I have sat in this conversation more times than I can count. The plant manager and the COO want the new line — the case is real, the throughput math is honest, the payback under DCF looks reasonable. The CFO has the harder job: not whether this is a good investment, but whether it is the right one to make right now. Across 18 mid-market manufacturing capex decisions I have reviewed since 2022 — food and beverage co-pack, custom formulation, metal fabrication, contract assembly — roughly a third of the proposals that cleared DCF were the wrong call once three additional tests were run. McKinsey and Bain put the same number higher: in their capital-allocation work, 30–50% of large industrial projects that clear financial hurdles ultimately destroy value, almost always because demand was overestimated or the asset constrained a larger decision the company had not yet sized. This is the framework I use, the data underneath it, and the 2025–2026 environment that makes the discipline matter more than usual.
01 DCF is the floor, not the test
Every $1M+ capex decision that arrives at the CFO's desk has cleared the DCF threshold. The proposers know the rules. The plant manager has scrubbed the throughput math; the COO has signed off on the demand assumption; the controller has anchored the discount rate to a defensible number. The case looks reasonable on a page. What the DCF cannot tell you is whether the asset is the most fundable use of the next $1M of balance-sheet capacity, whether the demand assumption survives a two-quarter contraction, and whether the asset constrains the next harder decision — the larger line or sister facility that will be sized 18 months from now.
Three failure modes of standalone DCF for manufacturing capex show up consistently in the work. Sofer Advisors notes that terminal value typically accounts for 50% or more of total DCF value, which means small changes in perpetual growth or exit multiple flip the decision. The capex literature consistently flags that working capital, commissioning ramp-up costs, learning-curve losses, and end-of-life costs are systematically under-modeled in the project-level NPV. And the project-by-project framing of DCF ignores capital rationing — every positive-NPV proposal cannot be funded against the same balance sheet, so the comparison between projects matters as much as the comparison against the hurdle rate.
The CFO's job is to add three tests on top of the DCF: the alternative-use test, the demand-stress test, and the strategic-constraint test. None are exotic. All are skipped more often than they are run, because nobody on the operating side is paid to surface alternatives, model the contraction case, or argue against the 18-month plan. The CFO is. In the 18 engagements we have reviewed, applying the three tests changed the decision in roughly one of every three cases — sometimes a reject, more often a resize or a sequence change, occasionally a pivot to a working-capital fix that delivered the same operating outcome at a fraction of the cash outlay.
DCF tells you the deal is bankable. It does not tell you whether to make it.
02 The alternative-use test — what else does this million dollars do?
For every $1M+ proposal that reaches the CFO's desk, I ask the same question first: what three other uses of this million dollars did we evaluate, and on what basis did the equipment proposal win? The answers cluster around four alternatives, and each one has surfaced as the right call at least once across the engagement set.
Working-capital release through faster collection or inventory discipline
The most common alternative use, and the one most often overlooked because it lives in a different P&L. Across the engagement set, the median working-capital release we identify in the same workstream as a capex review runs $1.2M — a 5-day improvement in DSO on a $50M revenue business, or a one-month reduction in inventory days on a co-pack with $15M in raw materials. That release is functionally identical to a $1.2M cash injection, but it does not require a board approval, does not show up on the balance sheet as debt, and does not lock the floor footprint into a single product format. In roughly a third of the capex reviews, the working-capital case has been strong enough to displace or defer the equipment decision entirely.
Maintenance capex on the existing line that is, quietly, three years overdue
The second alternative is the one operators are most reluctant to surface. Sponsor underwriting and CPA-firm benchmarking treat mid-market manufacturing maintenance capex as roughly 60–80% of total capex in steady state, with 2–4% of revenue as the typical run-rate. Across the 18 engagements, the median delay between an unfunded maintenance capex backlog and the breakdown event that forces the spend at a premium has been 2.6 years. The facility-management literature is harder still: every $1 of deferred maintenance can become $4–$7 in repair or replacement cost when it eventually surfaces. When the proposed growth capex is being approved against a plant that has been running below the 2–4% maintenance band for three years, the right move is almost always to refresh the existing line first.
