Manufacturing ERP software: how to choose one

Manufacturing ERP software is the same category name as ordinary business ERP and a different product underneath. The difference is not the module list. It is that the system has to hold a bill of materials, explode it into demand, and then keep the arithmetic true while a machine breaks, a supplier ships short and somebody reworks a batch by hand.

That is why generic accounting-first ERP fails on a shop floor, and why the price spread between vendors reaches a factor of six for the same headcount. What the vendor decided to count, per seat or per site or per transaction, matters more to the invoice than anything in the demo.

Every price below was read on the vendor’s own pricing page on 16 September 2026. All of them will drift, so read the model rather than the number.

Manufacturing ERP software in numbers

Eight figures worth carrying into a vendor conversation, all read on 16 September 2026.

• $49 per user per month is where dedicated manufacturing ERP software starts when it charges by seat, on MRPeasy Starter, with a minimum of one user
• $299 a month flat is Katana’s Core plan, which includes unlimited users and one location and charges sales orders by usage
• $199 a month is what Katana’s Manufacturing Management add-on costs on top of that, and traceability is another $249
• $80 versus $110 per user per month separates Dynamics 365 Business Central Essentials from Premium, paid yearly, and manufacturing sits in Premium
• $8 per user per month is a Business Central Team Members seat, which is what decides how much of the shop floor you can afford to put on screens
• $79 per 10 users is MRPeasy’s bundle rate from the eleventh user onward, which turns a per-seat model into something closer to flat
• $0 is the licence cost of ERPNext under AGPL-3.0, with managed hosting on Frappe Cloud starting at $5 a month per site
• No number at all is what Acumatica publishes: its pricing page quotes on applications, transaction volume and storage, and states outright that you do not pay for user seats

Six vendors, four pricing units. That spread is a licensing decision, not a quality signal.

What manufacturing ERP software adds to plain ERP

The dividing line is the bill of materials and everything downstream of it. General ERP records that you bought steel and sold a bracket; manufacturing ERP software knows that the bracket consumes 0.8 kg of that steel, takes eleven minutes on the press, and cannot be promised to a customer until both the steel and the press are free.

Those manufacturing ERP software modules are the ones to check line by line, because vendors in our ERP category use the same words for very different depth.

Module	What it has to do	What thin versions do instead
Bill of materials	Multi-level, with versions and effective dates	One flat level, no history
Routings and work centres	Operations, setup time, run time, capacity	A single lead time per item
MRP run	Nets demand against stock and open orders	A reorder point per item
Work orders	Issue, backflush, partial completion, rework	A stock adjustment
Shop floor reporting	Operator clocks in against an operation	Somebody types quantities at the end of the week
Costing	Standard, actual and variance by operation	Purchase price only
Lot and serial traceability	Forward and backward in one query	A text field for the batch number
Quality control	Hold, inspect, release, reject with reason codes	A checkbox

If MRP and capacity are absent, what is being sold is inventory software with a manufacturing label, and the gap will show up the first time two work orders want the same machine hour. That is a real distinction rather than a marketing one, and it is the reason inventory tools and manufacturing ERP sit in separate parts of any honest catalogue.

Costing is the module that decides whether the finance team trusts the system. Standard costing with variance analysis is the reason a manufacturer buys ERP at all, and it is the first thing missing from the cheap end.

What the bill is actually counting

Four units are in use across the market, and each one rewards a different shape of company. Nothing about the software changes; the invoice changes by a factor of six.

Model	Vendor example, 16 Sep 2026	Rewards	Punishes
Per named user	MRPeasy, $49 to $149 per user per month	Small office teams	Shop floor headcount
Per user, tiered by module	Dynamics 365 Business Central, $80 Essentials or $110 Premium, plus $8 Team Members	Mixed office and read-only staff	Everyone needing full rights
Flat plus add-ons	Katana, $299 a month Core, plus $199 manufacturing and $249 traceability	Many light users	Anyone needing every add-on
Consumption and applications	Acumatica, quoted, no seat charge	Large or seasonal headcount	Transaction-heavy operations
Licence free, hosting paid	ERPNext, $0 licence, hosting from $5 a month	In-house technical capacity	Teams without an admin

At 25 full users the same requirement costs $747 a month on Katana’s flat plan with two add-ons and $2,750 a month on Business Central Premium, from list prices read on 16 September 2026. Same shop, same work, one licensing decision.

