How to choose a software development company

Hiring a developer or agency is one of the highest-risk purchases a small business makes. The work is invisible until it's done, the failure mode is "we built the wrong thing," and the lock-in is real. Most owners evaluate vendors the way they'd evaluate a contractor: a couple of calls, a gut feeling, and whoever sounded most confident. That works for a kitchen remodel. It fails for software, because the qualities that make a good sales conversation are nearly uncorrelated with the qualities that make a good build.

This scorecard forces the evaluation onto evidence. Score each criterion from 1 to 5 based on what the vendor actually showed you (real code, callable references, pricing in writing) not what they promised. If you haven't seen it, score it lower. The anchors under each criterion describe what a 1 and a 5 look like in practice, so two people on your team can score the same vendor independently and compare notes.

Why the criteria are weighted

Not every criterion predicts project success equally. Communication and technical depth carry the most weight: 18% each, because nearly every failed engagement we've been called in to rescue traces back to one of the two: the team couldn't build it, or nobody knew it was going wrong until the budget was gone. References and pricing transparency come next at 14%, because they're the criteria you can verify before signing anything. Scope management and ownership terms matter most when something changes mid-project, and something always changes.

The verdict bands mirror the decision rules we use ourselves: 85+ is a strong candidate worth calling references on, 70–84 means dig into the weak spots before signing, 55–69 means the gaps are structural and need contractual fixes, and below 55 you walk away. A partial score is shown as you go, but don't trust the verdict until all eight criteria are scored: an unscored criterion is usually the one the vendor didn't want to answer.

Red flags that override any score

Some signals end the evaluation regardless of the total. A vendor who won't show you any code, won't produce callable references, wants 50% or more upfront, or tells you ownership transfers "after final payment" has told you how the engagement ends. Pressure to sign before you finish evaluating is the same signal. Use the scorecard to compare the vendors who pass those gates, and use the gates to shorten your list before you spend an hour scoring anyone.