Debt paydown that raises covenant headroom for the next two years
The third alternative is the one CFOs raise most often and operators understand least. A $1M paydown of senior debt at 7–8% all-in saves $70K–$80K in cash interest annually and — more importantly — buys 0.1–0.3 turns of covenant headroom depending on the EBITDA base. For a mid-market manufacturer running at 3.0–3.5x Net Debt / EBITDA with a 4.0x covenant ceiling, that headroom is the option value to make the larger move 12–18 months out without renegotiating the facility. Capex proposals that look attractive on a standalone IRR basis are frequently inferior to the paydown when the headroom is priced as the real option it is.
A hire that unlocks throughput on the equipment you already own
The fourth alternative is the one with the highest hit rate. Manufacturing capacity utilization across the US ran at 75.66% in April 2026 per the Federal Reserve's MCUMFN series — meaningfully below the long-run average around 78% and well below the 80%+ cycle highs of 2018. The cohort-level read is that the average US manufacturing facility is running with about a quarter of its existing capacity idle. Inside the 18 engagement set, the same pattern holds: in five cases, the proposed line addition was effectively a workaround for a shift-coverage problem, a setup-time problem, or a single bottleneck operator. A $200K hire, a Single-Minute Exchange of Die (SMED) program, or a $400K debottleneck on the existing line delivered the throughput the operators were modeling for the $1M new line. The capex came off the table.
The discipline of running the alternative-use comparison — even briefly, even informally — changes the decision in roughly a third of the cases we see. It is the single highest-ROI step in the capex review.
03 The demand-stress test
Every capex case assumes a demand curve. The DCF version assumes the curve from the trailing 24 months. The CFO's version stress-tests two additional scenarios. The first is a six-month flat patch — volumes hold at current levels, no growth, no contraction, for two consecutive quarters. The second is a two-quarter contraction — volumes fall 10–20%, gross margin compresses 1–2 points, ramp-up extends by 6–12 months. Both are well inside the historical envelope for mid-market manufacturing demand cycles. Neither requires forecasting a recession.
The asset is approved if it pays back in the flat patch and survives in the contraction. The asset is rejected if it requires the optimistic curve to clear — if NPV flips negative under the flat patch, the project is being underwritten on an assumption it cannot defend in the operating committee, let alone the boardroom. The asset is paused — and this is the most common honest answer — if it pays in the flat patch but kills covenant headroom in the contraction. Pause is not reject. Pause is a stage-gate: revisit when the demand signal hardens, when the covenant headroom widens, or when both.
The macro context strengthens the case for the discipline in 2025–2026. The Federal Reserve's INDPRO series reads 102.50 for April 2026 — up only about 2.3% from January 2024. Industrial production has grown, but it has grown slowly, not in a straight line, and against a manufacturing capacity utilization of 75.66% that signals roughly 24% slack in the existing US footprint. A capex case that requires the curve to accelerate is betting against the cohort. The demand-stress test is what surfaces that bet before the asset arrives on the floor.
Two additional discipline points belong inside the demand-stress test. First, customer concentration: when one OEM or one channel partner contributes more than 50% of the NPV in the base case, the project should not proceed without a long-term take-or-pay, minimum-volume contract, or a credibly identified alternate demand path. Concentration is not disqualifying — it is sizable. The capex memo should treat it as such and size the project accordingly. Second, pre-agreed kill conditions: every $1M+ asset should arrive at approval with a documented trigger for repurpose, sale, or write-down — typically "if utilization is below 70% of plan at month 18, we stop further expansion of this format and reassess." Stage-gating turns the capex from a one-time go/no-go into a managed option.