Microsoft is the clearest published case of tier-gating, because the manufacturing modules are the whole reason to move up a tier: the Dynamics 365 Business Central pricing page put Essentials at $80.00 per user per month and Premium at $110.00 per user per month, paid yearly, when it was read on 16 September 2026.

That $30 difference is manufacturing and service management, and it applies to every seat, including the ones that never open a work order.

Per seat, where the shop floor is the problem

Per-seat pricing is the one to model carefully, because manufacturing headcount is not office headcount.

Twenty machine operators who each need to report an operation are twenty users at list price, unless the vendor sells a cheap reporting seat, and the same arithmetic catches out teams buying time tracking tools by the head.

Business Central’s $8 Team Members tier exists for exactly that, and MRPeasy’s $79 per 10 users from the eleventh seat has the same effect by a different route.

Consumption, and who it actually suits

Consumption pricing deserves a careful read rather than a dismissal. The Acumatica pricing page, read the same day, publishes no price and instead quotes on the applications you select, your transaction volume and storage, and your deployment choice.

Acumatica states plainly that you pay for functionality rather than user seats, which is genuinely cheaper for a plant with many occasional users and genuinely more expensive for a high-transaction distributor.

Both patterns show up in invoicing and finance tools as well, so a team that has already been through one subscription negotiation will recognise the shape. The difference with ERP is the term length: three years is normal, and the unit you agreed to is the unit you live with.

What manufacturing ERP software costs at three sizes

Same five vendors, priced for three shapes of manufacturer, using only published list prices read on 16 September 2026. Implementation is excluded, and it is not small.

Setup	MRPeasy	Katana	Business Central Premium	ERPNext
5 office users, no floor reporting	$245 a month, Starter	$299 a month, Core	$550 a month	$5 a month hosting
12 office users, 20 floor terminals	$1,264 a month, Professional with bundles	$299 plus $199 add-on	$1,320 plus $160 for 20 Team Members	$20 a month hosting
25 office users, full traceability	$2,290 a month, Enterprise with bundles	$299 plus $448 in add-ons	$2,750 a month	$125 a month dedicated

Three things follow from that table, and none of them are visible in a feature comparison.

Flat pricing wins as headcount grows, and loses on depth. Katana stays near $750 a month at every size, because it charges for capability rather than people.

Against that, sales orders are billed by usage and each capability is a separate add-on, so a plant needing manufacturing, traceability and warehouse management is paying $896 a month on top of Core before anybody logs in.

Per-seat pricing punishes exactly the thing manufacturing needs, which is data from the people doing the work. If operators cannot afford a seat, they report through a supervisor at the end of a shift, and the MRP run is then working from yesterday’s reality.

The free licence is not the cheap option unless the technical capacity already exists. ERPNext costs nothing to licence and $5 a month to host on shared infrastructure, and it costs an administrator who can handle upgrades, backups and a custom report. Priced at a contractor’s day rate, that is more than the Business Central bill at the smallest size.

Discrete, process and project manufacturing want different software

This is the split that decides which shortlist is even worth reading, and most vendors serve one of the three properly. The words appear on every website; the data model underneath does not bend.

	Discrete	Process	Project
What is made	Countable units, assemblies	Batches, blends, volumes	One-off or engineered orders
Structure	Multi-level BOM, routings	Formula or recipe, yields	Work breakdown, milestones
Units	Each, sets	Weight, volume, potency	Hours, phases
Hard problem	Capacity and sequencing	Yield variance, co-products and by-products	Cost to complete, change orders
Traceability	Serial numbers	Lot genealogy, forward and backward	Document revisions
Typical failure	Nested BOMs the tool flattens	A formula forced into a BOM field	Progress billing done in a spreadsheet

Discrete manufacturing ERP software is the default that most products are built for, which is why electronics manufacturing ERP software and furniture or steel plants find a reasonable fit quickly.

