Across every band in this guide, roughly 10–15% of the total cost sits in the discovery and scoping phase, where the record model, integration inventory and data quality baseline are established. Skipping or shortening this phase does not reduce total cost; it moves risk into the build phase, where a wrong assumption about a field mapping or a permission rule is far more expensive to unpick.
The remainder splits roughly 55–65% into build (engineering and design), 10–15% into integration work specifically (which we separate from general build because its cost depends on the other system's API, not ours), and 10–15% into testing and hardening before go-live. A CRM handling financial or personal data at any scale needs a dedicated QA pass beyond unit testing: permission-boundary testing, so a user genuinely cannot see records outside their role, and data-integrity testing after bulk import.
A common mistake in early budgeting is to treat "the CRM" as a single line item and reporting, exports, and admin tooling as free additions. In practice, a reporting layer that lets sales leadership build their own dashboards without engineering involvement is a meaningful piece of scope in its own right, and belongs in the estimate from the start rather than being negotiated in as a change request once the team is already three months into the build.