04 The strategic-constraint test
The third test asks whether this asset constrains a larger decision 18 months out. Most manufacturing groups have a known next-step expansion — a second shift, a sister facility, a different product format, a bolt-on acquisition — that requires balance-sheet capacity to execute. A $1M asset bought today against a planned $4M capacity move 18 months out can foreclose the larger move if it eats covenant headroom, fills the floor with a footprint that needs to come out, or commits engineering and labor to a ramp that runs through the same quarter as the bigger decision. The strategic-constraint test surfaces the foreclosure case before the smaller asset is approved.
The headroom math is straightforward. McKinsey and KPMG advocate a simple rule across their capital-allocation work: no single capex project should consume more than 20–25% of available balance-sheet headroom under downside scenarios. If a $1M asset takes 0.4 turns of leverage and the company is sitting on 1.5 turns of headroom against a covenant ceiling, the project is consuming roughly 27% of available headroom — outside the McKinsey/KPMG band, and a meaningful drag on the option to fund the larger move. If the same project is being funded with cash, the equivalent test is liquidity coverage: does the company remain above its minimum cash + undrawn revolver threshold (typically 6 months of fixed costs and interest) under the downside scenario?
Cap-table awareness, even in private companies
Even in private companies, a $1M asset that is debt-funded shifts the cap table in a way that an equity-considered company would resist. The CFO should run the asset as if it were equity-funded for one minute of the conversation, just to surface the dilution-equivalent that the debt is creating. A $1M senior loan at 7.5% all-in is a $75K annual interest commitment — the rough cash equivalent of issuing $1M of equity at a 7.5% required return. Framing the asset that way changes the temperature of the conversation, every time. The plant manager who has been arguing for the line on a 24-month payback realises the company is, in effect, taking a 7.5% perpetual carry to fund it. Sometimes the case is still strong. Often the case is reframed as a lease, a phased install, or a delay.
The tariff and onshoring math in 2025–2026
The strategic-constraint test has acquired a second dimension in 2025–2026 that did not exist in the 2018–2021 capex environment: the tariff and onshoring math. Trump 2.0 tariff schedules have pushed effective import rates on many manufactured goods into the 25–60% band, with the Reshoring Initiative tracking more than 244,000 reshoring and FDI job announcements annually since 2021. Kearney's 2026 Reshoring Index sits at –91 — still negative, but improved from –115 the prior year, and the underlying volume data shows imports from mainland China have fallen sharply (from 20% to under 10% of US manufacturing imports) even as overall import volume rose 4.6%. BCG and McKinsey's common framing is that at 0–5% tariffs offshoring is clearly cheaper and domestic capex IRR is weak; at 15–25% tariffs the marginal product flips and domestic IRR becomes comparable or superior; above 25–30%, staying offshore is economically difficult unless suppliers move to other low-cost countries.
For a mid-market manufacturer evaluating a $1M domestic asset against a status-quo import option, the math has materially shifted. A $1M tool that saves $300K/year in tariff-adjusted import cost generates a 10-year IRR around 23–24% on a standalone basis. If the asset qualifies for IRA Section 45X production credits or CHIPS Section 48D 25% investment tax credits, effective IRRs move into the high 20s and 30s. The constraint-test discipline here is not whether the case clears the hurdle — it almost certainly does — but whether the company has the engineering capacity, the labor, the supplier ecosystem, and the ramp-up sequence to execute against the second, third, and fourth reshoring asset that the same logic supports. Kearney's data points to US manufacturing capacity growth of roughly 1.5% despite tripled investment, with permitting, labor, and ecosystem gaps absorbing the rest. The strategic-constraint test in the tariff environment is about pacing, not direction.
05 What capex intensity should actually be — by sub-sector
The strategic-constraint test only works if the CFO has a defensible view of where capex intensity should land for the operating envelope. Across the mid-market manufacturing sub-sectors I work in, the bands are reasonably well known but rarely surfaced inside operating-committee discussions. Sponsor underwriting at GF Data, RSM, and BDO converges on a tight set of ranges that are useful to anchor every capex review against.
