I have sat through four mid-market CPG in-house-transition decisions between 2020 and 2025 — two beverage, one supplement, one frozen-food — and three of the four arrived at the same place: the capex pencilled out in the deck, the gross-margin uplift looked clean, the strategic story sounded right, and somewhere between month nine and month twenty the brand discovered that QA and sanitation headcount needed to double, the ERP cutover ate two quarters of fill-rate, the SQF certification slipped twelve months, and the working-capital pickup was 22% of capex instead of the 10% modelled. The fourth was the supplement brand — the only one that did it well — and the reason it worked is the reason the other three struggled: they treated the decision as a margin question, and it is really a capital-displacement and overhead-absorption question first. The headline question is whether to bring production in-house. The working question is whether the brand can afford to tie up the capex, the working capital, and the management bandwidth an owned plant demands at this stage of the growth curve. Here is the framework we use when we advise on this today.
01 The wrong frame and the right frame
The wrong frame is the one founders open with most of the time: "Co-pack is taking five-to-fifteen percentage points of gross margin off my P&L. If I bring it in-house, I capture that margin." That statement is arithmetically true on a single hero SKU at high volume and full plant utilisation. It is rarely true on the blended portfolio inside the realistic eighteen-to-twenty-four-month window that a board will tolerate before asking when the EBITDA arrives. The frame quietly assumes that the in-house plant will run at 70-80% utilisation, that the overhead step-up is modest, that the working-capital pickup is small, and that the cash freed by the margin uplift exceeds the cash absorbed by the capex and the system build. Four assumptions, each of which gets stress-tested in the first twenty-four months. In our experience all four usually fail simultaneously.
The right frame starts with a different question: where in the capital allocation hierarchy does plant capex sit, against the alternative uses of the same dollar? Cascadia Capital and Auxo Capital both argue this directly in their CPG M&A commentary — for high-growth brands, the highest-return dollar is almost always demand-side (marketing, distribution, working-capital fuel for retailer growth) rather than vertical integration, until the growth rate naturally slows and the next EBITDA step-change has to come from structural margin. The corollary is that vertical integration is a late-game move for most mid-market consumer brands, not an early one — and when it is forced earlier, it is usually because the co-pack base is broken (capacity, quality, IP risk), not because the math says it pays. The diagnostic question is therefore: am I doing this because the math says it pays, or because the co-pack base is broken? The answers run different processes.
Vertical integration is a late-game move for most mid-market consumer brands. When it is forced earlier, it is usually because the co-pack base is broken — not because the math says it pays.
02 Capex commitment, honestly priced
The capex number is where most decks lose touch with reality. The published "$2M-$15M" range that gets quoted at industry conferences is real only for the lower tail of conversions — a single-line dry-snack or basic personal-care plant slotted into an existing food-grade shell with most utilities already in place. Once you specify the actual plant a mid-market brand needs in 2026 — GFSI-compliant, two-to-three lines, refrigeration where the category demands it, QA lab in the building — the all-in number moves materially. Food converted-warehouse builds run $6-15M; food greenfield $15-25M typical and $25-40M for the upper mid-market. Beverage is structurally more expensive: $11-20M for a converted cold-fill PET or can line, $20-40M for hot-fill or aseptic greenfield. Supplements run cheaper, $5-12M converted and $12-20M greenfield. Personal care and beauty fill-finish sit at $4-14M converted depending on format complexity.
Inside any of those totals the cost stack is roughly 35-50% building shell plus site plus core utilities; 30-45% processing and packaging equipment; 10-20% utility fit-out; and 10-15% soft costs (engineering, permits, validation, startup). The equipment line — the one that gets most of the attention in the business case — is rarely the line that surprises. A 100-150 cpm beverage filling line runs $1.5-2.0M; a 200-300 cpm line $2.5-3.5M; a 400-600 cpm line with depalletizer, pasteuriser, case packer and robotic palletiser $4-7M. Mixing tanks $60-350k depending on size; ribbon blenders $80-350k; HTST pasteurisers $200k-$1.5M; tunnel pasteurisers $750k-$2M. These numbers are knowable in advance. The numbers that surprise are the building-modification line (sanitary construction, drainage, refrigeration), the utility-upgrade line (electrical to 2-4 MW, compressed air, process water, fire protection) and the soft-cost line, which the deck always under-estimates by 30-50%.