Process manufacturing ERP software is the harder purchase: a formula is not a bill of materials, yields vary by batch, and co-products mean one run produces two sellable things plus waste.

Project manufacturing ERP software is different again, because the cost object is the job rather than the item. Teams that end up here often already run project management tools and need the ERP to own cost to complete rather than task lists.

What changes by industry, and what does not

The core engine is the same everywhere. What changes is the compliance surface and the unit of measure, and that is where an industry edition earns its premium or fails to.

Industry	What the software has to add
Food and beverage	Recipe scaling, allergens, shelf life, catch weight, recall in minutes
Bakery	Short shelf life, daily production plans, route and van sales
Pharmaceutical and medical device	Validated environments, electronic signatures, device history records
Chemical and coatings	Formulas by potency, safety data sheets, hazardous shipping rules
Cosmetics and personal care	Formula versioning, filling and packing lines, regional label rules
Garment and apparel	Size and colour matrices, cut plans, subcontract stitching
Plastics and rubber	Moulds and tooling, regrind and scrap, cycle time per cavity
Electronics	Reference designators, approved manufacturer lists, part obsolescence
Steel, glass and furniture	Sheet and coil nesting, offcuts, made-to-measure variants

Food manufacturing ERP software is the clearest example of why the industry label matters. Catch weight alone breaks a standard item master: the same case of product is a different weight and therefore a different cost and price every time, and a system that stores one number per item cannot represent it.

Regulated industries pay for evidence rather than features. Pharmaceutical and medical device work needs a validated installation and an audit trail that survives inspection, which is closer to the discipline in our compliance tools than to anything on a shop floor.

Those environments also make upgrades slow on purpose, and a vendor who pushes quarterly releases is not automatically a good fit.

Garment manufacturing ERP software and its neighbours in apparel have the opposite problem, which is combinatorial rather than regulatory. One style in eight sizes and six colours is 48 sellable items that share one design, one cut plan and one set of subcontractors, and a system without a proper matrix turns that into 48 unrelated records.

What does not change: costing, MRP and traceability are the same three things every manufacturer needs working, and every industry edition is a layer on top of them. A vendor whose industry story is strong and whose costing is weak is selling the wrong half.

Free and low cost manufacturing ERP software, and where it stops

Free manufacturing ERP software is real, and it is free in two distinct senses that get confused in every comparison. One is an open-source licence, the other is a limited tier of a commercial product, and they stop in different places.

The open-source licence

ERPNext is the open-source case: the licence is AGPL-3.0 and costs nothing, self-hosting is permitted, and Frappe Cloud starts at $5 a month per site for shared hosting, $20 for a virtual machine and $125 for a dedicated server, read 16 September 2026. What it does not include is somebody to own the upgrade path.

The limited free tier

Katana’s free plan is the limited-tier case: $0 for 30 SKUs with unlimited users, which is a genuine starting point for a workshop and a hard ceiling for anyone else. A limited tier is useful for learning whether the data model fits before money moves, and the same logic applies across software for small business generally.

Low cost manufacturing ERP software has a predictable failure mode, and it is not missing features. It is that the cheap tier holds one legal entity, one location and one currency, and the second site is what forces the migration you were avoiding.

Read the limits page before the feature page, the same way you would when comparing anything in analytics tools where row and event caps decide the real price.

What breaks during the move

ERP migrations do not fail on software. They fail on data that was never as clean as anybody believed, and on the assumption that two systems can run in parallel without doubling the work.

Bills of materials are the worst of it. Nobody has a current, complete, multi-level BOM for every active product; they have drawings, a spreadsheet, and one person who knows the substitutions. Cleaning that up is a project of its own and it belongs before go-live rather than during.

Nobody arrives with a current multi-level bill of materials for every active product. Cleaning that up is the single most common reason an ERP go-live date moves, and it is work that has to happen before the migration rather than during it.

