Most CRM cost articles answer a question nobody asked. They quote per-user list prices, add a vague "implementation costs extra" line, and stop exactly where the budgeting problem starts. The licence is the part you can already look up. The part that decides whether the project lands is everything around it.

So this article does the opposite. Below is a full line-item budget for a mid-market CRM rollout, with the hours behind each figure, the assumptions that generate them, and the costs that turn up in month four when they were not in anyone's spreadsheet in month one.

The model behind these numbers

Every figure here comes from a stated set of assumptions. Change the assumptions and the numbers move, which is the point: this is a model you can run against your own scope, not a price list.

• Company shape: 120 CRM seats, of which about 45 are sales, 30 service, 20 marketing and the rest management and read-only.
• Blended services rate: USD 110 per hour. That is a mixed onshore/offshore delivery team. A fully onshore team in a major market typically runs USD 150 to 220; a fully offshore team USD 45 to 85.
• Source systems: one legacy CRM plus a spreadsheet estate and one billing system that must stay in sync.
• Records migrated: roughly 180,000 across accounts, contacts, opportunities and historical activity.
• Integrations: four in scope at launch.
• Licence assumption: USD 65 to 150 per user per month depending on edition. Confirm current pricing with the vendor; published tiers move often and enterprise discounts at this seat count are real.

Year-one line-item budget

The table below splits the build into workstreams. "Lean" is a disciplined rollout with tight scope. "Typical" is what most mid-market projects actually land on. "Heavy" is a regulated or multi-entity business with legacy complexity.

Two things in that table matter more than the totals.

First, services outweigh licences in every column once you strip out the add-ons. The software is the cheap part of a CRM programme. The expensive part is reconciling how eleven people describe the same sales process and then encoding one version of it.

Second, data migration and integrations together are 30 to 35 percent of the services budget. Those two lines are also the ones most likely to be underestimated in a proposal, because they are the ones a vendor can least accurately scope before seeing your data.

What each line actually buys you

Discovery is not a formality

The deliverable is a written process map with named owners and decision rules, signed off before configuration starts. Skipping it does not save the 120 hours. It moves them into the build phase, where changes cost two to three times more because they land on top of work already done.

A useful test: if discovery produces no disagreements, it was not deep enough. Every mid-market business has at least two teams who believe they own the same handoff.

Data migration is a data quality project wearing a migration costume

The extraction and load is the easy half. The expensive half is deciding what happens to the 14,000 contacts with no source, the accounts that exist three times under slightly different names, and eleven years of activity history that nobody will read but everyone is afraid to delete.

Budget for at least three migration rehearsals into a sandbox. Teams that budget for one always end up running three anyway, just without the time allocated for it.

Integrations are priced by auth and edge cases, not by count

"Four integrations" tells you almost nothing. A supported native connector between two mainstream SaaS products might be 8 hours. A bespoke two-way sync with a twenty-year-old ERP that authenticates over a VPN, has no webhooks, and needs conflict resolution rules can be 200. Ask for each integration to be quoted individually with its direction, trigger mechanism and failure behaviour named.

Automation build is where the return lives

Configuration makes the CRM usable. Automation makes it worth paying for: lead routing that hits an SLA, quote approvals that stop leaking discount, service escalations that fire before a customer complains. This is the line most likely to be cut under budget pressure and most likely to be regretted. If you need to trim, trim scope elsewhere and keep the automation that removes manual handoffs. We cover the failure pattern here in why marketing automation projects fail at implementation rather than strategy.

Training is a change budget, not a video budget

Recorded modules are cheap and largely ineffective on their own. What changes behaviour is role-specific live sessions, manager coaching on the new reports, and a named internal champion per team who has time formally allocated. If your training line is under 5 percent of the services budget, you are funding a launch, not an adoption.

The six costs that blow CRM budgets

1. Source data worse than described. Discovered in week five, priced in week one at zero. Mitigate with a paid data audit before the SOW is signed.
2. "Standard" integrations that are not. Any system older than a decade, or any system with a customised schema, should be assumed bespoke until proven otherwise.
3. Process redesign smuggled in as configuration. When discovery reveals the current process is broken, fixing it is a separate project with its own budget. Naming that early is cheaper than absorbing it silently.
4. Reporting requirements gathered late. Executives ask for their dashboard two weeks before go-live, and the data model does not support it. Gather reporting requirements during discovery, because they constrain the data model.
5. No sandbox discipline. Building in production feels faster for three weeks and then costs a month.
6. Zero budget for the humans. The team still has a day job during rollout. Backfill, overtime or slipped targets are real costs whether or not they appear in the project budget.

Year two and three: the shape changes

Year-one thinking causes bad decisions because the cost profile inverts once you are live.

Three-year total cost of ownership on a typical mid-market rollout therefore sits close to USD 890,000 — roughly 2.4 times the year-one number, and about 55 percent licences over the full period. Anyone comparing CRM options on year-one cost alone is comparing the wrong number.

The internal admin line deserves attention. A CRM without an owner degrades quickly: fields proliferate, adoption drops, reporting stops being trusted. Half an FTE is the realistic floor at 120 seats.

How to pressure-test a quote you have received

Take any implementation proposal and check it against this list. Every "no" is a number that will find you later.

• Are hours shown per workstream, not just a project total?
• Does the data migration line state an assumed record count and a number of rehearsals?
• Is each integration listed separately with direction, trigger and auth method?
• Is there an explicit UAT allowance, with who runs it and for how long?
• Is hypercare a defined period with a defined response time, or the word "support"?
• What is the change-request rate, and what counts as a change?
• Who owns the data model after go-live, and is knowledge transfer scoped?
• What happens to the price if record volume comes in 50 percent higher than assumed?

A vendor who answers all eight without hesitating is usually more expensive on paper and cheaper in practice.

Where you can genuinely spend less

Cutting cost is not the same as cutting scope badly. These four reductions rarely hurt the outcome:

• Migrate less history. Two years of closed activity plus full open-record data covers most real needs. Archive the rest somewhere queryable and cheap.
• Sequence integrations. Launch with the two that block daily work. Add the rest in a phase two with better information and less schedule pressure.
• Buy the edition you use. Mixed-tier licensing across roles is often available and frequently ignored. Read-only managers do not need full seats.
• Keep reporting out of the CRM if you already have a warehouse. Duplicating a semantic layer inside a CRM is an expensive way to get a second version of the truth. See the data plumbing nobody budgets for.

What almost never works: cutting testing, cutting training, or cutting discovery. Those three are where mid-market CRM projects go quietly wrong.

When the answer is not a CRM implementation

If your pipeline problem is that nobody agrees what a qualified lead is, a CRM will document the disagreement in a more expensive format. If your service problem is understaffing, workflow automation will route tickets faster to people who still cannot answer them.

Fix the process question first, on paper, cheaply. Then implement. That sequence has never once made a project cost more.

If you want a view of how this budget compares to running the work with an internal team, we ran the same exercise in in-house versus agency versus hybrid, costed per outcome. Our own delivery approach for this kind of work is described on the enterprise software and business automation pages.