What has to be true before a second plant becomes a platform?
Buying a manufacturing business the first time is an operating transfer. The second time is a systems transfer. First 90 days on plant one are the entrance exam.
Readiness tests before you bid on plant two
- Concentration and job-cost extracts on demand, not as a science project
- A plant lead who is not you
- One quality system you can explain to a CB
- A peg you have already true-up'd once without a fistfight
- Debt capacity that is not a folklore 7(a)
Practice, not a law: many published add-on theses start around $1–3 million EBITDA. New platforms often want $3–5 million+ before a first-time platform is interesting. Strategics can go smaller if the capability is scarce. Do not name firms. Seller-side add-on pricing is how the other side thinks. Rollover and the second bite is how sellers hear the same story.
| Platform ingredient | Ready | Not ready |
|---|---|---|
| Ops bench | Plant one runs without you | You are still the estimator |
| Systems | Shared job-cost extract | Two whiteboards, two religions |
| Certs | Compatible gates or a funded integration | AS9100 plus a lapsed ISO and a speech |
| Customers | Overlap you have consent language for | Two 35% heroes who compete |
| Capital | Named stack above or beside SBA | "We'll 7(a) the second one somehow" |
| Culture | One quoting rulebook | Tribal rules in both buildings |
IBBA Q1 2026: 83% of deals over $5 million attracted three or more offers (Source: IBBA). Your second bite will be competed. A buyer who still cannot produce a schedule from plant one looks like a tourist.
Multiples, SBA, and the temptation to prepay the roll-up
On this site, EBITDA $3–10M general fab sits at 5.0x–6.5x; precision / plastics 5.5x–7.5x; AS9100 / ISO 13485 6.5x–9.0x — advisor-survey and reported-deal ranges, not an appraisal. GF Data Q1 2026 manufacturing recap averaged 7.2x, 5.9x under $25 million EV, 8.8x at $100–250 million EV (Source: Iconic GF Data recap). Larger-EV series stay in their lane. Do not pay a $100 million-EV cell for a second $2 million-EBITDA shop because you wrote "platform" on a slide.
SBA 7(a) maximum remains $5 million with a 10% injection under SOP 50 10 8 (Source: SBA). Affiliation can block a casual second 7(a). Combined 7(a)+504 can reach $10 million of SBA-backed financing under the 2026 notice for borrowers who fit both programs (Source: SBA). That is still not a license to stack folklore. Confirm with the lender.
A quality of earnings on plant two should test the same job-cost versus tax questions, plus any "synergy" you plan to put in the model. Synergies you have not demonstrated are not earnings.
Labor and search-fund gravity
Deloitte/MI: as many as 3.8 million manufacturing employees needed 2024–2033, 1.9 million potentially unfilled (Source: Deloitte/MI). Two thin benches do not make a deep one. Cross-training on plant one is a prerequisite, not a slide.
Searchers sometimes buy a second company. Stanford's 2024–25 median purchase of $16 million at about 6.2x EBITDA on a ~25% margin company (Source: Stanford GSB) is the first-bite study profile. It is a search-fund median, never a job-shop multiple, and not a roll-up formula. If your board expects 25% blended margin after two 11% shops, rebuild the board packet.
Main Street SDE medians (2.75x on about $268k in BizBuySell 2021–2025) do not become LMM cells because you bought two of them (Source: BizBuySell via Iconic). Two wage-replacement shops are two jobs.
Integration that is not a slide
The second close is the easy day. The hard months are quoting rules, item masters, and quality manuals that do not match. Pick one job-cost religion and convert the other plant on a calendar, not "eventually." If plant two's ERP cannot export won-job versus actual, you bought another whiteboard. Budget the reconstruction the way you would budget a QoE.
Certificates do not merge because you say "platform." A CB will want to know which legal entity holds which certificate, which sites are in scope, and whether special processes moved. Asset-purchase number two may need a new listing. Stock-purchase number two may keep a registration you do not understand. Confirm with the registrar before you announce a combined capability to customers.
Customer overlap is a consent problem. Two plants shipping the same OEM under different quality agreements can create a conflict the OEM will price. Do not send a joint press release. Do the commercial work. FAR 42.12 still applies if either book has government contracts. ITAR registration still needs amendments. Foreign-person control still needs 60-day DDTC notice under § 122.4(b) (Source: DDTC).
Working capital on plant two needs its own peg, not a copy of plant one's average. Seasonality, consignment, and POC methods differ. The peg page is still the method. Copy-paste is how you buy rust twice.
