1. Write a one-page project brief
Explain the business problem, users, required outcomes, known technology constraints, timeline, sensitive data and what success looks like. Do not hide the budget if it is fixed; a realistic range helps suppliers propose an appropriate approach.
2. Shortlist for relevance, not logo count
A provider that has built ten ecommerce stores may not be right for a dispatch platform. Ask for work that resembles your risk profile, integrations or user flow.
3. Interview the delivery people
Speak with the developer, QA lead or operations manager who will actually work on the engagement. Ask them to explain a relevant project, a failure they handled and how they would structure the first two weeks.
4. Verify the environment
For sensitive work, ask about access control, endpoint security, backups, incident procedures, subcontractors, data location and how staff accounts are removed when someone leaves.
5. Run a paid trial
The trial should resemble the real job and produce something useful. Evaluate comprehension, communication, documentation, quality and how feedback is handled.
6. Contract for the end as well as the beginning
Define ownership, confidentiality, data handling, third-party components, payment, acceptance, warranties where appropriate, termination, knowledge transfer and offboarding.
7. Start with a controlled milestone
Do not hand over every credential on day one. Grant the minimum access needed and expand it as trust and requirements justify.
Interview the assigned people, not only the sales team
Ask who will write the code, run the support queue, prepare the reconciliations or manage the campaign. Meet at least the delivery lead and one hands-on contributor before signing a long engagement. If the provider cannot identify the team yet, understand how staffing will be approved later.
Check references against a similar operating model
A reference from a large enterprise may not tell a small founder how the provider handles a five-person project. Ask for examples similar in size, technology, data sensitivity and decision speed to your own work.