Analyzing labor and key employees

How do you test labor and key employees before you rely on the roster?

Buying a manufacturing business is an operating transfer. Outsiders feel this first. Operators who already know the process still lose deals when they skip I-9s, OSHA logs, and the CBA.

The bench versus the census

Census CBP 2023 counted 284,452 manufacturing establishments; 68.4% have fewer than 20 employees (Source: Census CBP). On those files the bench is the deal. A 14-person shop with one programmer is a constraint. Deloitte and The Manufacturing Institute estimated as many as 3.8 million manufacturing employees needed from 2024–2033, with 1.9 million potentially unfilled if the skills gap holds (Source: Deloitte/MI). NAM's 2024 Q1 outlook, cited in the same release, had 65% of manufacturers naming talent as their primary challenge.

BLS OEWS May 2025 national median annual wages: machinists $58,750; welders, cutters, solderers, and brazers $53,750; CNC tool programmers $68,120; CNC operators $50,690 (Source: BLS OEWS). Night shift, certified work, and scarce metros pay more. If the seller's add-back assumes cousin rates, rebuild the SDE. A quality of earnings should test replacement cost, not repeat the brochure.

ISM's August 2026 Employment Index printed 51.2 with PMI 54.6 (Source: ISM). Climate is not your roster.

Priced items that are not culture

I-9 files. ICE audit exposure is a diligence request. Sample the files. Ask how they are stored. This is not a vibe check.

OSHA 300 logs. Repeat citations and process-safety issues (if applicable) are spend. A pretty safety poster is not a log.

CBA and multiemployer withdrawal liability. A buyer can be a successor obligated to bargain; assuming the CBA is a separate decision. Withdrawal liability is a priced item, not a surprise at the pension-estimate letter.

Federal WARN (29 U.S.C. § 2101): employers with 100+ full-time employees; generally 60 days before a plant closing (50+ employment losses) or a mass layoff (500+, or 50–499 if at least 33% at the site). Mini-WARN statutes exist in several states with different headcounts and days. Confirm with counsel before you plan a "reorganization" in month one. The first 90 days are an operating plan, not a layoff brochure.

Immigration and deemed export. Foreign employees on controlled technical data are an EAR/ITAR issue, not only an HR issue.

Stay math, MBOs, and the first quarter

Retention is cash, title, and a role that survives the first argument about quoting rules. Vest stays over 12–24 months. If the seller's spouse runs AP and will not remain, budget a controller. If the lead estimator is 68 and "open to helping," write the hours.

Management buyouts already have the bench and often lack the equity. The inverse file — you have equity and no bench — is the outsider problem. Do not mix the two theses.

SBA 7(a) maximum is $5 million. SOP 50 10 8 wants a 10% injection (Source: SBA). Lenders will still ask who the operator is. A personal guarantee does not replace a plant manager.

Stanford's 2024–25 median target had about 30 employees on a ~25% margin, ~$2.5 million EBITDA company (Source: Stanford GSB). That is a search-fund median, never a job-shop multiple, and a reminder that study companies have benches. A 12-hour owner-estimator shop is not that median.

The working-capital peg is not a labor document. The people who can finish the WIP are. If the only person who can close a traveler is leaving, you bought incomplete jobs at full value.

Plant-visit protocol: no interviewing the second-shift lead without the seller's script. You are confirming the cell exists, not announcing a sale to the floor.

Compensation, temp labor, and the recast you inherited

Ask for the last twelve months of overtime, temp-agency spend, and contract programmers. If the recast treated overtime as "one-time catch-up" every quarter, put it back. If temps are 15% of direct labor, you do not have a bench — you have a vendor. That can be fine. It is not "lean." It is a cost that will follow the work.

Piece-rate and gainshare plans change behavior at close. If the seller ran an informal bonus out of petty cash, your first payroll will start an argument. Write the plan or end it on purpose. Do not discover it when someone walks.

Supervisors who are working leads — common under Census CBP's 68.4% small-establishment share (Source: Census CBP) — are not "excess G&A" you can cut. They are the second shift. A QoE that deletes them as add-backs is selling you a hole. SBA 7(a) DSCR inside a $5 million cap still wants those people to exist (Source: SBA).

Non-competes and customer non-solicits for key employees are counsel items and state-specific. Do not copy a form from another state. Do not assume a programmer's non-compete survives. Retention cash often works better than paper you cannot enforce.