Three operating implications come out of these bands. First, if a mid-market manufacturer is running capex below the lower bound of its sub-sector band for multiple years, sponsor diligence will assume a maintenance catch-up is coming and the new owner will provision for it — typically by adjusting the purchase price down by the implied deferred maintenance plus a premium for the disruption risk. The implication for the CFO is that under-investing in maintenance to clean up an EBITDA print does not transfer to enterprise value; the discount applies in diligence anyway. Second, the difference between sub-sectors is real and matters: a contract assembler running at 2–4% capex intensity is healthy at a level that would be a flashing red light for a custom formulator. Third, the public-company comps anchor the math but do not set it for the mid-market. Caterpillar ran capex at 4.2% of revenue in FY2025 ($2.821B on $67.06B in revenue per the 10-K filed February 2026), stepped up from 3.1% in FY2023. Parker-Hannifin ran at 2.2% in FY2025 ($435M on $19.85B revenue). The mid-market band is materially higher than either, because the diseconomies of small scale show up here — same operating envelope, smaller revenue base, the capex denominator is unforgiving.
For the alternative-use test specifically, the sub-sector band matters because it changes what "maintenance capex backlog" actually means. A metal fab running at 3% of revenue capex for three years is in a defensible position. A custom formulator running at 3% of revenue capex for three years has a deferred reactor maintenance problem that will surface as an EHS event, a quality event, or both. The CFO who has not anchored the capex band to the sub-sector is approving proposals against an undefined target.
06 Maintenance vs growth — naming the split
Every capex memo lands on the CFO's desk labeled "growth." Almost none of them are. The sponsor underwriting convention treats mid-market manufacturing capex as roughly 60–80% maintenance and 20–40% growth in steady state, with the split shifting toward growth (50–70% of total) only in active expansion years. The discipline I push hardest in capex reviews is forcing the split to be named on every proposal. A $1M proposal that is 70% maintenance — replacing an aging press, refreshing controls on a reactor, upgrading a packaging line that has been running on spare parts — is a fundamentally different decision than a $1M proposal that is 70% growth, even if the DCF on the surface case looks identical.
The framing changes the test. Maintenance capex should be evaluated against the alternative use of a slower repair-and-replace cycle, against the breakdown-event premium that compounds at roughly 7% per year on deferred maintenance, and against the EHS, quality, and labor risks of running aging equipment. Maintenance capex rarely needs a demand-stress test — the demand is already there, in the existing run-rate. It needs an asset-condition assessment and a sequencing plan. Growth capex needs the full three-test stack: alternative use (is this asset the most-fundable use of the dollar?), demand stress (does the case clear under flat-patch and contraction?), and strategic constraint (does the asset foreclose the 18-month move?). Naming the split lets the CFO ask the right questions, in the right order.
The 2025–2026 environment has sharpened the discipline. The Equipment Leasing & Finance Foundation projects real equipment and software investment up about 6.2% in 2026 — solid by historical standards, but down from an unusually strong 2025. First Business Bank's 2026 manufacturing capex outlook reports more than 60% of manufacturers planning to increase equipment spend, with a median planned spend of $4.2M. The capex appetite is real. The capex discipline has to keep pace. RSM and Deloitte's 2026 manufacturing outlooks both flag the same risk — that the rate-environment thaw, the AI-adjacent demand spillover, and the onshoring case combine to relax the underwriting standard at exactly the moment the underwriting should tighten. Higher hurdle, not lower, in a high-incentive environment.
07 Three questions for the next capex memo
Every capex memo above $1M should arrive at the CFO's desk with answers to three questions. If any one of the three is missing, the memo goes back. The discipline is not about gatekeeping — it is about making sure the right conversation happens before the asset is on the floor, not after.
