Buy or build? A framework for EU manufacturers

This is a framework, not a sales page. We build custom software, and a page arguing that everyone should build custom software would be worthless to you. What follows uses published prices from both routes, names the situations where packaged software is clearly the right buy, and gives you a way to reach your own answer without a vendor in the room.

The two routes side by side

Packaged figures are the values shown on each vendor's own pricing page at the time of checking and vary by country and billing term. Microsoft notes its published prices are informational and may not reflect actual list price by region. Implementation, migration and localisation are separate in every column.

Variable one: headcount

Per-user pricing is not a flaw — it is a fair way to charge, and for a small team it is by far the cheapest way to get a full ERP. The point is simply that it scales with your hiring, indefinitely. A manufacturer moving from 10 to 30 licensed users triples the recurring bill without changing anything about the software.

Two published mechanisms soften this and are worth knowing. Microsoft publishes a Team Members licence at $8.00 per user per month for people who read records, approve, or enter limited data — which describes a large share of shop-floor headcount, and getting the licence mix right is often the single largest saving available. Katana publishes Core from $299 per month with unlimited users, removing headcount from the equation entirely at the cost of a higher floor.

The custom route removes it too, by construction: there is no per-user fee and no user cap. Whether that advantage outweighs the higher day-one cost is arithmetic, and the answer flips somewhere between about twenty and fifty users depending on which tier you would otherwise buy.

Variable two: exceptions

Write down every requirement that does not appear in any vendor's published feature list. Not what you would like — what you actually do, today, that the standard model does not represent. Subcontract routing. Mould life. Batch merging rules. Customer-specific quality documentation. Dealer-specific pricing logic.

Then count them:

• Zero to five exceptions. Configure a packaged product and stop reading. The exceptions are absorbable, and building would be an expensive way to reinvent things that already exist.
• Five to fifteen. A judgement call, and the tie-breaker is centrality. If those exceptions are administrative, configure them away. If they are the reason customers choose you, they deserve to be first-class rather than bolted on.
• More than fifteen. You are about to build a bespoke system inside someone else's framework — with the framework's constraints and none of the ownership. At this point, price building it properly and compare honestly.

One caveat that applies at every count: the exception list must be written by the people who do the work, not by the people who buy the software. Exception lists compiled in a management meeting are consistently too short.

Where data residency changes the answer

For an EU manufacturer, hosting location is usually settled by policy rather than preference. The relevant question is not buy-versus-build but where each option can physically run.

With a SaaS product you choose from the regions the vendor offers, and that list is a hard boundary. Odoo publishes three deployment options on its own pricing page — Odoo Online, Odoo.sh and on-premise — which gives more room than a cloud-only product. With a custom-built system, hosting region is entirely your decision, including a server in your own building if that is what your policy or your customers require.

If data residency is a hard requirement, apply it as a shortlisting filter before you compare features or prices. Discovering a residency conflict after selection is one of the more expensive ways an ERP project can restart.

The hybrid pattern

Most manufacturers who ask this question well end up not at either pole but at a boundary. The pattern that works:

1. Buy the compliance core. Accounting, statutory reporting, tax filing, standard purchasing. A vendor amortises legislation tracking across thousands of customers; you cannot match that, and you should not try.
2. Build the differentiating edge. The shop-floor terminal screen operators use a hundred times a day. The scheduling rules specific to your machines. The customer portal. These are where general-purpose software costs you in small daily frictions that never appear on an invoice.
3. Budget the interface deliberately. This is why API access matters so much in tier comparisons: Odoo places its External API in the Custom plan, MRPeasy in its Unlimited tier. Compare at the tier that actually includes it, or your integration budget is wrong before you start.

KodDelta's published bands for the edge: $3,000-6,000 for a single process (2-4 weeks), $8,000-15,000 for a multi-module system (4-8 weeks), $20,000+ for an end-to-end platform, and $5,000-12,000 for a retrieval-augmented AI assistant over your own documents. Maintenance is optional at 12-25% per year.

Test your own situation in three questions

1. How many licensed users in three years? Multiply by the tier you would truly buy, then by sixty months. If that number is smaller than a one-time band plus five years of optional maintenance, buying wins on cost and you can stop there.
2. How long is your exception list, written by the people who do the work? Under five, buy. Over fifteen, price building. In between, ask which exceptions your customers would notice if they disappeared — those are the ones worth owning.
3. Where must the data live, and is that negotiable? If it is a hard requirement, filter on it first. It eliminates options faster and more cleanly than any feature comparison.

With those three answers the decision is usually clear within a single conversation. If they point to a packaged product, we will say so — a project that should not exist helps neither of us.

Sources

• Odoo Pricing — plans, per-user rates and deployment options (odoo.com)
• Dynamics 365 Business Central Pricing (microsoft.com)
• MRPeasy Pricing (mrpeasy.com)
• Katana Pricing (katanamrp.com)
• Acumatica Pricing (acumatica.com)
• European Commission — GDPR adequacy decisions (commission.europa.eu)
• KodDelta pricing and delivery bands

Sources checked in August 2026. Vendors change prices without notice; verify the linked page before deciding. This page is general information, not legal advice. Product names belong to their respective owners.

Frequently asked questions

Is buying always cheaper than building?

On day one, yes — almost without exception. Over five years it depends on two variables: how many people need licences, and how many of your processes fall outside the standard model. Per-user subscriptions grow with headcount forever, while a one-time build does not. Neither answer is universal, which is why the decision has to end in arithmetic.

What do the packaged options actually publish?

Odoo publishes Standard at €11.90 and Custom at €17.90 per user per month billed yearly, with all apps included. Microsoft publishes Business Central Essentials at $80.00, Premium at $110.00 and Team Members at $8.00 per user per month. MRPeasy publishes $49 to $149 per user per month. Katana publishes Core from $299 per month with unlimited users. Acumatica publishes nothing and requires a quote.

How does GDPR affect the decision?

It affects where the system runs more than whether it is bought or built. With a SaaS product, hosting region is the vendor's decision within the options it offers; with a self-hosted or custom system, you choose the region outright. Odoo publishes on-premise as a deployment option alongside its cloud. If data residency is a hard requirement, confirm it before shortlisting rather than after.

What is the threshold where building starts to make sense?

A practical rule from real projects: count the requirements that no published feature list covers. Under five, configure a packaged product. Over fifteen, you are building a bespoke system inside someone else's framework and should price building it properly. Between the two, weigh how central those requirements are to how you win business.

Should statutory accounting ever be custom-built?

Almost never, and we say so at the quotation stage. Tax rules, statutory reporting formats and mandatory e-invoicing regimes change constantly, and packaged vendors amortise that tracking across thousands of customers. Buying legislation compliance is always cheaper than building it. Build the part that differentiates you, not the part that is the same for everyone.

What does the custom route cost?

KodDelta publishes one-time bands: $3,000-6,000 for a single-process solution delivered in 2-4 weeks, $8,000-15,000 for a multi-module system in 4-8 weeks, $20,000+ for an end-to-end platform in 3-6 months, and $5,000-12,000 for a retrieval-augmented AI assistant. Maintenance is optional at 12-25% per year, there is no per-user fee, and source code is delivered under contract.