ERP selection for a $5-20M manufacturer: the 6-week evaluation plan

At a glance

"Go figure out our ERP options" does not require an 18-month enterprise RFP built for companies ten times your size, nor a trade-show coin flip. At $5-20M, what works is a bounded, week-by-week plan you can run with the team you already have: six weeks, one deliverable per week, ending in a defensible decision and a signed-off implementation path. This is that plan. We sell one of the systems you might evaluate, ERPNext, and the one section where that shows up is labeled; the method works regardless of what you pick.

Why ERP selection goes wrong at this size

Three failure modes account for most of the wreckage. The common thread is that the evaluation never forced the systems to run your business, on your data, against your numbers. The six-week plan below exists to force exactly that.

• The over-scoped process. A mid-market company copies an enterprise methodology: 400-requirement RFPs, scoring matrices nobody reads, a 12-month timeline during which the business changes and the champions leave. Diligence theater, not diligence.
• No process at all. The owner sees a demo at a trade show, likes the salesperson, signs. Fit against the company's actual 15-20 core processes was never tested, so the gaps surface during implementation, where they are most expensive.
• Demo theater. The vendor drives every demo through their polished happy path with their sample data. Everything looks great because it was designed to. Your weird-but-critical process, the one that makes you money, never appears on screen.

How long should ERP selection take for a $5-20M manufacturer?

Six focused weeks is enough: one week documenting your 15-20 core processes, one building a shortlist of three, two running scripted demos and sandbox pilots, and two on references, five-year TCO, and the decision. Selections that stretch 6-18 months usually stall from scope creep, not diligence.

The constraint that makes six weeks work is focus: one owner, a small selection team, and a hard rule that the evaluation tests core processes rather than cataloging every feature anyone might ever want.

Week 1 - document the 15-20 processes that run your business

Not a 400-line RFP. A one-to-two-page process inventory: the 15-20 workflows that, if the ERP handles them well, the business runs, and if it handles them badly, the project fails. For a typical manufacturer that means quote-to-order, BOM and routing maintenance, work order execution, purchasing and receiving, inventory moves and cycle counts, lot and serial traceability, shipping and invoicing, month-end close, job costing, and your two or three genuinely weird processes, the ones you would have to explain to a new hire for an hour. Those weird ones matter most; they are what demos are designed to avoid.

For each process, write what triggers it, who touches it, what working looks like, and one real example with real data you can hand a vendor. Assemble the selection team now: one executive sponsor, one finance owner, one ops owner, one person who will live in the system daily. Four people, not twelve.

Prefer not to start from a blank page? Our free expert audit is effectively Week 0 as a service: we map your processes with you and tell you whether ERPNext belongs on your shortlist, including when it does not.

Week 2 - build the shortlist of three and kill the rest

Apply four brutal filters to the long list, in order. Three finalists, no more; every additional vendor adds a week of demos and sandbox time for diminishing information. Write one sentence per killed vendor explaining why, so you do not relitigate the long list later.

• Size fit. Does the vendor have referenceable customers at $5-20M revenue? Enterprise systems at your size mean you are the smallest, least-important customer; micro-business tools mean you will outgrow it mid-implementation.
• Industry fit. Manufacturing-capable out of the box, with BOMs, work orders, and MRP, not via a partner ecosystem.
• Budget fit. Rough five-year all-in against your budget. Kill anything 3x over before wasting a demo slot. If you are on QuickBooks plus an inventory bolt-on, our QuickBooks and Fishbowl comparison shows how that stack's replacement math typically shapes a shortlist.
• Deployment fit. Cloud, self-hosted, or either, matched to your IT reality rather than ideology.

Week 3 - scripted demos with your data, you drive the agenda

Send each finalist the same demo script built from your Week-1 process inventory: here are eight scenarios, in this order, using this data we are providing; show each one end-to-end, including the accounting entries it creates. Attach your sample items, a real and sanitized BOM, and your ugliest recent order.

In the demo, hold the line. When the vendor detours to their dashboard tour, pull them back to scenario four. Score each scenario pass, pass-with-configuration, or fail on a shared sheet, the same sheet for every vendor, scored independently by each team member before comparing. The phrase to be suspicious of is "that's configurable"; the correct response is "great, configure it and show me." A vendor who cannot demo your scenario with two weeks' notice is telling you something about implementation.

Week 4 - sandbox or pilot with your top two

Demos are theater even when you write the script; sandboxes are evidence. Get hands-on access to your top two systems. Most credible vendors will provide a trial instance, and open-source options like ERPNext can be spun up free. Load ten items, one BOM, one customer, one vendor, and run one real order end-to-end: quote, order, work order, material issue, receipt, shipment, invoice, payment. Have the daily-user team member drive, not the vendor.