And then there is maintenance capex, which the build deck almost never models honestly. Steady-state maintenance capex runs 5-7% of gross PP&E for food and beverage (mechanical, wet, lots of wear), 4-6% for supplements, 3-5% for personal care and beauty. On a $20M food plant with serious automation, that is $1.0-1.4M every year — before any expansion capex, before any line upgrades, before any of the inevitable retailer-driven packaging refreshes that pull capital into the plant. The plant is a permanent capex consumer; the deck usually models it as a one-time spend.
03 The capacity flexibility math nobody runs cleanly
Co-pack and owned-plant have structurally different cost shapes, and the cost-shape difference is the source of most of the strategic argument on either side. Co-pack is mostly variable: conversion fee of $0.15-0.75 per unit for food, $0.20-1.00 for beverage, plus minimum-run fees of $5-50k per run. You can typically 3-5x volume with the same co-packer inside 12-24 months by adding shifts and moving from campaign to regular slot. You can also scale down — at a cost. MOQs of 50-100k units per SKU per run, line-reservation retainers if you under-utilise contracted capacity, and at worst a renegotiated rate card that erodes the margin advantage you started with. The point: co-pack absorbs the demand volatility for you, and prices for it.
Owned plant inverts the cost shape. The fixed-cost stack — depreciation on a $20M asset over ten years, salaried plant staff (manager, QA/QC, maintenance, schedulers, supervisors), direct labour base at 1-1.5 shifts, utilities, maintenance, sanitation, insurance, certifications — runs $7.5-12M per year before a single dollar of materials moves through the building. That cost is there at 30% utilisation and it is there at 90% utilisation. The unit economics flip on absorption, not on theoretical efficiency. Run a stylised plant with $10M of annual fixed cost and 20M units of capacity: at 50% utilisation the fixed cost per unit is $1.00 (worse than the $0.80 co-pack baseline); at 75% utilisation it is $0.67 (better by $0.13 per unit, worth $2.0M of annual margin); at 90% utilisation it is $0.56 (saves $0.24, worth $4.3M). The economic argument for in-house is entirely an argument about utilisation. The economic risk is entirely an argument about realistic volume forecasting.
The seventy-percent-by-month-eighteen test
The single highest-leverage question on this decision is whether the realistic volume plan, after stress-testing for retailer-launch timing and DTC growth, allows the plant to operate at 70%-plus utilisation inside 18-24 months. If yes, the in-house economics work — typically a 13-17 percentage-point gross-margin improvement over co-pack at scale, translating to $0.40-0.50 per unit of incremental gross profit. If no, the in-house plant is a capital trap: a fixed-cost base larger than the variable-cost saving it is supposed to generate. The brands that get the answer wrong are the brands that anchor on a base-case forecast that was built before the rate-of-growth slowdown that always shows up in the second half of a fast-growth brand's curve.
For sellers building the SKU-by-SKU model that supports this question, the landed-cost teardown is the place to start — the work the SKU profitability teardown lays out at /blog/sku-profitability-dtc-landed-cost/ is the same operating discipline applied to the in-house decision. Without a clean view of contribution per SKU per channel today, the forecast that supports the plant build is a finger in the wind, and the plant build will be the largest capital decision the brand makes on the back of one.
04 The hybrid case that often wins
For most mid-market consumer brands, the right answer is not all-co-pack or all-in-house — it is hybrid. Core hero SKUs (the top 20% by volume, contributing 70-80% of revenue) move in-house once the volume threshold is crossed; the long tail — seasonal items, low-volume variants, channel-specific SKUs, innovation pipeline — stays with co-packers. The economics are cleaner than either pure model. Captive in-house GM on hero SKUs at 70%-plus utilisation runs 45-55%; co-pack GM on the long tail runs 35-45%. The blended portfolio outcome on a 70/30 core/tail split is typically 42-50% — call it 300-700 basis points better than all-co-pack and 200-400 basis points worse than fully-utilised all-in-house, with materially less capex, less working capital, less governance complexity and more capacity flexibility on the long tail.
The brands that have executed this well — Olipop's core flavours in-house with seasonal extensions co-packed, Vital Proteins' margin-critical core SKUs centralised while innovations stay external, Liquid Death's biggest pack formats brought in-house with newer formats outsourced — share a discipline that is harder than it looks. They run a real SKU-segmentation rule (top-20%-by-velocity in, everything else out), they enforce minimum-run discipline on the captive line, they hold a service-level priority that protects hero SKUs in capacity-conflict scenarios, and they refresh the segmentation quarterly as the SKU mix evolves. The brands that have struggled with hybrid — and there are more of these than the success stories — have run it as a portfolio-of-convenience rather than a portfolio-of-discipline, and the consequence is the worst of both worlds: a captive line that is under-utilised because the segmentation rule keeps slipping, and a co-pack base that is over-priced because the volume that was supposed to consolidate there got pulled in-house.