Opening inventory has to be counted, not exported. Starting balances carried over from a system whose quantities were already wrong produces a new system that is wrong on day one, and the trust never recovers.

The costing method is a decision, not a setting. Moving from actual to standard costing changes reported margins on every product, and the finance team needs to agree on that before the first period closes rather than argue about it afterwards.

Integrations are where the timeline goes. Accounting, ecommerce, shipping and carrier tools and the label printer on the packing bench all have to keep working, and each connector is a small project with its own owner.

Dual running costs more than people plan for. Two to three months of entering every transaction twice is the honest estimate, and it needs staffing rather than goodwill.

Cloud based manufacturing ERP software removes the server question and none of the above. What it does change is who owns the upgrade calendar: the vendor decides when the version moves, and a validated or heavily customised plant has to be ready for that rhythm.

Cloud or on-premise, in 2026 terms

Cloud is the default and the exceptions are specific rather than sentimental. Two of them are real: a validated environment where every change needs requalification, and a plant whose connectivity genuinely cannot be trusted with production reporting.

	Cloud, vendor hosted	Self-hosted or on-premise
Upgrades	Vendor schedule, little say	Your schedule, your effort
Cost shape	Subscription, predictable	Capital plus staff, lumpier
Customisation	Extensions within limits	Anything, including regret
Offline shop floor	Needs local buffering	Works on the local network
Who fixes it at 2am	The vendor, under an SLA	Whoever has the credentials
Realistic fit	Almost everyone	Validated, remote or heavily bespoke

A middle road exists and is underused: vendor-hosted ERP with local buffering on the floor terminals, so a dropped connection queues operations rather than stopping production. Ask how that queue behaves when it reconnects, because an MRP run fed by a two-hour backlog of operations is not the same as one fed live.

What to check in the trial

A trial that only proves the software installs is wasted. Five tests separate manufacturing ERP software that fits from software that demos well, and all of them use your own data rather than the vendor’s sample company.

• Load one real product, fully. A multi-level BOM with a substitution, a routing with setup time, and a work centre with capacity. If the import needs a consultant, that is the answer to a different question you were going to ask later.
• Run MRP with a deliberate shortage. Short one purchased component and see what the system proposes, and whether it explains why.
• Report a partial completion and a rework. Nothing exposes a thin work order model faster than half a batch finished and two units sent back.
• Close a costing period and look at variance. By operation, not by product. A tool that cannot show where the money went is not a manufacturing system.
• Trace a lot in both directions. From a finished case back to the supplier’s delivery, and from that delivery forward to every case it touched. One query, under a minute, in front of somebody who is timing it.

MRPeasy publishes a 15 plus 15 day trial with no credit card, which is enough for all five tests if the BOM data is prepared first. Most vendors at the mid-market end will extend a trial when the plan for it is specific, and the plan above is specific.

What to ask the vendor before signing

Six questions, all of which have a wrong answer that you want to hear early.

• Which seat types exist and what does the cheapest one allow? This decides whether the shop floor reports in real time or at the end of a shift.
• What does the price do at renewal, in writing? A three-year term with an uncapped uplift at the end is a different deal from the one on the slide, and renewal is one of the things a pricing page keeps quiet about.
• Who owns the integrations after go-live, and at what rate? Connectors break when either side updates, and the answer is a name or a support tier.
• How does the upgrade calendar work, and can it be deferred? Relevant for regulated plants, irrelevant for most others, and worth knowing which you are.
• What does implementation cost, as a range, with the assumptions that produced it? Katana publishing a $2,000 onboarding package is the exception; most of this market quotes per engagement.
• Can the data leave, in what format, and does that include BOMs and history? An export of the item master is not an exit.

Comparing those answers side by side is the point of keeping a shortlist in one place. That is what our ERP catalogue and the wider product directory are for, and every outbound link there is nofollow because nobody pays for position.

Two more shortlists usually turn out to be necessary rather than optional. Field operations end up needing their own tools, which is why field service software sits apart from ERP, and the customer-facing half of the business almost never fits the ERP’s own contact table, which is the subject of what CRM software actually is if that part is also on the list.