Training records, special processes, and the person who is the approval

Weld procedures, heat-treat recipes, clean-room gowning, and allergen training are quality-system evidence and labor evidence. If the only qualified welder is leaving, the procedure is a PDF. NADCAP and customer approvals can be person-specific. Budget requalification time in the first 90 days.

Apprenticeship and tuition deals can be assets if people stay, or liabilities if they leave with a clawback fight. Read the agreements. Do not advertise a "training culture" without the files.

Immigration status and visa expiry belong on a private schedule, handled by counsel, not on a whiteboard in the break room. Deemed-export risk on foreign employees who see controlled drawings is an EAR/ITAR issue. ITAR § 122.4(b) is about foreign-person control of a registrant — a different clock — but the floor-level deemed-export issue is live on day one (Source: DDTC).

Stanford's ~30-employee / ~25% margin median (Source: Stanford GSB) is a search-fund profile with a bench. It is never a job-shop multiple. If you are buying a 12-person owner-estimator shop, your labor thesis is the deal. IBBA's 83% three-offer statistic (Source: IBBA) will not hire your programmer. Fed utilization 76.0% (Source: Fed G.17) will not either.

A labor model line that belongs in every LOI packet

Replacement GM. Stay pool. Programmer hire or stay. Controller if the spouse leaves. Temp-to-hire conversion if temps are structural. Workers-comp class re-rate placeholder. WARN counsel if you already plan a reduction — federal 100+ FT and 60 days (29 U.S.C. § 2101). That packet is not culture. It is the investment.

SBA 7(a) DSCR inside $5 million (Source: SBA) uses this packet after the QoE. Combined 7(a)+504 can reach $10 million (Source: SBA). Same packet. Main Street 2.75x SDE (Source: BizBuySell via Iconic) feels each hire as a turn. Stanford $16 million / 6.2x / ~25% margin / ~30 employees (Source: Stanford GSB) is a bench profile, never a job-shop multiple.

Deloitte/MI 3.8 million / 1.9 million (Source: Deloitte/MI) is why the packet is not optional. BLS OEWS May 2025: machinist $58,750; welder $53,750; CNC programmer $68,120 (Source: BLS OEWS). IBBA's 83% three-offer statistic (Source: IBBA) will not staff the cell. ISM Employment 51.2 (Source: ISM) is climate. Fed utilization 76.0% (Source: Fed G.17) is climate. Your roster is the deal.

Jason Taken is a business broker, not an employment lawyer. Educational information only. Confirm WARN, successorship, and immigration with counsel. No outcome is guaranteed.

Questions the roster should answer before you model synergy

Who can quote if the owner is sick. Who can program if the lead is on vacation. Who runs nights. Who holds the top three accounts. What the OSHA 300 log actually shows. Whether I-9s would survive a desk audit. Whether a CBA and a withdrawal estimate exist. Whether anyone is a person-specific special-process approval.

Synergy speeches that skip those questions are not a model. They are a hope that labor markets will gift you a bench. Fund stays. Price replacement wages. Put temps that are structural back into the run-rate. If the spouse runs AP and is done on closing day, budget a controller before you exclusive.

A 14-person shop does not have excess G&A you can cut because a CIM called a working lead "supervisory." That person is the second shift. Delete them in a recast and you bought a hole.

HedgeStone will ask for the four names on the first call. If two of them are the seller, the stay plan is the deal, not a later workstream. Cross-training matrices and stay cash belong in the LOI packet, not in a culture slide you show after close.

Successorship, withdrawal, and the labor schedule nobody wants to open

If a CBA exists, ask counsel for the successorship clause and a withdrawal-liability estimate before you exclusive. A multiemployer plan you did not model is not a culture issue. It is a balance-sheet hole that can exceed a year of SDE on a small shop. The same packet should flag work-authorization expiry dates and any person-specific special-process approval. Those are not HR niceties. They are how a cell stops shipping in month two. Put the four names — who quotes, who programs, who runs nights, who holds the top three accounts — on the first page of the labor model. If two of those names are the seller, the stay cash is the deal. A 25% margin study cell still has to staff the cell. The study median does not hire, and a CIM org chart that calls a working lead supervisory is not a reason to delete second shift.