The manufacturing ERP requirements that decide a mid-market project sit outside the feature matrix. Nine factors matter, and three decide the outcome: whether the system costs production the way your plant runs, whether it survives a demo on your own data, and who implements it. Test each factor before you score anything.
A vendor’s features matrix usually says yes in every cell. That is why three demos in, the finalists start to look the same as each other, and better than what you have. One buyer in Crestwood’s research put it plainly after a third demo: “they all seem to have this setup.”
The worry that should drive the evaluation sits somewhere else. Crestwood has analyzed 28 of its own meetings with ERP buyers. Implementation risk, the question of whether the business can get from here to there, came up in 23 of them. Product-fit risk, whether the software can do the job, came up in 20. A feature checklist scores the second worry and ignores the first.
This list gives you the nine manufacturing ERP requirements worth scoring, what to ask about each, how to test it on your own data, and the warning sign that tells you a finalist has failed.
The nine manufacturing ERP requirements at a glance
Six of the nine are fit factors a demo can prove or disprove: production model, costing, BOM and routing, inventory and traceability, planning, and shop floor capture. Three are project factors that no feature list asks about: what you keep and must connect, growth and who can use the system, and the team that implements it. Features decide whether a system makes the shortlist. These nine decide which finalist works in your plant.
| Factor | What to ask | How to test it | Warning sign |
|---|---|---|---|
| 1. Production model fit | Which of our production modes does the system run natively: make-to-stock, make-to-order, configure-to-order, batch? | Run one order of each mode you actually ship, start to finish | A custom screen or a workaround appears for your most common order type |
| 2. Costing method | How does it cost a finished part: standard, actual or average, and where do labor and overhead absorb? | Close one month on your own routings and read the variance report | Variances only make sense after an export to a spreadsheet |
| 3. BOM and routing depth | Multi-level BOMs, revisions, alternate routings, a configurator if you build options? | Load your deepest BOM and change one component revision mid-order | The engineering change sits outside the system |
| 4. Inventory and traceability | Lot and serial control, multi-location stock, backflush or issue, recall in one query? | Trace one finished lot back to its raw material receipts | A trace takes more than one screen and one person |
| 5. Planning | Does MRP plan from real demand, lead times and capacity, or from a reorder point? | Feed it a demand spike and read what it tells you to buy and when | Planners re-key the MRP output into a spreadsheet to trust it |
| 6. Shop floor capture | How do time, quantity and scrap get recorded at the work center? | Have an operator clock a job and report scrap on the floor device | Output reaches the system at shift end, by hand |
| 7. What you keep and must connect | Which systems stay, and how does each one connect: EDI, CAD or PLM, shipping, eCommerce? | Push one real EDI order through to an invoice | An integration is described as “on the roadmap” |
| 8. Growth and who can use it | Second plant, second entity, more users: what changes in setup and in cost? | Ask for a written quote at today’s headcount and at double it | Users are rationed, so data entry pools in a few people |
| 9. The implementation team | Who configures, trains, tests and supports, by name and by phase? | Meet the people who will do the work, and call two of their manufacturing clients | The team you meet in the sales cycle is not the team on the project |
Fit factors you can test on your own data
A fit factor passes or fails in a demo, provided the demo runs on your parts, your routings and your orders rather than the vendor’s sample company. If a finalist will not run the test, treat that as the result.
1. Production model fit
Start with how you make things. Make-to-stock, make-to-order, configure-to-order and batch production each put different weight on the same modules. If your plant runs more than one mode, a system that handles the main one natively and the second through a workaround will carry that workaround for as long as you run it.
Watch for the moment a vendor says a capability is a configuration away. One buyer heard the word configurator for the first time during an evaluation and concluded it was “probably something we do need.” Another described a common frustration with industry-specific systems: they do the core process well and “everything else suffers.”
2. Costing method
Costing is the factor where the floor and the books meet, and it is the one two systems with identical feature lists differ on most. Ask how a finished part gets its cost: standard, actual or average. Ask where labor and overhead absorb, and how purchase price, usage and efficiency variances report.
Then close a month in the demo. A controller reading the variance report on your own routings learns more in twenty minutes than a features matrix says in twenty pages. Firewire Surfboards described its starting point as having no insight into the finances of its weekly production. That is a costing failure, and nobody finds it on a checklist.
3. BOM and routing depth
Multi-level bills of material, revision control, alternate routings and, if you build options, a configurator. Load your deepest BOM into the demo and change one component revision on an order already released. The test shows whether engineering change lives inside the system or beside it.
