The Acquisition Target Screen: Thesis to a Ranked, Buyable Longlist

This playbook takes one approved acquisition thesis and turns it into a ranked longlist of 15 to 30 private companies that fit the thesis and could actually be bought. Every row carries its evidence, and every company you cut gets a dated reason. It is for corp dev analysts, strategy teams and talent-intelligence researchers who have to defend that list at the next strategy review. Each stage says what to do, roughly how long it takes, and what done looks like.

What does a finished acquisition target screen look like?

A finished screen is two documents: a ranked longlist of 15 to 30 private companies, and a rejection log covering everything else in the universe. Each longlist row carries:

• its source and the date it was pulled
• a size figure labelled fact or estimate
• its ownership evidence
• a fit grade and an actionability grade

Each rejected company carries a dated reason code.

Most screens skip the second document, and that is why the same rejected names come back every quarter. The log is part of the deliverable, not housekeeping. A review that sees only the survivors has no way to check that the cut was fair.

How steep should the cut be? Published examples give start and end counts but never the attrition at each gate. Nobody has a public benchmark for how many names should fall at the size filter compared with the ownership filter. What the examples do show is the overall shape:

I derived the pass rates from the published start and end counts.

What the review calls each stage varies. The practitioner benchmark calls 50 to 200 companies a "longlist", which roughly matches your filtered survivor set. The 15 to 30 this guide hands over sits at the top of what that benchmark calls a shortlist. Name the stages in your own deck so the review argues about companies, not words.

The acquisition target screen, step by step

The screen runs in seven steps, from a frozen thesis to a ranked handover, and the order of the steps matters more than the tools you use. BCG describes the same arc in three phases: map the industry segments where attractive candidates sit, pick out specific targets from the long list, then prioritise next moves. The seven steps split those phases into gates you can check.

I have no published source for how long each step takes. The figures below are my estimates. The first two steps take about 0.5 to 1 day and 2 to 4 days. The later steps scale with universe size, so time yourself on the first 20 companies at each step and multiply.

Sources disagree on one point of order. Vendor screeners tend to filter on revenue and legal form first, because those are the fields their databases hold. Thesis-first practitioners fix the criteria before any target is named. I side with thesis-first. Filtering on database fields first quietly lets the database decide what the thesis means.

One practitioner notes that "low accuracy is acceptable" at the longlist stage. That holds for size estimates, which you will tighten later. It does not hold for the filters themselves.

Which hard filters go in writing before the first company is named?

Hard filters are the pass/fail criteria taken from the thesis, written down, dated, and signed by corp dev before the universe is built. They cover geography, what the company sells, who it sells to, a size band, the ownership types in scope, feasibility flags and named exclusions. One boutique advisor describes screening "against criteria agreed before any target was named". That is the standard to copy.

Why before? Once names exist, filters drift toward the companies people already like, and a favourite that fails a filter tends to get the filter rewritten. Signing first also gives you a version number. When the thesis changes next quarter, every rejection logged under the old version can be re-tested against the new one instead of re-argued.

A hard filter is binary. If you would grade something on a scale, such as product adjacency, quality of customer base or cultural fit, it belongs in fit scoring later, not here.

THESIS SCREEN: HARD FILTERS
Thesis name:
Filter version:            Date frozen:
Analyst:                   Signed off (corp dev):

H1 Geography (HQ and/or operations):
H2 What the company sells (in / out):
H3 Customer segment served:
H4 Size band, stating the unit (headcount or revenue):
H5 Ownership types in scope (founder / family / sponsor-backed / VC-backed / corporate carve-out):
H6 Feasibility flags (HSR band, UK turnover band) - flag only, never cut:
H7 Named exclusions (existing partners, prior rejections still valid, conflicts):
Hold-age cutoffs for actionability grading (your judgment, written here):

REJECTION REASON CODES
R1 Outside geography
R2 Sells outside thesis scope
R3 Wrong customer segment
R4 Below size floor
R5 Above size ceiling
R6 Ownership type out of scope
R7 Named exclusion
R8 Duplicate or no longer trading

Rejection log columns:
Company | Registry no. | Reason code | Evidence | Date | Filter version

Two notes on writing these filters:

• Use headcount for the size band where you can. Most private-company revenue in this screen will be estimated, while headcount is observed. A headcount floor is easier to audit than a revenue floor built on an estimate.
• Write feasibility as a flag. Merger-control thresholds turn on deal value measured in ways you cannot see from the outside. The size section explains why a hard cut here throws away real targets.

