Begin with the job, not a feature list.
Write down one complete task: who starts it, which information it needs, the systems it touches and what proves it is finished. Include an exception, not just the happy path.
A vendor can check many feature boxes and still force the team to coordinate the real work manually. Conversely, an unglamorous existing product may solve the task well without a custom build. The comparison needs to reflect the way the business operates.
Compare three paths.
Configuration and integration are separate options from a full replacement. A mixed approach can keep a useful system of record while adding the workflow it lacks.
| Path | A useful fit | What to check |
|---|---|---|
| Configure an existing tool | The workflow is standard and the tool supports it | Licensing, usability, permissions and reporting |
| Connect tools | Each tool works, but the handoff does not | Record ownership, APIs, sync failures and maintenance |
| Build custom software | A valuable requirement remains unmet | Ownership, support, adoption and total operating cost |
Count the cost of the operating model.
Compare implementation, licenses, integrations, migration, training, hosting, support and future changes over the same period. Include the people who will own the system after launch.
For an existing product, investigate the cost of the plan that actually includes the required features. For a custom system, ask which third-party services remain recurring costs and how the source code, data and hosting can be transferred. “You own it” is not a complete operating model.
Test the hardest requirement before committing.
Choose a representative task that a promising option might fail: an unusual permission, a messy import, a client review loop or an exception that crosses systems. Test it with real users and redacted data.
A prototype can answer whether a workflow is usable. A technical spike can answer whether a connection works. Neither is evidence that the entire system is ready for production. Define what the test proves and what remains unknown.
Use a simple decision memo.
Record the preferred option, alternatives considered, expected benefit, principal uncertainty and conditions that would change the decision. Name an owner for the first release and for ongoing support.
If the business cannot identify a valuable difference that custom software would create, buying or configuring may be the better answer. If the work is genuinely distinctive, a focused custom platform can be worth exploring.
- One representative workflow and one difficult exception
- A like-for-like cost comparison
- An adoption and support owner
- A small test before the large commitment