The warning sign is custom screens. One buyer’s team had new screens built to avoid the standard manufacturing order screens while still pulling the bill of material. That is a system being worked around by the people it was bought for.
4. Inventory and traceability
Inventory nobody trusts was raised in 16 of the 28 buyer meetings Crestwood analyzed, which ties it with three other patterns for second place, behind administrative work that grows with volume. The details repeat: on-hand quantities that do not match the shelf, components in stock when the system expects the parent item, lot picking where “sometimes it works, sometimes it doesn’t.”
Ask for lot and serial control, multi-location stock and a clear choice between backflushing and issuing material. Then trace one finished lot back to its raw material receipts. Food manufacturers have a dated reason to run that test: the FDA’s Food Traceability Rule requires records of key data elements at critical tracking events, provided to the FDA within 24 hours of a request or another time it agrees to, and the FDA has said it will not enforce the rule before July 20, 2028.
5. Planning
Ask whether material requirements planning works from real demand, lead times and capacity, or from a reorder point with a better name. Planning and forecasting done by hand, or not at all, came up in 10 of the 28 meetings. One buyer set min and max points from a six-month average usage calculation kept outside the system.
The test is a demand spike. Add a large order to the demo and read what the system tells you to buy, make and reschedule. If your planners would re-key that output into a spreadsheet before trusting it, the module has failed.
6. Shop floor capture
Time, quantity and scrap have to reach the system where the work happens, or every report after that is an estimate. Nobody can see the operation was a pattern in 16 of the 28 meetings. One buyer described checking efficiency targets with a report run by hand “every couple of weeks.”
Have an operator clock onto a job and report scrap on the device your floor would use. If the vendor’s answer is a terminal in the office, the data will arrive at shift end and the visibility you are paying for will arrive with it.
Project factors no feature list asks about
Three factors decide whether a system that fits on paper goes live and stays live. None appears on a features matrix, and all three show up in the objections buyers raised in Crestwood’s research. Score them in writing, with names and dates, on the same sheet as the fit factors.
7. What you keep and must connect
Buyers arrive with systems they have already declared untouchable. Across 25 of the meetings Crestwood analyzed, buyers named 57 things they had to preserve, usually an adjacent system they liked, a control they are audited on or a report shape they depend on. List yours before the first demo, then ask how each one connects: EDI, CAD or PLM, shipping, eCommerce.
Connections are where the hours go when they fail. Summit Brands onboarded more than 120 trading partners across three entities in under 10 months after replacing a homegrown EDI setup with an integrated one, and freed more than 120 hours of staff time each month. Push one real EDI order through to an invoice in the demo.
8. Growth and who can use it
Ask what changes when you add a second plant, a second legal entity or fifty more users, in setup and in cost. Get the answer as a written quote at today’s headcount and at double it. Multi-entity complexity came up in 11 of the 28 meetings.
Who can use the system matters as much as how many. Advanced Plastic Corp described limited user licenses that led to inefficiencies, and reporting with no drill-down. Firewire Surfboards’ CFO said the licensing model “weighed heavily in our decision.” When users are rationed, data entry pools in a few people, and that is how knowledge ends up trapped in one person.
9. The implementation team
Partner risk came up in 13 of the 28 meetings, and where it appeared it was usually the loudest thing in the room, because it was almost always driven by a specific past experience. Score the team with the software. Ask who configures, who trains, who writes the test plan and who answers the phone after go-live, by name and by phase.
When Green Bay Packaging evaluated four ERP applications to replace a 30-year-old mainframe, its Corporate Director of Accounting described the requirement in one sentence: “We had to have a system that was flexible, user-friendly and came with expert support.” Two of those three are about software. The third is about people.
How to score the nine factors across finalists
Weight the factors before the first demo, script every demo from your own transactions, run the same script on each finalist, and score with the evidence written beside the number. Order matters more than the scale you use, because weights set after the demos drift toward whichever vendor presented best.