Where does a target universe come from beyond industry codes?

A target universe is the complete, dated set of companies that could plausibly pass the hard filters. Build it from several independent channels so that no single source's blind spots become the list's blind spots. Industry classification codes are one channel, and on their own they are unreliable.

The evidence on codes is consistent. A Census working paper on how businesses classify themselves in the Economic Census found that many respondents wrongly picked sector 81, Other Services. An IRS Statistics of Income study notes that NAICS reporting is prone to error, and that because there is no tax consequence the errors often go unnoticed. No published misclassification rate exists for private acquisition targets, so you cannot correct for the error statistically. You can only work around it.

Pull from every channel below. Record the channel and the pull date on every name.

• Classification codes. These give you broad coverage and a defensible starting point. They miss companies that filed under the wrong code, often into a catch-all sector.
• Industry association member lists. Membership shows the company identifies with the industry. Members may include suppliers, consultants and companies far outside your size band.
• Trade exhibition exhibitor lists. Practitioners use these to find strong mid-sized companies that codes and databases under-represent. A stand also shows the company has a budget for selling.
• Recruitment media. Active hiring is a growth signal and surfaces companies the other channels miss. It can mislead when the hiring is backfill or a single office.
• Company registries. These give you legal entity, status and a registry number to deduplicate on. They say little about what the company actually does.

Deduplicate on registry number wherever one exists, because trading names and legal names diverge. Done means 120 to 800 unique names, each tagged with its channel and pull date.

Then run the five-names test. Ask two operators who know the space, from the business unit or the deal team, to name five companies they would expect on the list. Confirm all of them appear in the universe. If a well-known competitor is missing, it is almost always a code problem, and it tells you which channel to widen before you go any further.

How do you estimate a private company's size from public headcount and filings?

Multiply the company's public headcount by its sector's revenue per employee from the Census Statistics of U.S. Businesses (SUSB). Replace that estimate with a filed or reported figure wherever one exists, and label every value FACT or ESTIMATE. SUSB is the official US source of business counts, employment and payroll by industry and size.

SUSB publishes employment and payroll every year but receipts only for years ending in 2 and 7. Your revenue-per-employee ratio can therefore be up to five years old before the next receipts year is published. Pull the sector figure yourself from the most recent receipts tables for the thesis sector. I don't quote sector ranges here because I have not verified them.

No published error band exists for estimating revenue from headcount, so measure your own. For every survivor where you also have a factual revenue figure, compute estimate divided by actual. Report the median and the spread of that ratio in the handover. A review can work with a stated error. It cannot work with false precision.

The method breaks in predictable places:

• Capital-intensive firms: headcount understates revenue, so the estimate comes in too low.
• Staffing and labour-heavy service firms: headcount overstates revenue, so the estimate comes in too high.

For the headcount input, count people on public profiles who currently work at the company. Refolk returns people across LinkedIn, GitHub and the open web from a plain-English question, so a per-company headcount is one query instead of an hour of manual counting. Treat it as a floor, because not every employee keeps a public profile, and record the date of the count.

Turning estimated size into a feasibility flag

Merger-control thresholds turn estimated size into a feasibility flag you can compute for every row. At the latest US revision, an HSR filing is mandatory and suspensory above a $133.9M size of transaction. HSR is the US premerger notification regime. Suspensory means the deal cannot close until the waiting period ends. The UK regime, by contrast, is voluntary.

Here is how to read the table:

• US. The FTC revises the HSR thresholds every year. At the last revision the minimum rose from $126.4M to $133.9M, an increase of 5.93%. Between $133.9M and $535.5M, a size-of-person test also applies at $267.8M and $26.8M. The filing fee is $35,000 for deals of at least $133.9M and below $189.6M. At the 12.0x median US buyout EV/EBITDA multiple reported for the latest full year, $133.9M implies roughly $11.2M of EBITDA. Targets clearly below that can close without the 30-day wait or the fee. Re-derive the figure whenever the multiple or the threshold moves.
• UK. The CMA regime is voluntary and non-suspensory. The CMA has jurisdiction where target UK turnover exceeds £100M or the 25% share-of-supply test is met. A hybrid test also catches cases where one party has at least a 33% UK share of supply and over £350M of UK turnover. The £10M de minimis exemption, which applies when each party is below £10M of UK turnover, is the only clean exclusion. Everything else is a risk flag.
• EU. I have not covered EU thresholds here. Check them with counsel if any survivor has EU revenue.

