Find the first useful problem
Follow one week of work: enquiries, orders, stock movements and approvals. Record who enters information, where it is copied and how completion is checked. A useful first project has an owner, a clear input and a measurable result. Compare staff time, corrections and delays before deciding what to replace.
Compare the same period and scope
A build price and a monthly subscription are not directly comparable. Use the same planning period for both options. Include setup, migration, integrations, training, hosting, support, usage charges and future changes. Custom software has operating costs too. A subscription can be good value when the workflow fits and data is easy to export.
Keep what already works
Your accounting, email or payment tool may not need replacement. Check whether a supported integration can remove duplicate work while preserving existing records. Confirm API access, rate limits, export formats and provider charges. A vendor saying it has an API does not prove that your required operation is supported.
Agree ownership and the exit route
Ask who controls the hosting, domain, source code and administrator accounts. Specify third-party licences, documentation and export formats in the proposal. Test whether your team can retrieve useful records, not just receive a database file. Define what happens when maintenance ends or the provider changes.
Pilot before expanding
Select a small group, representative data and a way to return to the previous process if needed. Compare the agreed baseline with the pilot: completion time, correction rate and overdue work. Resolve exceptions before adding departments. A successful demo is the start of evaluation, not evidence of business-wide adoption.