How much does an ERP cost?

The short answer

For a Swiss company of 10 to 50 people, an ERP costs somewhere between CHF 25,000 and CHF 300,000 in the first year, and the spread is not vendor greed — it is the range of what companies actually ask a system to do. A ten-person services firm with clean data and no stock sits at the bottom. A fifty-person manufacturer with two sites, a warehouse, and twelve years of data in four old systems sits at the top.

What the two have in common: the software subscription is the smallest part of the bill. Licences for SME cloud ERP run CHF 50 to 150 per user per month. The implementation — configuring the system to your business, migrating your data, connecting the tools you keep, training your people — typically costs one to three times the first year’s licences, and then your own team’s hours come on top of that, unbilled and unbudgeted.

This guide takes the bill apart line by line: the four cost blocks, what they look like in numbers for three typical companies, why implementation dominates, the costs that never appear on a quote, how long it all takes, and — because most articles on this topic are written by people selling the implementation — when it is cheaper to wait.

The four cost blocks

Every ERP bill, whatever the vendor calls the line items, is made of four blocks. Two of them are on the quote. Two of them are not.

The percentages are the typical split published by implementers for classic projects. Notice what they imply: the block you can compare on a price list is a fifth of the money, and the block that is easiest to underestimate — your own people — is roughly as large as the entire external implementation.

What it looks like in numbers

Three typical Swiss companies, with the ranges you will actually find in quotes from Swiss and German implementers. These are ranges, not offers — your own numbers depend on the drivers in the sections that follow. They are here so that the first quote you receive has something to be compared against.

Two things stand out. First, the subscription is never the story: even in the largest scenario, a year of licences is a third of the first-year bill. Second, the internal hours are real money. Eight hundred hours of your best people is half a year of a full-time salary — spent not on customers, but on the project.

If you want to sanity-check a quote quickly: independent research on classic ERP projects puts the total first-year cost for mid-sized companies at a low single-digit percentage of annual revenue, and finds that more than a quarter of projects exceed the budget they started with. Plan the contingency — 10 to 15 percent is the usual advice — before you need it.

Why implementation costs more than the software

The software is generic. Your company is not. Between the two sits a translation job, and the translation is done by people who bill by the day.

A classic implementation runs through the same phases everywhere: an analysis of how your business works and a concept of how it should work in the system; the configuration of that concept — master data, workflows, documents, roles, reports; the migration of your existing data; the interfaces to the tools you keep; testing; training; and the weeks around go-live when the old and new worlds run side by side. Every one of those phases is consulting time, and consulting time in Switzerland costs CHF 150 to 250 an hour.

This is also why two vendors with identical per-user prices can be three times apart on the total. The licence measures the product. The implementation measures the distance between the product and your business — and how much of that distance has to be closed by hand.

The uncomfortable corollary: in a classic ERP, every future change to your processes goes through the same channel. A new field, a new approval step, a report the system does not have — each one is a change request, each one is billed. The implementation is not a one-off cost; it is the first instalment of a relationship in which adaptation always has a price.

The costs nobody quotes

A vendor’s quote covers the vendor’s work. These are the costs that land on your side of the table — predictable, well documented, and absent from almost every proposal.

None of these are reasons not to buy an ERP. They are reasons to budget for the project you are actually going to have, rather than the one on the quote.

How long it takes

Time is a cost too — every month of implementation is a month of your best people split between two jobs. The published timelines for SME projects are consistent: two to four months for a small company on a starter package with standard processes, four to eight months for a mid-sized company with stock and interfaces, and eight to fourteen months for manufacturing or multi-site companies. Classic projects in the wider German-speaking market run six to eighteen months.

Where the time goes is instructive. The configuration itself is rarely the long part. The calendar is consumed by the analysis phase — agreeing how the business should work — by data cleaning, by waiting for decisions, and by the weeks of parallel running before anyone dares to switch the old system off. In other words: by the distance between your business and the software, again.

A rule that holds surprisingly well: if a vendor’s timeline is much shorter than the ranges above, ask which of those phases they are assuming you will do yourselves.

What makes a project expensive

Headcount is the number vendors ask for first, and it is the least interesting one. These are the drivers that actually move the bill — in rough order of how much.

Cloud or on-premise?

The cost question most often asked after the price itself. The honest answer is that the software price is the wrong place to look.

Cloud means a subscription per user, no server hardware, updates included, and the vendor carrying the infrastructure and its security. The first years are cheaper and the spend is predictable. Over a decade the subscription may add up to more than a one-off licence — the comparison vendors of on-premise software like to make.

On-premise means a one-off licence plus 15 to 22 percent of it every year in maintenance, servers to buy and replace, backups, security, and somebody — in-house or external — whose job it is to keep it running. Infrastructure alone is put at 10 to 20 percent of a classic project’s total.

For a company of 10 to 50 people, the deciding factor is rarely the licence arithmetic. It is who carries the operating risk, and whether you want an IT function you would otherwise not have. Most SMEs today land in the cloud for that reason alone — and the questions worth asking a cloud vendor are about data location, exit terms, and what happens to your price at the next renewal.

How to read a quote

Seven questions that turn a price into a budget. Any serious vendor can answer them; the ones who cannot are telling you something too.

And one thing to do with the answers: add your own hours, at a real hourly cost, to the vendor’s number. The result is the first figure in this whole process that deserves to be called the price.

When it’s cheaper to wait

This is the section the implementers’ cost guides leave out. An ERP you introduce too early costs everything above and returns almost nothing, because there is not yet enough business to run through it.

Wait if your processes still change every month — locking them into software means paying to configure guesses, then paying again to change them. Wait if the pain is a single bad tool rather than the handoffs between tools; replacing one tool costs a fraction of an ERP. And wait if nobody can yet describe how an order should flow through the company — that conversation costs nothing, and it is the one part of the implementation no vendor can do for you.

But be honest about the other side of the ledger. Waiting has a price too: the hours spent re-typing and reconciling, the errors that reach customers, the decisions taken on stale numbers, the migration that grows with every month of data. We quantified it in When do you need an ERP? — five people losing four hours a week each is over a thousand hours a year. Against the numbers above, that is not a small figure. It is roughly the implementation cost of the smallest scenario, every single year, spent on nothing.

How AI changes the price

Look at the four blocks again. The one that is advertised — the licence — is the small one. The ones that decide the bill are implementation, adaptation and migration: human hours spent translating your business into a rigid system, and moving your data into it. That cost structure is not a law of nature. It is a property of how ERP has been built for forty years.

AI-native ERP attacks exactly those blocks. When the people who use the system can reshape it in plain language — a new field, a new workflow, a new view, without a consultant and without a change request — the implementation stops being a translation project and the running costs stop being a meter that ticks every time your business changes. When migration is assisted by a system that understands what your old data means, the most underestimated line in the project shrinks with it.

This does not make an ERP free. Your data still has to be understood, your processes still have to be decided, and your people still have to learn. What changes is the multiplier: the biggest cost blocks stop scaling with consulting days. And when adaptation is cheap, the two reasons companies wait — the price of the implementation and the fear of cementing their processes too early — both lose their force at the same time.

Frequently asked questions