Capital, culture, and the second-bite fantasy
Cash at close on plant two still follows IBBA's majority-cash pattern (Source: IBBA). Notes and earn-outs still appear when concentration or a certificate is the open item. Do not fund plant two with a revolver that assumed plant one's unused availability. Field exams travel. So do cross-default clauses.
Culture is quoting ethics and safety, not a pizza. If plant one runs scientific molding and plant two runs "good enough," the combined margin will be the worse one until you pick. Labor markets do not care that you have two logos. Deloitte/MI's 3.8 million / 1.9 million figures still apply (Source: Deloitte/MI). BLS OEWS May 2025 still puts machinists near $58,750 nationally (Source: BLS OEWS). Two thin benches compete with each other for the same hires.
The second bite — a later sale of the combined company — is a thesis, not a multiple you prepay. Rollover is how sellers hear it. You should hear the inverse: if you overpay for plant two to "set up" a sale, you have already spent the second bite. Main Street 2.75x SDE shops (Source: BizBuySell via Iconic) do not become a GF Data 7.2x average because there are two of them.
Fed utilization 76.0% (Source: Fed G.17) and ISM PMI 54.6 (Source: ISM) describe climate for both plants. They do not integrate item masters. Do the systems work or do not buy the second file.
A sequence that respects both plants
Year one: close plant one, finish the 90-day list, true-up the peg, keep the certificate live, and prove you can extract concentration without the seller. Year two: write a buybox for plant two that names process overlap, geography you can staff, and a capital path that is not a leftover 7(a) daydream. Then hunt. IBBA's 83% three-offer statistic on deals over $5 million (Source: IBBA) will still be there.
Do not "build pipeline" for plant two while plant one's programmer is interviewing. Deloitte/MI's 3.8 million / 1.9 million figures (Source: Deloitte/MI) make that arrogance expensive. BLS OEWS May 2025 welder median $53,750 (Source: BLS OEWS) is not a rounding error when you need two weld cells.
Quality systems on a sequence: get plant one's next CB audit done before you distract the quality manager with a second manual. Customers notice missed audits. Access premiums in the 6.5x–9.0x EBITDA $3–10M aero/medical cell exist only if audits stay clean — advisor-survey ranges, not an appraisal.
Real estate on a sequence: two plants are two cap-rate decisions. 504 occupancy is per project. Combined SBA-backed financing can reach $10 million (Source: SBA). Affiliation can still block a casual second government-guaranteed stack. Confirm before you exclusive on plant two.
Main Street 2.75x SDE shops (Source: BizBuySell via Iconic) can be add-ons to a larger platform. They do not become the platform by themselves. Stanford's $16 million / 6.2x / ~25% margin median (Source: Stanford GSB) is a first-bite study profile, never a job-shop multiple, and not a reason to bolt a 9% fabricator onto a 25% model and call it synergy.
Jason Taken is a business broker, not your platform COO. Educational information only. No outcome is guaranteed. Stay-versus-add lives on single-plant vs multi-plant platform if plant one still cannot export concentration.
Questions plant two should not be allowed to dodge
Can plant one produce a concentration schedule today. Can plant one ship if you are in the hospital. Do the quality systems even rhyme. Is the capital path a leftover 7(a) daydream. Are the two hero customers competitors. If any answer is no, you are collecting shops.
You are not building a platform. Wait. Get plant one's next CB audit done before you distract the quality manager. Do not send a joint press release. Do the commercial consent work. Write a new peg for plant two. Copy-paste is how you buy rust twice.
Two thin benches compete with each other for the same hires. Cross-training on plant one is a prerequisite, not a slide. The second bite is a later sale. It is not a multiple you prepay on the first or second close.
HedgeStone will not run a second-plant process for a buyer whose first ERP still cannot export concentration. Finish plant one. Then write the next buybox. Two wage-replacement shops are two jobs. They do not become a platform because a slide said so.
Sequence the registrar before you sequence the LOI
Get plant one's next CB audit done before you hand the quality manager a second manual. Customers notice missed audits. Access premiums in the advisor-survey aero and medical cells exist only if those audits stay clean. They are not an appraisal. Two plants are two cap-rate decisions if you own dirt, and two occupancy tests if you still want 504. Affiliation can block a casual second government-guaranteed stack even when combined 7(a) and 504 can reach $10 million on paper. Confirm that before you exclusive on plant two. Main Street 2.75x SDE shops can be add-ons. They do not become a platform by themselves. Write a new peg for plant two. Copy-paste is how you buy rust twice. If plant one still cannot export a concentration schedule, you are not ready for a second teaser. The longer platform-through-acquisition guide is the buy-and-build sequence once plant one can run without you.