When hybrid breaks
Three failure modes recur. Capacity bleed — the captive line not full enough, depreciation and salaried labour spread thin, the "cheap" in-house unit becomes more expensive than the co-pack alternative; usually traceable to a hero SKU that lost velocity faster than the plan accounted for. Dual-overhead drag — the brand now runs two operating systems (internal plant management plus external vendor management), each with its own QA flow, forecast cadence, procurement load and scheduling friction; this is especially punishing for brands without a real S&OP and S&OE cadence and is the leading reason hybrid models fail to deliver the modelled overhead absorption. Governance complexity — the SKU-segmentation rule erodes, capacity-allocation decisions get made ad-hoc, MOQs slide, change-control protocols slip, and the portfolio drifts back toward an undifferentiated production base that costs more than either pure model.
05 The talent and system overhead the deck always misses
The single biggest pattern across the failed in-house transitions we have seen is overhead under-build. The capex deck names the line operators and the plant manager. The overhead reality is QA and food-safety headcount of 4-8 FTEs running $1.0-2.0M per year (3-5% of revenue for a brand under $50M); regulatory and certification capex of $250-750k unexpected in the first three years; supply-chain headcount of 2-3 dedicated roles costing $300-600k per year; and ERP plus MES systems running $800k-$3.1M over three years with a real go-live timeline of 12-18 months rather than the modelled 6-9. Aggregate manufacturing-related overhead at low utilisation runs 8-12% of revenue, stepping down toward 6-9% only after plant utilisation crosses 70%. That gap — the difference between modelled overhead at 5-7% and reality at 10-12% in the ramp years — is where the gross-margin gain gets eaten.
QA and food-safety under-build
The minimum viable food-safety org for a GFSI/SQF-aspiring plant is one QA/Food Safety Manager (PCQI and SQF Practitioner; $140-190k fully loaded), two QA supervisors plus three QA techs ($350-450k loaded), a sanitation manager plus leads plus crew ($300k-$1.0M depending on shifts) and an external lab budget that runs $50-300k per year depending on category risk. The deck usually budgets one QA manager and a handful of techs at half the realistic loaded cost, and the budget breaks the moment the brand pursues SQF or BRCGS certification — which it will, because Kroger, Costco and Target will require it. Audit fees plus documentation plus the inevitable hygienic-design fixes (segregation walls, drains, hand-wash stations, allergen storage, metal detection or X-ray upgrades) add another $150-500k of unplanned capex in the first 24 months.
ERP and MES — the line that breaks the timeline
The ERP and MES build is the line that most consistently breaks the timeline. NetSuite, Acumatica or D365 Business Central licences run $50-250k per year; implementation partner fees $500k-$1.5M for a real mid-market CPG configuration with multi-site, trade spend and lot traceability; internal backfill of two-to-four FTEs for 9-18 months that the deck rarely capitalises. The modelled go-live is 6-9 months; the real go-live to stabilised state is 12-18 months. The operational risk during cutover — BDO and RSM both flag this as a top operational risk in mid-market manufacturing transitions — costs 1-3% of revenue in lost sales, OTIF penalties and retailer fines during the cutover quarters. On a $50M brand, that is $500k-$1.5M of cost that the financial case never named. The same operational discipline that surfaces deduction recoveries in the trade-spend playbook at /blog/cpg-trade-spend-deduction-recovery-playbook/ applies here: the system cutover is a margin event, not just a project.
And then there is working capital, which the deck usually models at 10% of capex pickup and which actually runs 15-25%. On a $12M plant build, that is $1.8-3.0M of incremental inventory and spares — raw and packaging safety stock, work-in-process while you stabilise yield, finished-goods buffer while you learn run lengths, MRO stores you did not need when the co-packer carried them. This is real cash, and it sits on the balance sheet the day the line commissions.
06 The volume threshold below which co-pack wins
Pulling the capex, the absorption math, the hybrid economics and the overhead reality together: there is a knowable revenue threshold below which co-pack is the right answer for almost every mid-market consumer brand, and a knowable revenue threshold above which in-house starts to pay. The thresholds vary by category, but the bands are not wide.
