How to choose a development agency without getting burned.

Plenty of agencies are good. The problem is that the good ones and the bad ones produce documents that look identical to a founder who has not commissioned software before. Here is how to tell them apart before the money moves.

Read the quote properly

A quote worth trusting is specific about what it excludes. A quote to be nervous about is a large number attached to a list of features.

• Is the scope itemised, with an estimate per item? One total for everything means nobody has thought it through, or nobody wants you to see the assumptions.
• What happens when the estimate is wrong? It will be. The answer should be a change process you agreed in advance, not a conversation in month four.
• Who is actually doing the work? The senior people in the pitch are often not the people writing the code. Ask for names, and ask what percentage of their week you are getting.
• Is testing in the number, or extra? If quality assurance is a line item you can decline, you are being sold a way to make the quote look smaller.
• What happens after launch? Support, hosting, and the bugs found in week one. An agency that has not priced this has not finished thinking.

The contract terms that matter

• Intellectual property assigns to you on payment. In writing, for all code, designs and infrastructure configuration. This is the term that ends deals in due diligence when it is missing.
• The code lives in your repository, in your account, from day one. Not handed over at the end. Day one.
• Cloud and third-party accounts are in your company name, with your billing details and your admin access.
• An exit clause with a handover obligation. What you receive, in what state, and within how many days if you stop.
• No lock-in through hosting. If they host it and you cannot leave, the rate goes up later and you have no answer.

Warning signs before you sign

• A fixed price for a scope nobody could estimate confidently. Somebody is going to lose money, and the contract decides who.
• No discovery phase offered, or one thrown in free. Free discovery is a sales meeting.
• Reference clients who are all still mid-project. Ask to speak to one who finished a year ago.
• The technology stack is the same for every client regardless of the problem.
• Pressure to sign this month for a discount that expires.
• No questions about your business model. If they have not asked how you make money, they are not designing for it.

What good looks like

A good agency pushes back on your scope. They tell you which third of the feature list to cut, they demonstrate working software every fortnight rather than showing progress reports, and they are comfortable with someone technical on your side reviewing their work. That last point is a strong filter all by itself: an agency that resists independent review is telling you something.

Agency, contractors or in-house

1. 1Agency when the work is bounded, you want it delivered rather than managed, and you can afford to pay for coordination as well as code.
2. 2Contractors when you have someone able to direct them daily. Without that, you are buying capacity with no steering.
3. 3In-house when the software is the business and will keep changing forever. It is slower to start and cheaper to run.

Whichever you choose, the failure mode is the same: nobody on your side of the table able to tell whether the work is good. That role costs a fraction of the build, and it is what a fractional CTO does most days.