How do you grade actionability: ownership, hold age, succession, prior process?

Actionability is the likelihood that a company's owners would engage in a sale within your planning horizon. Grade it from four public signals: ownership type, sponsor hold age, founder succession and prior sale-process history. Fit says you would want the company. Actionability says you could get it. A longlist graded only on fit is a map of the thesis, not a target screen.

Ownership and control records

Ownership type comes from registries and filings, and each one has a specific blind spot.

• UK PSC register. PSC stands for person with significant control. Companies House now holds the only PSC register, as companies no longer keep their own under the Economic Crime and Corporate Transparency Act, and individual PSCs must verify their identity. What it proves: someone holds more than 25%. When it lies: the threshold is strictly more than 25%, so a sponsor holding exactly 25% or less does not appear. The register can never be blank, so a company with no PSC must say so, and that statement can hide a fragmented cap table. Check the shareholder list in the confirmation statement.
• US Form D. This is the notice a company files when it raises money privately under an exemption. It names the executive officers, directors and promoters, the exemption claimed and the size of the offering, and it is due within 15 calendar days of the first sale. What it proves: the company raised outside capital and who ran it at the time. When it lies: a missing Form D does not mean the company is bootstrapped. The SEC has brought enforcement actions against multiple companies for failing to file. A secondary report of an academic study claims more than half of privately funded companies did not file. I have not verified that figure, but even the direction is reason enough not to trust a missing filing.
• Sponsor portfolio pages. Treat these as leads and confirm them in a registry. I have not verified US beneficial-ownership reporting as a usable source for this screen. Check its current status before relying on it.

Sponsor hold age

Hold age matters because the pool of sponsor-owned companies is ageing. Published private-markets data put the median hold for exited US PE-backed companies at about six years, down from a seven-year peak. The median for companies still held is 3.8 years, a multi-year high. Another dataset reports an average holding period of 6.1 years. US sponsor-held inventory stood at 12,552 companies, about nine years of exits at the current pace. Many portfolio companies are sitting inside a normal exit window.

No published source names a hold age at which a company becomes sale-ready, so the cutoff is your judgment and belongs on the signed filter sheet. My default is to score 2 at or above the at-exit median, 1 between the still-held median and the at-exit median, and 0 below it. When it lies: PitchBook tracked 147 exits to continuation funds in a single year, totalling $95.8 billion. A continuation fund is a new vehicle, run by the same sponsor, that buys a company out of its existing fund. A company that has just moved into one has had its hold clock reset, whatever its original entry date says.

Founder succession

Founder succession is the strongest actionability signal for companies without a sponsor. A founder who has run the company for decades, with no visible successor, has an ownership horizon whether or not they talk about it. The evidence lives in public profiles: the founder's tenure and career start, and whether anyone below them holds a president, COO or deputy title. When it lies: a successor is already running operations, or the next generation of the family is visibly in the business. In either case the owners have planned for continuity, not a sale.

Finding these people one company at a time is the slowest part of the screen. With Refolk you can ask for the whole segment at once and then match the results against your survivors.

Prior sale process

Prior-process history should count for more than hold age. Of sponsor-backed companies brought to market, half did not transact, largely because of valuation disconnects. A company that ran a process and failed has shown that its owners want to sell, and the reason it failed tells you roughly where the price gap was. A completed sale or a recent move to a continuation vehicle points the other way for the near term.

How do you rank the survivors and hand over the longlist?

Rank by multiplying fit (1 to 3) by actionability (0 to 8). Cut to the top 15 to 30, and move thesis-fit companies that can't be bought yet to a dated watch list instead of deleting them. The multiplication is deliberate: a perfect fit with zero actionability scores zero and drops out. Padding a list with names nobody can buy is the failure this screen exists to prevent.

