68–73%. That's how often ERP implementations in manufacturing-heavy companies fail to meet their stated objectives. Budget overruns run 189–215% of the original quote. And up to 30% of the features you paid for are never actually used.

Nobody puts that slide in the sales deck.

You're Too Big for QuickBooks. Not Big Enough for SAP. Enjoy.

This is where most 50M–500M CPG brands live: In a gap the software industry largely pretends doesn't exist.

You're running omnichannel fulfillment across grocery, club, mass, DTC, and marketplaces, each with different order profiles and retailer penalties. You've got co-manufacturers with variable lead times and opaque execution. You need lot traceability, shelf-life management, recall readiness. Trade promotion is probably your largest SG&A line. And your EDI setup is a pile of partner-specific maps someone built three years ago that nobody fully understands.

That's not a simple operation. QuickBooks isn't built for it. Neither is Fishbowl or Unleashed, even though they'll tell you otherwise until you're a year in and buried in spreadsheet workarounds.

61% of CPG executives describe their supply chain tech stack as: multiple legacy systems, multiple ERPs, and ubiquitous spreadsheets. That's not a technology problem. That's a market failure.

But the Tier-1 ERP path? That's its own kind of trap.

What Actually Happens When You Buy a Big ERP

Here's a cost breakdown nobody leads with.

For a 50-user NetSuite deployment in manufacturing: roughly \(269K year one, ~\)144K annually after that. And that's if everything goes cleanly. It often doesn't. Implementation services routinely run 2–3× the annual license fee on their own.

Microsoft Dynamics 365 for a typical mid-market org: \(375K to \)1.75M in year one. Of that, only 15–20% is licensing. Consulting runs 40–50%. Data migration, integrations, training. That's another 15–25%. You're basically buying a consulting engagement that happens to come with software.

SAP's GROW with SAP promises 4–6 month cloud deployments. Then you meet your integration bill.

And this is before accounting for the 68–73% chance the project doesn't deliver what you scoped.

For a CPG running mid-double-digit EBITDA, the math is brutal. You're committing \(300K–\)1M+ over 1–2 years with a coin-flip on success. Accenture and McKinsey have both documented what 'successful' ERP implementations often look like five years later: multiple legacy systems, multiple ERPs, and spreadsheets filling the gaps.

The Consultant Is the Product

Let's be direct about what's actually happening here.

The software vendor isn't your real supplier. The system integrator is. And they get paid whether the project succeeds or not.

Dynamics 365 breakdowns put consulting at 40–50% of year-one cost. NetSuite's own documentation says implementation services often run 2–3× annual license spend. So when the project runs late — and it frequently does — you call the consultant. Who bills hourly. Who has every incentive to scope the next phase.

This isn't a knock on consulting firms. Complex implementations are genuinely hard work. But the economic structure is backwards for how mid-market CPG companies actually operate.

You don't have a 50-person IT change management org. You have a supply chain VP, maybe a director of IT, and a team who also has to run the actual business while the ERP is being implemented. The 'consultant tax' is real, and it compounds over time.

AI Just Changed the Math on Building

Here's where it gets interesting.

For most of the past decade, the calculus was clear: building a custom platform required 12–18 months of development, investment 2–3× what a packaged program costs, and breakeven over 5+ years. Plus 60–80% of ongoing IT budgets eaten by maintenance forever. Forget it.

That calculation has shifted.

GitHub Copilot controlled study (Microsoft/GitHub): developers completed coding tasks 55.8% faster: 1h11 vs 2h41 for the same work. McKinsey's research puts code writing and documentation at roughly 2× faster with AI tools. ZoomInfo's internal data: 63% of developers completing more tasks per sprint, median 20% reduction in completion time.

Stack those numbers on top of mature cloud infrastructure, API-rich SaaS for commodity functions, and composable architecture where you swap services instead of re-platforming — and the economics of a targeted, modular build start looking very different from 2019.

First release in 4–6 weeks instead of 12–18 months. Lower cost on follow-on capabilities. No consultant holding the keys.

The Autonomous Planning Upside Nobody Is Modeling

This number is worth sitting with.

Autonomous supply chain planning, where demand shifts and promotion changes ripple automatically through inventory, production, and procurement, delivers up to 4% revenue uplift, 20% inventory reduction, and 10% supply chain cost cuts, per McKinsey's research on consumer goods companies.

For a \(200M CPG, that's roughly \)8M in revenue and meaningful working capital release. Not a rounding error.

But you can't get there with six siloed systems and a spreadsheet connecting them. Autonomous planning needs end-to-end data and orchestration. It's not a product you buy and install. It's a capability you build over time with clean data plumbing underneath it.

The companies capturing this aren't waiting for their ERP go-live to get rescheduled.

The Architecture That Actually Works for This Segment

The answer isn't 'ditch your ERP.' That's not realistic and it's not the point.

The pragmatic play for a \(50M–\)500M CPG is to keep a lean core ERP (NetSuite, D365, SAP public cloud) for what it's actually good at: financials, GL, basic order-to-cash, simple manufacturing. Don't try to squeeze CPG-specific workflows into it. You'll lose.

Instead, build a platform layer underneath: API gateway, event bus, common data model. Then build the capabilities that actually differentiate your operation as modular services around that platform:

• Demand forecasting and inventory planning that consumes POS, orders, and promotion calendars

• Omnichannel order orchestration with allocated inventory visibility across DCs, 3PLs, and co-mans

• Trade promotion and deduction management with post-event ROI

• Co-manufacturer collaboration with real-time schedule, yield, and quality data

• EDI/retailer integration orchestration with monitoring and exception handling

You do it in phases. Foundation first (3–6 months). Planning next. Then order management. Then trade. Each phase has a measurable outcome. Nobody holds your roadmap hostage.

The core rule: buy commodity functions, build what differentiates you. AI makes 'build' cheap enough that this is now a real option in the \(50M–\)500M range. It wasn't five years ago.

The Risks Are Real. Don't Skip This Part.

AI-generated code speeds you up. It can also speed you into a mess if you're not disciplined.

The 60–80% maintenance trap doesn't care that Copilot wrote your boilerplate. If you don't enforce architecture standards, keep services small and decoupled, and run real CI/CD and code review, you will end up with a custom system that costs as much to maintain as your old ERP did.

You also need actual engineering talent. AI doesn't replace senior architects. A lot of mid-market CPG companies don't have a strong product engineering function in-house, and that's a real constraint. The mitigation is building a small senior core team and using implementation partners for setup — but with explicit knowledge transfer milestones, not perpetual SI dependence.

Scope creep will kill you faster than any consulting overbill. The moment you start re-implementing accounting features because someone asked for a custom GL report, you've lost the plot. Build only where you differentiate. Have a governance body that can actually say no.

So Where Does This Leave You

The structural gap for mid-market CPG is real and well documented. SMB tools break under real operational complexity. Enterprise ERPs have failure rates above 70% and economic structures that don't fit your capital profile or your change management capacity.

AI-assisted development and composable architecture have created a third path that didn't make economic sense before 2023.

It's not easy. It requires engineering investment, clear priorities, and genuine discipline about scope. But for CPG operators who are tired of either drowning in spreadsheets or gambling $1M on a consultant-heavy ERP program — there's finally a different conversation worth having.

The companies figuring this out right now aren't waiting for the software industry to solve the mid-market problem.

They're solving it themselves.