Crestwood calls step four a Day in the Life Demo: five to seven of the buyer’s own workflows, each run from a real trigger to a real outcome, built from what discovery surfaced rather than from a standard script. Ask every finalist for the same thing, whoever they are.
| What you score | Evidence that earns a 3 | Evidence that earns a 1 |
|---|---|---|
| Fit factors (1 to 6) | Ran on your data, in the demo, and the result matched what your team does today or does it better | Shown on the vendor’s sample company, or described on a slide |
| Project factors (7 to 9) | In writing: integration method, a quote at double headcount, named project team and two references you called | Verbal assurance, a roadmap date, or “we’ll assign the team after signature” |
| Weight | 3 for the factor your last system failed on, 2 for factors that touch month-end, 1 for the rest | Weights set after the demos |
Crestwood uses the Three-Legged Stool framework to check the result: Partner, Platform and People. If any leg is weak, the project is at risk regardless of which platform scores highest. Factors 1 to 6 test the Platform leg and factor 9 tests the Partner leg. The People leg is tested by who on your team runs each demo script, and whether a colleague could cover for them during a vacation.
What a feature checklist leaves out
A requirements list built from your current system is mostly inherited. When Crestwood classified the 78 customizations its buyers described, 13 reflected how the business actually competes, 30 were necessary controls, 20 were legacy workarounds and 15 could not be classified because nobody asked why they existed. Treat that ratio as directional, and classify your own list before it becomes a scoring sheet.
Sort every requirement into three piles: keep it because it is how you win work, keep it because an auditor or customer requires it, or retire it because it exists only to get around the old system. The third pile should never reach a demo script. A buyer in the research said it directly: they did not want to “take what we have today and copy and paste” it into a new system.
The other thing a checklist leaves out is the system you already have. In four meetings in the research, when a requirement looked unmet, the buyer reached for their current workaround within seconds rather than for another vendor. Each finalist is also competing with the workaround your team already runs, and the score has to show why it is worth leaving.
Where Crestwood stands on this list
Crestwood sells, implements and supports two of the platforms a mid-market manufacturer is likely to shortlist, Acumatica and Microsoft Dynamics 365 Business Central. That is why this list ranks no products and names none inside the factors. Either platform can win or lose on these nine, and the factor that decides it is usually in your plant, not in the brochure.
Before the evaluation, most teams hold a features matrix where every cell says yes and a decision that will be made on demo impressions. After it, they hold a weighted score with evidence beside every number, a costing test run on their own routings, and a named project team they have already called references for. The bridge between the two is knowing which of the nine factors carries the most weight for your plant, and that is a question about how the business runs today.
The business may not have a people problem. It may have an operating model problem. Crestwood’s Business Optimization Assessment looks at workflow, reporting, system utilization, connected systems, and finance and controls, and you receive a written, prioritized summary within 24 hours that is yours to act on with Crestwood or without it. There is no cost. We sell, implement, support, and optimize it. An in-house team of industry experts. One hand to shake.
For the platforms and industry detail behind this list, see Crestwood’s manufacturing ERP page.
Frequently asked questions
What are manufacturing ERP requirements?
Manufacturing ERP requirements are the capabilities and conditions a system must meet to run a plant’s production, costing and inventory. For a mid-market manufacturer, nine matter most: production model fit, costing method, BOM and routing depth, inventory and traceability, planning, shop floor capture, connections to systems you keep, growth and user access, and the implementation team.
What are the main features of a manufacturing ERP?
The core features are bills of material and routings, production orders, material requirements planning, inventory with lot and serial control, shop floor data capture, product costing and financials in one ledger. Most shortlisted systems list all of them. The difference between finalists shows up when each feature runs on your own data, which is why the test matters more than the tick.
What is the most important factor when choosing ERP for a manufacturer?
Costing method is the fit factor to prove first, because it is where production and the books meet and it fails last, at the first month-end after go-live. The implementation team is the project factor to prove first. In Crestwood’s research, implementation risk came up in 23 of 28 buyer meetings, more than product fit.
How do you test an ERP system before buying it?
Script the demo from your own transactions and run the same script on every finalist. Use one order of each production mode you ship, your deepest bill of material, a lot trace, a demand spike and a month-end close. Score each result with the evidence written beside it, then call two manufacturing references for each finalist.
Should a manufacturer choose an industry-specific ERP or a general one?
Choose the one that passes your production model and costing tests and still covers finance, planning and integration without workarounds. An industry-specific system can run a narrow process well and leave everything else weak. A general system can need configuration for your core process. Run the same nine tests on both and let the evidence decide.
Does the FDA Food Traceability Rule affect ERP selection?
For food manufacturers handling foods on the FDA’s Food Traceability List, yes. The rule requires records of key data elements at critical tracking events, provided to the FDA within 24 hours of a request. The FDA has said it will not enforce the rule before July 20, 2028. Test a full lot trace in every demo.
