The licence is not the cost of the system

Buyers argue about per-seat price because it is the number the vendor puts in the email. It is also the number that is easiest to compare in a spreadsheet and the number that is least like the bill you will pay.

TCO write-ups of CRM-shaped systems keep landing in the same range: licences are about 20–35% of year-one spend once you add implementation, integration, data work, training, and administration. Salesforce-class deployments are often described as 3–5× the licence line. HubSpot-class stacks tend to land lower, then climb as you cross tiers. None of that is a quote. All of it is a warning not to treat the proposal as the project.

• Licences28% of year-one spend
• Implementation & integration45% of year-one spend
• Internal time & admin27% of year-one spend

Licence share of year one: typically 20–35% once implementation and admin are included (CRM TCO syntheses, e.g. Weaver CRM Graveyard citing SI/analyst ranges). The split below is a planning model for a mid-market CRM-shaped project — not your invoice.

If 55% of CRM projects miss their planned objectives (Johnny Grow, 2025), a large part of that miss is here: the organisation budgeted for seats, then discovered the partner, the data, and the change management were the actual product.

Buy, configure, or build

These are not ideologies. They are cost shapes.

Buy when the job is a known category (CRM, accounting, helpdesk) and three vendors already compete on it. You want their roadmap, their security questionnaire, and the option to leave. You will pay rent forever. That is the deal.

Configure when the category is right and your process is 80% standard. The remaining 20% is fields, workflows, and discipline — not a custom platform. Most “we need it custom” requests in a 50-person company are this, plus a fear of looking unstructured.

Build when the workflow is the business, no vendor owns the job, and you are prepared to own the thing after the agency leaves: hosting, auth, the next change request, the person who understands it. Building a second CRM because you did not like the first one’s demo is how companies collect systems.

How it usually goes

1. 01A gap in the current tool becomes a ‘platform’
2. 02A developer quotes screens, not operations
3. 03No owner named for year two
4. 04The licence you were fleeing is still running, plus a new bill

How it should go

1. 01Name the job that no bought system can do
2. 02Price 3 years of run cost, not 3 months of build
3. 03Buy the commodity around it (auth, email, CRM of record)
4. 04Only then write the piece that is actually yours

Implementation is a second vendor

For Salesforce, Dynamics, SAP, and similar, the delivery partner is a buying decision. Public “implementation packages” are rare; SI ranges for the same logo can differ by 10× depending on who you hire in Dubai versus who you hire from a global bench.

So: get the implementation number in the same round as the licence. Ask what happens when the partner’s project team demobilises. If the answer is “your admin,” make sure that admin exists as a role, not as optimism.

• Projects that miss the planned objective55%
• Projects that hit objective, time, and budget25%

Johnny Grow (2025): 55% missed planned CRM objectives. Cross-source reviews (Gartner/Forrester and later syntheses) keep putting people, adoption, and planning above ‘we picked the wrong logo.’ Treat cause-splits as directional.

A simple test

Write the job on one line. Then:

• If three serious vendors already sell that job, buy.
• If one of those vendors can do it with configuration you can maintain, configure.
• If none of them can, and you would still need the thing if every vendor vanished, build the smallest version, next to a bought system of record — not instead of one.

Re-run the test when someone says “the CRM doesn’t do X.” Often X is training, or a report, or a WhatsApp pipe, not a new product.

What to do this week

• Take your last software proposal and add three lines: partner implementation, internal time (hours × real loaded cost), ongoing admin. If you cannot fill them, you do not have a cost. You have a seat price.
• For any “we should build it” idea, write the three-year run cost before the build estimate.
• If you are mid-stack and unhappy, separate wrong product from wrong implementation. They are different branches: switch vs fix vs rebuild.

The tech check has all three doors — find/switch, optimise, build — because they are not the same project. Picking the door on purpose is most of the work.

Sources

• Johnny Grow, CRM failure report (2025): 55% / 25%
• CRM TCO syntheses (licence as 20–35% of year one; Salesforce-class often 3–5× licence) — planning ranges, not quotes
• Friction vendor-db implementation notes: partner-led delivery, wide SI variance in the GCC