-on sourcing leans on the opposite: tenure on the corporate filing, owner age proxies, no recent capital raise, no website refresh in five years. The platform gets bought because it is about to grow; the add-on gets bought because the owner has stopped trying to.
CohnReznick argues that the best sponsors win with add-ons because they integrate aggressively. True. But integration is downstream of the sourcing decision. Pick the wrong neighbor and integration cost eats the multiple-arbitrage spread.
Integration inside the diligence loop
Solomon Edwards calls integration the make-or-break factor for add-ons. In LMM practice, what that means is the integration lead sits inside diligence, not after it.
On a small add-on going into a mid-size platform, the CFO doing the close is also the CFO running the integration. The ops lead reviewing route density is the same ops lead who will absorb the trucks. If diligence and integration are different people, diligence underwrites a deal integration cannot execute, and the spread compresses.
The practical move: the standing integration team gets a fixed line in the process Montage Partners describes, and the sourcing team only surfaces candidates that pass an integration screen before they hit the deal log. Geography, payroll system, route overlap, customer concentration on the platform's existing accounts. Targets that fail do not move forward, regardless of how attractive the multiple looks on paper.
The owned-pipeline argument, applied here
We have written before about why a firm should own its sourcing pipeline instead of renting a list. The add-on motion is where that argument stops being abstract. A platform deal gets sourced once; renting that list is annoying but survivable. An add-on motion gets sourced many times over a hold, against a target set defined by the platform's own operating data. CLA's add-on guide tells you to "start with the end in mind" and write a strategic vision. Fine. The strategic vision the firm should actually be writing on day one is which neighbors get bought, in what order, off whose data.
That layer cannot be a vendor product. The vendor does not have the platform's branch coordinates, customer postal codes, or technician routes. Even if it did, the vendor sells the same data to every sponsor running a similar thesis. The targeting edge in add-on sourcing comes from crossing public operator data with proprietary platform data, and that cross only exists inside the firm.
The build itself is small relative to a platform diligence: a target index scoped to the platform's geography, a scoring function tuned to the integration screen, and the plumbing so the index refreshes against the platform's operating data on whatever cadence the deal team wants. After the first add-on closes, the same system runs against the expanded footprint for the next one. The marginal cost of sourcing each subsequent add-on, run this way, gets close to zero.
Two pipelines
Here is what the finance-training pages miss. Platform sourcing and add-on sourcing share almost no infrastructure. Platform sourcing is a national funnel, vertical-filtered, tuned to growth signals, run a few times a year against a target universe that does not change much from quarter to quarter. Add-on sourcing is local, tuned to integration economics, run on a tighter cycle against a target universe defined by a portfolio company the firm did not own twelve months ago and will not own in five years.
The signals, the data inputs, and the cadence all invert, and the team running it is different too. For the add-on motion it has to include someone from the platform's operating side, because the integration screen is the gate. An LMM sponsor that builds one good pipeline and runs both motions through it underperforms the sponsor that builds two. Trying to make one pipeline do both jobs is where the multiple-arbitrage math quietly dies.
