ERP implementation cost guide

Buyer guide

Six things implementation fees actually buy.

Understanding the composition is what lets you tell an inflated quote from a thorough one, because the two look similar on a single line.

Discovery and data assessment

Usually 10–15% of the fee and the phase most often shortened. Data problems were the primary cause in 24% of stalled projects and were almost always found during load rather than during a diagnostic that had nominally happened.

Requirements and design

15–25%. Where requirements are derived from interviews rather than transactions, this phase is cheap and the configuration phase is expensive.

Data migration

20–30%, and the line with the widest variance. Two to five years of history, remapped and reconciled, is genuinely more work than a single opening balance load.

Configuration and build

25–35%. Rises sharply where the answer to every gap is a customisation rather than a process change — the pattern behind 14% of stalled projects.

Testing and parallel running

10–20%, and the first line cut when a project is behind. Cutting it is how a migration reaches go-live without anyone having proven the numbers tie.

Training and handover

5–10%. Frequently underfunded, which is why so many implementations produce a system nobody outside the project team can operate.

Negotiate terms, not rate

Buyers spend most of their effort on day rate. Partners have limited room there and substantial room on terms, and the terms are what actually protect you when something goes wrong.

Named team with a substitution clause. Replacements require your approval and equivalent seniority. Partner capacity and turnover was the primary cause in 8% of stalled projects.
Discovery-failure allocation. Anything the partner should reasonably have found during scoping is theirs to absorb; genuinely new requirements are yours. Write that distinction down.
A data-condition contingency. Agree in advance what happens if the data is worse than the diagnostic found, rather than negotiating it under pressure in month three.
Phase gates with an exit. The right to stop after discovery, having paid only for discovery, is worth more than any rate concession.
Knowledge transfer as a deliverable. Documented and accepted, not assumed. A partner you cannot leave is a dependency you bought without pricing it.

A rate concession saves you a few percent. A phase gate with an exit saves you the whole project when the answer turns out to be no.

Overruns are predictable

Median variance to original implementation quote across our sample is +27%. That is not a reason to distrust partners; it is a reason to budget for it and to ask each candidate what their own historical figure is.

Partners who track that number and will state it are a different kind of firm from those who deflect. “Every project is different” means either they do not measure it or the number is bad, and both are worth knowing before signing rather than in month four.

What actually drives overrun

Almost never the software. In our sample of stalled projects, the software being genuinely incapable was the primary cause 4% of the time — the rarest cause and the most commonly blamed afterwards, because it assigns responsibility outside the building.

The real drivers are requirements that kept moving (31%), data worse than anyone knew (24%), and no internal owner able to decide across departments (19%). Add those and 74% of stalled projects failed for reasons that existed before anyone configured anything.

How to reduce it before you sign

Pay for a real data diagnostic before committing to a fixed price. A week and a few thousand dollars removes the largest single source of overrun.
Derive requirements from transactions rather than interviews.
Name one person per decision area who can decide within a day.
Cut scope to what you need in the first six months. Everything else is phase two, and phase two is cheaper than a delayed phase one.
Do not let testing and parallel running be the line that absorbs a schedule slip.

Our own numbers, for comparison

We quote implementation at roughly 0.25× year-one subscription against a market median of 1.4×. The honest reason is that we do less: no partner margin, a narrower product, and an integration-first approach that defers the migration entirely for many customers. A NetSuite implementation covers considerably more ground than ours. Compare three-year totals against what you actually need rather than comparing the multiples.

Questions

Common follow-ups.

What is a normal implementation multiple?

Median 1.4× first-year licence across our 63-quote sample, with more than 2× variance for comparable scope. NetSuite tends higher, Odoo lower.

How much overrun should we budget?

Median +27% in our sample. Ask each candidate partner for their own historical figure — the willingness to answer is as informative as the number.

What causes overruns?

Requirements that kept moving, data worse than anyone knew, and no internal decision-maker. Together 74% of stalled projects. The software itself is 4%.

Should we pay for a diagnostic first?

Yes. A week and a few thousand dollars removes the largest single source of overrun, and the findings are yours whether or not you proceed.

What term matters most?

A phase gate with an exit after discovery. It is worth more than any rate concession because it caps your exposure when the answer turns out to be no.

What implementation actually costs

Reviewing a proposal?