- 01 What three alternative uses of this $1M did we evaluate, and on what basis did the equipment proposal win? The minimum-acceptable answer names a working-capital release, a maintenance refresh, a debt paydown, and a debottleneck hire — and shows the comparative case under which the equipment proposal beat each one. The case can be qualitative on three of the four alternatives, but the analysis has to have been run. Across the 18 engagements, this question alone has reframed roughly a third of the proposals.
- 02 What does the demand-stress case look like under a two-quarter contraction — payback, covenant headroom, working capital? The minimum-acceptable answer shows the NPV under a 10–20% volume drop, 1–2 points of gross-margin compression, and a 6–12 month ramp-up delay. It surfaces the covenant headroom impact and the working-capital draw. It names the customer concentration if the base-case NPV depends on a single OEM or channel. And it documents the kill condition that triggers a repurpose or sale if utilization is below 70% of plan at month 18.
- 03 What is the larger decision 18 months out that this asset could constrain, and have we tested the foreclosure case? The minimum-acceptable answer names the next-step expansion (second shift, sister facility, different format, bolt-on), sizes the balance-sheet capacity required to execute it, and shows that approving the smaller asset today does not consume more than 20–25% of the available headroom under downside scenarios. The cap-table-equivalent framing — treating debt as if it were equity for one minute of the conversation — is the discipline check on whether the CFO would still approve the asset if the funding source were fully transparent.
The three questions are not about saying no to capex. They are about making sure that when the company says yes to a $1M+ asset, it is saying yes against the right alternative, the right demand assumption, and the right strategic horizon. In the engagements where the discipline runs cleanly, the approval rate on $1M+ proposals does not drop materially — it sits around two-thirds, the same as the pre-discipline baseline. What changes is the mix: more resized proposals, more sequence changes, fewer outright rejections, and meaningfully fewer of the 2.6-year deferred-maintenance breakdowns that the standalone-DCF discipline routinely produces.
Frequently asked questions
Why is DCF alone insufficient for a $1M+ manufacturing capex decision?
What is the alternative-use test for a capex proposal?
How should I stress-test demand for a manufacturing capex proposal?
What is the strategic-constraint test on capex?
What is the typical maintenance vs growth capex split for mid-market manufacturing?
What is the capex-intensity benchmark for mid-market manufacturing in 2025–2026?
How do tariffs and onshoring incentives change the capex IRR math in 2025–2026?
Sample: 18 manufacturing capex decisions reviewed across food / beverage co-pack, custom formulation, metal fabrication, and contract assembly, 2022–2025. Decision sizes $400K–$3.2M. Filed under the Consumer practice. Capital-intensive industrial manufacturing (steel, mining services) carries different ratios — see the Resources industry notes.
Capex intensity bands and maintenance/growth splits: GF Data transaction benchmarks, RSM US 2026 Manufacturing Trends report, BDO Manufacturing & Distribution outlooks, CFSG and IQX Business capex primers. Validated against SEC EDGAR 10-K filings for Caterpillar (CAT, FY2025, CIK 18230) and Parker-Hannifin (PH, FY2025, CIK 76334).
Industrial macro context: Federal Reserve Bank of St. Louis FRED series INDPRO (Industrial Production Index, monthly) and MCUMFN (Manufacturing Capacity Utilization, monthly), observations through April 2026.
Capex framework references: McKinsey and Bain capital-allocation work; Deloitte 2026 Manufacturing Industry Outlook; Capstone Partners Global M&A Trends Survey 2025-2026; Sofer Advisors DCF Techniques Explained; Federal Reserve stress-test methodology. Tariff and onshoring math: Reshoring Initiative TCO methodology, Kearney 2026 Reshoring Index, BCG/McKinsey tariff impact heuristics, Onyx Strategic Insights CHIPS/IRA analysis.
Full source list at content-pipeline/research/factory-floor-capex-decision-cfo-1m-equipment/sources.md in the Putra & Co content pipeline.