Questions people ask about manufacturing ERP software

What is manufacturing ERP software?

It is business software that combines standard ERP functions, which means accounting, purchasing, sales and inventory, with production planning built on a bill of materials. The distinguishing parts are MRP, routings and work centres, work orders, shop floor reporting, lot or serial traceability, and costing that reports variance by operation rather than only purchase price.

How much does manufacturing ERP software cost?

Published list prices in September 2026 run from $0 for an open-source licence to roughly $2,750 a month for 25 full users on a per-seat product.

The unit matters more than the rate: per-seat vendors start near $49 to $110 per user per month, flat-rate vendors sit near $299 plus add-ons regardless of headcount, and consumption-priced vendors publish nothing and quote on transaction volume.

Is there genuinely free ERP software for a manufacturing company?

Yes, in two forms. ERPNext is licensed under AGPL-3.0 and can be self-hosted at no licence cost, with managed hosting from $5 a month per site, and several commercial vendors publish a free tier, such as Katana’s $0 plan for 30 SKUs. Both are real starting points, and both trade money for either technical capacity or a hard ceiling.

What is the difference between manufacturing ERP and inventory software?

Inventory software tracks quantities and locations; manufacturing ERP explodes a bill of materials into demand and schedules it against capacity. The practical test is the MRP run: if the system cannot net demand against stock and open orders, or cannot tell you that two work orders want the same machine hour, it is inventory software with a manufacturing label.

Does industry-specific manufacturing ERP software matter, or is it marketing?

It matters where the unit of measure or the compliance surface changes, and it is marketing everywhere else.

Catch weight in food, potency in chemicals, size and colour matrices in apparel and validated environments in medical devices are all things a generic item master cannot represent. Costing, MRP and traceability are identical across industries, so an industry edition with weak costing is the wrong purchase.

How long does a manufacturing ERP implementation take?

Small deployments with clean data and no customisation run three to four months; mid-market projects with integrations and industry requirements run six to twelve. The variable is almost never the software. It is how complete the bills of materials are, how many integrations have to keep working, and whether anyone has been given the time to count opening inventory properly.

Should a small manufacturer buy cloud or on-premise?

Cloud, unless there is a specific reason not to. The two real exceptions are a validated environment where every version change needs requalification, and a site whose connectivity cannot be trusted with live production reporting.

Everything else that used to argue for on-premise, including customisation and data ownership, is now handled by extensions and export terms that belong in the contract.

What breaks most often after go-live?

Integrations and reporting discipline. Connectors fail quietly when either side updates, so somebody has to own them by name. Reporting discipline fails when floor seats were too expensive to license, at which point operations get entered hours late and every planning number derived from them is stale in a way that nobody notices until a promise is missed.

Can manufacturing ERP software replace a CRM?

Partly, and usually not well. ERP holds the customer as an account that owes money, with orders, credit limits and shipping terms, which is enough for a business whose selling happens through a handful of long-standing accounts. Pipeline, activity history and forecasting are a different data model, and teams that sell actively end up running both.

What does an MRP run actually do?

It takes demand, which is sales orders and forecast, explodes it through every level of the bill of materials, subtracts what is already in stock and already on order, and proposes purchase and production orders with dates. The output quality depends entirely on input accuracy: wrong lead times or stale stock figures produce a confident plan that cannot be built.

How many users should be licensed at the start?

Everyone who enters or approves a transaction, plus the floor reporting seats, and nobody else. Read-only tiers exist for the rest, such as Business Central’s $8 Team Members seat, and viewer licences are the cheapest way to stop the reporting workarounds that ruin data quality. Adding seats later is easy; removing them mid-term usually is not.

Where do vendor prices come from, and how current are they?

Every figure in this article was read on each vendor’s own published pricing page on 16 September 2026, and none of it came from an aggregator.

Prices in this market change more often than the pages suggest, mid-market ERP is frequently discounted against list, and anything published without a date should be treated as a rough guide rather than a quote. Our own software catalogue records what vendors publish, and nothing else.