- 01 Ambient and frozen food: Owning a plant typically pencils at $75-125M annual revenue on a single product platform, when co-packer conversion is consuming 10-15% of net sales and baseline volume is predictable enough to underwrite 70%-plus utilisation by month eighteen. Below $75M, co-pack is the right answer and the capital is better deployed into distribution and marketing. Above $125M with a clean utilisation forecast, the in-house economics are strong.
- 02 Beverage: The threshold is higher — typically $100-200M revenue — because co-pack conversion runs $2-4 per case all-in and the capex per unit of capacity is the most expensive of the four categories. Many fast-growth beverage brands stay co-packed past $200M because the rate of growth keeps the demand-side dollar more valuable than the in-house margin dollar. The brands that have moved in-house earlier — and made it work — have almost always done so by acquiring an under-utilised facility at a discount rather than building greenfield.
- 03 Supplements: Owned in-house typically pencils at $100-150M revenue on a platform, with the threshold pulled lower when IP-sensitivity or specialised processes (certain probiotics, gummies, novel dosage formats) make co-pack uncomfortable. Below the threshold, the supplement category has unusually deep contract-manufacturing capacity at competitive rates and the asset-light model is structurally preferred by sponsors.
- 04 Personal care and beauty: Many brands stay outsourced past $200M revenue, particularly in beauty where contract-manufacturer capacity is deep and the strategic-buyer cohort (L'Oréal, Estée Lauder, Coty) often has under-utilised plants of its own and will not credit a brand's captive manufacturing in a sale process. In-house fill-finish may start to pencil at $75-150M for brands with specific IP or innovation-speed advantages, but is rarely the path of highest enterprise-value return.
Across all four categories, the buyer-mix effect that Cascadia and Auxo both call out matters in the exit. Strategic buyers typically have under-utilised plants of their own and do not pay a premium for the brand's captive manufacturing unless it is genuinely advantaged (specialised process, locked-in IP, location-specific cost advantage). Financial buyers often penalise heavy asset intensity with a lower multiple versus an asset-light brand of identical margin profile, because the asset-light brand is easier to scale, easier to refinance and easier to exit. This means the plant decision is not just an operating question — it is a multiple question, and the multiple penalty for the wrong-stage in-house transition can exceed the operating-margin gain over a typical hold period.
Strategic buyers will not credit your plant unless it is genuinely advantaged. Financial buyers often penalise heavy asset intensity. The plant decision is a multiple question as well as an operating question.
07 How we run the decision in practice
When we are asked to advise on this question, we run a five-step process. Step one: quantify the current co-pack economics. Conversion cost per unit, co-pack share of net revenue (typically 10-18% is what we see), three-to-five-year volume plan by SKU and format with base, upside and downside scenarios, and a clear-eyed view of co-packer constraints — capacity, MOQs, lead times, changeover inflexibility, IP risk. The brands that do this badly are the brands that have never properly priced their co-pack base.
Step two: model an owned-plant P&L honestly. Capex range by category from the bands above; fixed-cost stack ramped over years one to three; variable cost per unit (labour plus utilities plus maintenance) against the co-pack benchmark; utilisation ramp under three scenarios; fixed cost per unit at each utilisation step; total COGS per unit and gross margin at each step; annual incremental EBITDA versus the co-pack baseline; NPV and IRR on the plant project against a realistic cost of capital. Honest in this context means including the QA over-build, the ERP cost, the working-capital pickup and the 12-18 month implementation timeline.
Step three: run the capital-displacement comparison. Take the same dollar of capex and model it deployed into marketing and trade (MER 1.5-3.0x for strong omnichannel brands), distribution and channel expansion ($0.5-2M slotting plus promo per major retailer yielding $10-25M of incremental annual retail sales at NielsenIQ benchmarks), and inventory and working-capital fuel (improved OTIF protecting 1-3 percentage points of ACV in key retailers). Convert each to incremental EBITDA and enterprise value at the brand's exit multiple. Compare against the plant IRR. For a brand still in the high-growth phase, the demand-side dollar wins almost every time.
