Ask three ERP vendors for a quote and you get three numbers that look comparable and are not. Each one is a licence figure plus an implementation estimate, and the implementation estimate is the half that will move. Everyone in the room knows it will move. Nobody can say by how much, so the decision is made on the half that will not.
There is a better way to think about the cost, and it starts with an uncomfortable fact: the implementation half is mostly not about the vendor. It is about you. Four things in your own business set most of it, and every one of them can be counted before a single vendor is invited in.
The enemy is a quote that prices the wrong half.
An ERP implementation has two kinds of cost. The software half is licences, hosting and the vendor's own configuration hours. It is quoted, it is comparable across vendors, and it is the smaller part of what you will spend.
The services half is everyone's time deciding how the business actually works so the system can be configured to it. Which customer record wins when two disagree. What happens to an order that ships short. Who may approve a credit and above what amount. Every one of those is a decision, every decision takes people away from their jobs, and every decision that has not been made when the project starts gets made under deadline pressure, in a change order.
We have written about the line items nobody budgets for when leaving an old suite, and about the readiness conditions that decide whether an implementation succeeds. This post is narrower and more useful before you shop: the cost itself is predictable from four counts, and you can take them yourself.
The four counts that set the cost.
Here is what actually drives the services half, in the order we look at them.
How many systems have to connect to it.
Count every system that will exchange data with the ERP: the CRM, the e-commerce platform, the warehouse or 3PL, payroll, the bank, the shop floor, the field service tool. Each one is an integration with two ends, a mapping, an exception path and a test. The count is the single best predictor of implementation cost we know, and it is a number you already have.
How many exceptions are written down.
Walk the order-to-cash and procure-to-pay paths and count the places where the process departs from the happy path: the partial shipment, the disputed invoice, the customer on hold, the rush order that skips credit check. Then count how many of those have a written rule with a threshold and an owner. The gap between the two counts is work the implementation will do for you, at implementation rates, in production. Closing it beforehand is the cheapest cost reduction available.
How clean the master data is.
Take the customer list, the item list and the vendor list. Count duplicates, records with no owner, fields that are blank on more than a fifth of records. Each of those becomes a migration task, and migration tasks are where implementations go over, because the data is discovered to be worse than anyone claimed after the schedule is set. A weekend with the three lists and a deduplication rule changes the quote more than a vendor negotiation does.
Who owns each business object.
For customers, items, orders, invoices and payments, name the one system that may write each and the one person accountable for it. Where you cannot, you have found a decision the implementation will have to make, and decisions made mid-implementation are the expensive kind. We have written about how to settle system of record per object; it is the foundation of the count.
Put the four numbers on one page and you have an estimate of the services half that is more honest than any vendor's, because it is built from your business rather than from a rate card and a template.
What this gives you before you ever talk to a vendor.
A number you can defend. The services estimate is derived from counts a finance director can verify, not from a sales engineer's judgement.
Comparable quotes. When every vendor is estimating against the same four counts, the differences between their numbers mean something.
A cost you can reduce before you spend it. Every exception written down, every duplicate merged and every owner named before the project starts is a line that never appears in a change order.
An implementer you can hold to a number. If the counts are known, an estimate that moves is the implementer's problem, not yours.
That last point is the one we would push hardest, because it is the commercial commitment that separates an implementation partner from a vendor. We guarantee our estimates: if we get one wrong, we absorb the difference. We can only do that because we take the four counts first, and it is the reason a fixed-scope ERP Readiness Review exists as a product rather than a sales call.
When the cost is not the question.
If your ERP decision is driven by a dated obligation, a compliance requirement or an entity consolidation with a deadline, the cost is what it is and the counts tell you how much schedule risk you are carrying, which is still worth knowing.
If the four counts come back small, few systems, exceptions documented, data clean, owners named, then your implementation cost really is mostly the software half, and you should negotiate the vendor hard and stop reading articles like this one.
And if you are not sure you need an ERP at all, count first. A business with two systems to connect and clean data may need its systems to talk to each other more than it needs a new one.
How we know this.
We do business systems integration for mid-market operators, and the four counts are the first thing we take in every ERP conversation, because they are what our guaranteed estimate is built on. When an estimate has gone wrong for us, it has been because a count was wrong, never because a rate card was.
Where this goes.
The ERP market is about to get more interesting for a US mid-market operator, with a new entrant committed to arriving this year and incumbents repricing at renewal. None of that changes the arithmetic above. Whichever vendor you evaluate in 2027, the services half of the cost will be set by your systems, your exceptions, your data and your owners, and all four can be improved now, while the vendor question is still open.
The operators who spend this window taking the counts and closing the gaps will get comparable quotes, smaller ones, and an implementation that configures a business it already understands. If you want the four counts taken properly, start with a no-risk discovery. You pay for it only if you proceed, and you leave with the numbers either way.
Know whether your business is ready for an ERP before you sign for one.
The ERP Readiness Review is a 90-minute working session plus a written scorecard across master data quality, documented exceptions, integration scope and data ownership. Fixed scope, no obligation. Or see how we approach it.
Book the ERP Readiness Review
More on the same problem:
Frequently asked questions.
What is a typical ratio of ERP implementation cost to licence cost?
Industry guidance has long put services at one to three times the first-year licence for a mid-market suite, and higher for an operationally complex business. The ratio is a poor planning tool, though, because the services number is set by the business rather than the product. Two companies buying the same product at the same licence can pay very different implementation costs, and the four counts on this page explain most of the gap.
Which ERP implementation costs are internal rather than invoiced?
The time of the people who decide how the business works so the system can be configured to it, which is typically the best operations and finance staff for six to twelve months. Data cleanup done by your own team. Testing. Training. Parallel running of the old and new systems through at least one close. These never appear on a vendor quote and are usually larger than the vendor's services line for a mid-market company.
How can a company reduce ERP implementation cost before signing?
Write down every process exception with a threshold and an owner, merge duplicate customers, items and vendors, name the one system that may write each business object, and inventory every system that will connect with its direction and volume. Each of those is a change order that never happens. The work takes weeks at internal cost and removes the same work at implementation rates, in production, under deadline.
Why do ERP implementation quotes from different vendors not compare?
Because each vendor estimates against its own template and its own assumptions about your business, and the assumptions differ. One includes data migration and another lists it as client responsibility; one counts three integrations and another says marketplace connector. Give every vendor the same four counts and the same written exceptions, and ask which cost items their number covers. Only then do the differences between the quotes mean something.
How does a fixed ERP implementation estimate work?
The implementer counts first: systems to connect, exceptions documented, data quality, owners named. From those counts the people who will do the work build the estimate, and the implementer commits to it, absorbing any overrun that comes from their counts being wrong. The client's side of the commitment is that the counts were honest and the decisions get made. An estimate that is not built from counts cannot be fixed, whatever the contract says.