You are answering three questions no demo can. Can our people navigate this without a vendor driving? Do the accounting entries land where our controller expects? Where did we get stuck, and was the vendor's answer to our stuck-point fast and honest?

Week 5 - reference calls and TCO math

References: two or three per finalist, at your revenue size, in or near your industry, and you pick from a list rather than just the vendor's favorite. Ask what broke during implementation, what month-end actually looks like, how support response times are now two years in, and what they would do differently. The single most revealing question is "show me one measurable before and after." A real reference has one instantly. When prospects ask us, we point at Texas Stone Designs, where production scheduling went from 24 hours to 5 minutes and CEO Micah Grant will say so on the phone, or at Mr Goli Soda's roughly 80% ERP cost reduction moving off Odoo and franchise onboarding cut from 60 to 35 days. Demand that same specificity from every vendor on your list, including us.

TCO: five-year, all-in, the same spreadsheet for all finalists. Include subscription or license fees at your year-5 user count (per-user pricing compounds, so model growth, not today), implementation, data migration, integrations, hosting, annual support and maintenance, upgrade costs, and internal hours. Data cleanup and parallel running are real labor; budget roughly 1.5x workload for finance during the parallel-run weeks. Label every number a vendor would not publish as estimated, and note which vendors forced you to estimate, because that is Week 6 input. Our ROI calculator structures this math if you want a starting template.

Week 6 - decision scorecard, negotiation, and implementation plan

Assemble the one-page scorecard: process pass/fail counts from demos and sandbox, reference quality, five-year TCO, implementation timeline, and a risk note per vendor. Score independently, then discuss divergences; that discussion is where the real decision gets made.

Before signing, review the implementation plan with the same rigor as the software: named team, week-by-week phases, your responsibilities such as data cleanup hours and training time, parallel-run approach, and go-live criteria. An implementation quote produced without a discovery conversation is a guess you will pay for later. Negotiate with your runner-up still warm, since that is your only leverage, and get exit terms in writing: data export format, what happens at non-renewal, and who owns customizations.

What should be on an ERP evaluation checklist?

Your 15-20 core processes with pass/fail demo results per vendor; data migration scope; integration needs; five-year all-in TCO covering licenses, implementation, hosting, support, and internal hours; implementation timeline and methodology; references from companies your size; and exit terms and data ownership. Score each vendor on the same sheet.

The discipline matters more than the artifact: same sheet, same scenarios, same questions for every vendor. The moment each vendor gets evaluated on their own terms, you are back to comparing demos instead of comparing fit.

Red flags that predict a failed implementation

Any two of these predict trouble: refusing to publish or state pricing, no references at your revenue size, demos that avoid your scripted scenarios, "that's configurable" as the answer to everything, implementation quotes without a discovery phase, and pressure to sign before you have run one real order through a sandbox.

Unpublished pricing is not automatically sinister, but it transfers all negotiating leverage to the vendor and usually signals pricing that flexes to whatever your budget appears to be. Make opacity cost them points on the scorecard. We publish ours on the pricing page, partly because we believe sellers who make buyers guess deserve the skepticism they get. Reference-opacity is worse than price-opacity: a vendor with no referenceable customer at your size is proposing you as the experiment.

"Everything is configurable" is technically true of every modern ERP, which is precisely why it is an empty answer. Configurable at what cost, in whose hours, maintained by whom through upgrades? A good vendor answers a specific process question with a specific mechanism, or an honest "that would be custom work, roughly X hours." Vendors who respect you say no, or say it costs extra, in the sales cycle. The ones who never do are deferring every no until after the contract.

Where ERPNext fits in a shortlist, and when it should not

Declared interest: this is the section where we are a vendor, not an advisor.

ERPNext earns a shortlist slot for $5-20M manufacturers when the profile is outgrowing QuickBooks plus bolt-ons, needing unified accounting, inventory, and manufacturing in one database, and a budget where per-user licensing fails the five-year TCO test. It is open source with zero per-user fees, so the TCO line most ERP decisions hinge on simply is not there, and our typical implementations for this segment run 6-8 weeks (our typical result at this size and scope, not a universal promise).

It should not be on your shortlist if your Week-1 process inventory includes hard requirements for native finite-capacity scheduling (APS), PLM with engineering change control, automated multi-entity consolidation eliminations (ERPNext's are manual journal entries at close), turnkey US bank feeds (statement imports are the default; live feeds are scoped integration work), or US payroll in the ERP (the pattern is Gusto or ADP alongside). We maintain the complete list, with workarounds and who each limit blocks, at ERPNext's honest limits; run your Week-1 inventory against it before you spend a demo slot on us.

Get the free expert audit is the Week-0 shortcut: we map your processes, give you a straight fit or no-fit answer on ERPNext, and you keep the process map either way.