Step four: factor in the strategic and operational considerations the financial model cannot fully capture. IP and process secrecy. Quality and regulatory risk tolerance. Supply assurance and retailer-penalty risk. Optionality to co-manufacture for others as a way to fill spare capacity at 20-30% gross margin on conversion. Step five: decide on the phased path — usually some version of stay-and-optimise (cut 5-10 points from COGS through spec, pack and scheduling work), or partial insource (bring one process step in-house while leaving filling outsourced, or acquire a small under-utilised facility rather than build greenfield), or full vertical integration (once revenue is at the category threshold and ≥70-80% utilisation is underwritable inside two-to-three years and the brand has largely exhausted higher-ROIC demand-side investments).
Most of the brands we have advised on this have ended at step five with one of the partial paths — phase the decision, prove the utilisation, preserve capital flexibility — rather than the full vertical integration that the deck started with. The brands that have ended at full integration have done so with eyes open about the overhead step-up, the implementation timeline and the multiple-penalty in a sponsor-led exit. The companion CFO-perspective work on the $1M-equipment capex decision in manufacturing at /blog/factory-floor-capex-decision-cfo-1m-equipment/ applies the same operating logic to the line-by-line equipment decision once the in-house path is chosen.
08 What we are watching into 2026 and 2027
Three structural shifts are pulling the threshold numbers around in 2026 and into 2027. First, tariff-driven raw-material volatility is making co-pack contracts harder to price and harder to renegotiate at the typical two-year cycle — some of the brands we work with are seeing co-pack conversion fees rise 8-15% on renewal, which compresses the margin advantage of staying outsourced and pulls the threshold revenue band lower by 10-20%. Second, retailer consolidation and the rising bar on OTIF and fill-rate (Costco, Walmart, Target all tightened in 2025) is making supply-assurance arguments stronger for hero SKUs — captive control of the top 20% by velocity is becoming a defensive position rather than a margin position. Third, sponsor preference for asset-light businesses has continued to widen the multiple penalty for asset-heavy mid-market consumer brands; the 1.5x EBITDA multiple penalty we saw in 2024 has widened toward 2.0x in the 2025 deal cohort, which is material against the operating-margin gain a captive plant generates over a three-to-five-year hold.
The Q2 2026 read on the mid-market consumer M&A multiple environment is documented in the DTC and CPG multiples post at /blog/dtc-cpg-mid-market-ma-multiples-q2-2026/ and remains the single best framing for the multiple question that runs alongside the operating question on this decision. The manufacturing M&A multiples Q2 2026 post at /blog/manufacturing-ma-multiples-q2-2026/ is the comparable manufacturing-sector frame for what the asset-heavy multiple penalty actually looks like in current transactions. Read together, those pieces and this one are the working library on the production-model decision for mid-market consumer brands in 2026.
Frequently asked questions
At what revenue level does it make sense for a CPG brand to bring production in-house?
How much does it really cost to build a mid-market CPG plant?
What gross margin uplift can I realistically expect by going in-house?
Why is hybrid (core in-house, long-tail co-pack) often the right model?
What overhead categories get most underestimated in in-house transitions?
How does going in-house affect my valuation in a sale process?
What is the capital-displacement argument against in-house production?
Capex ranges: Cascadia Capital Advanced Manufacturing and CPG commentary; Auxo Capital Advisors CPG M&A practice notes; Plante Moran private-equity manufacturing benchmarks; RSM working-capital and capex strategy commentary; NetSuite manufacturing capex guidance; Clarkston Consulting asset-intentional manufacturing framework.
Operator commentary and named-brand examples: BevNet, FoodNavigator-USA, Packaging World, Manufacturing Dive plant-opening coverage 2023-2026; Intrepid Investment Bankers in-house manufacturing analysis; Edible-Alpha co-pack vs in-house operator guidance.
Margin benchmarks: McKinsey CPG insights on rethinking costs and asset-light strategy; PwC Strategy& Consumer Packaged Goods Outlook 2026; Trust Cultivar and Kickfurther CPG gross-margin benchmarks; Whipstitch Capital sponsor commentary on margin/capital trade-offs.
Overhead and system-build benchmarks: BDO and RSM mid-market manufacturing operational-risk and ERP-implementation commentary; Eurofins, Mérieux NutriSciences external lab pricing references; SQF/BRCGS/FSSC 22000 certification scheme fee schedules.
Full source list at content-pipeline/research/copack-vs-inhouse-cpg-capex-margin/sources.md in the Putra & Co content pipeline. Named-brand commentary is illustrative and drawn from public operator interviews; specific production-model details for any named brand are subject to change.