ACTIONABILITY (0 to 8)
Ownership type     2 = founder/family majority confirmed in records
                   1 = sponsor- or VC-backed, or UNKNOWN
                   0 = corporate subsidiary with no carve-out signal
Sponsor hold age   2 = at or above your at-exit cutoff
                   1 = between your cutoffs, or not sponsor-owned
                   0 = below cutoff, or recent continuation transfer
Founder succession 2 = long-tenured founder, no visible successor
                   1 = unclear, or not founder-run
                   0 = successor in place or next generation running
Prior process      2 = prior process failed on price
                   1 = none known
                   0 = recent completed sale or continuation transfer

SCORE = Fit (1 to 3) x Actionability (0 to 8)

LONGLIST ROW
Rank | Company | Registry no. | Source channel | Pull date
Fit (1-3) | Fit rationale (one line)
Size | FACT / ESTIMATE | Size evidence | Headcount + date
Owner type | Owner evidence | Hold age | Succession signal | Prior process
Actionability (0-8) | Score | HSR / CMA flag | Open questions

Send four things to the review together:

• the ranked longlist
• the watch list
• the rejection log
• a one-paragraph error note giving the median and spread of your estimate-to-actual size ratio

Who else is screening the same targets?

The competing buyers for a private target are other corp dev teams and sponsors. Counting them tells you how contested the longlist will be. In Refolk's index of professional profiles, 787 US profiles carry the title Corporate Development, against 121 in the UK.

All columns come from Refolk's index. Hub share is derived from a 25-profile sample in each country.

The practical reading: the UK buyer-side set is small and two-thirds London-tagged in the sample, so you can list the competing corp dev teams by name for a UK longlist. In the US the set is larger and concentrated at big technology acquirers. A query such as "Heads of corporate development at US vertical SaaS companies that made bolt-on acquisitions" returns the people you will be bidding against. I left out a count restricted to senior titles because it returned an obvious filter artefact.

How the target screen goes wrong

Most failed screens fail in one of seven predictable ways. Each one produces a list that looks complete and isn't. Each has a cheap check, and the checks belong in the workflow, not in a post-mortem.

1. A universe built from codes only

A company that fits the thesis but filed under NAICS 81 never appears. What it looks like: an operator reads the longlist and asks where a well-known competitor went. Check: run the five-names test with two operators before cutting, and widen the channel that missed any name.

2. Headcount-to-revenue estimates read as facts

The estimate comes in too low for capital-intensive firms and too high for staffing firms, and SUSB receipts can be five years old. What it looks like: a company passes the size floor on an estimate and then shows half that revenue in its accounts. Check: cross-check against filed accounts, Form D offering size or a press-reported figure, and mark every unconfirmed row ESTIMATE.

3. Hold age used alone

A company held for six years looks like it is for sale. But half of sponsor-backed processes did not transact, and some exits go to continuation funds instead of buyers. What it looks like: a high-actionability row whose sponsor has just moved it into a continuation vehicle. Check: look for a prior failed process or a continuation transfer before scoring hold age.

4. Misreading the PSC register

A sponsor holding exactly 25% or less does not appear, and a "no PSC" statement can hide a fragmented cap table. What it looks like: a company graded founder-owned that turns out to have a minority sponsor with consent rights. Check: read the shareholder list in the confirmation statement, not just the PSC page.

5. Reading a missing Form D as bootstrapped

Failing to file is common, so a missing Form D proves nothing. What it looks like: a VC-backed company scored as independently owned. Check: score ownership on missing evidence at no more than 1, and mark it UNKNOWN.

6. Treating the HSR threshold as a hard cut

Value is measured at the ultimate parent entity level and includes assets, so a deal near the line can cross it. What it looks like: a strong target cut at $125M of estimated value that counsel would have cleared or filed. Check: flag rows within 20% of $133.9M and keep them in play.

7. No rejection log

Without one, cut names come back every quarter and the team re-argues them from scratch. What it looks like: a review spending twenty minutes on a company that failed the geography filter last time. Check: confirm that every removed row has a dated reason code and the filter version it failed.

A longlist padded with companies nobody can buy is a thesis map, not a target screen.

Keeping the screen current between strategy reviews

A screen goes stale in three predictable ways, and each has a scheduled fix:

• Ownership changes. Re-run the ownership and actionability steps on the longlist and watch list every quarter.
• Thresholds move. The FTC revises the HSR thresholds every year. Re-derive the EBITDA equivalent from the current threshold and the current median multiple.
• SUSB publishes a receipts year. Receipts appear for years ending in 2 and 7. When a new receipts year comes out, refresh your revenue-per-employee ratios and re-measure your estimate error.

When the thesis changes, bump the filter version. Then re-test only the companies rejected under criteria that changed. Leave the rest of the log alone.

Watch-list triggers should be concrete and dated. Examples: a sponsor reaching your at-exit hold cutoff, a founder succession change appearing on public profiles, a new Form D, or a confirmation statement showing a shareholder change. A watch list without triggers is just a second longlist nobody reads.