How to Choose a Software Development Agency: Questions to Ask Before You Sign

Short answer: ask five questions, and listen less to the answers than to their specificity. Who writes the code? Who owns it when we finish? What is excluded from this quote? What happens on the first scope change? What does year one of support cost? Agencies that answer with names, numbers, and written terms are the ones to keep talking to.

What follows is a checklist for deciding whether to sign or walk away. Picking the wrong agency is expensive. Only 50 percent of projects delivered value worth the effort and expense in 2025, and 13 percent failed outright, according to the Project Management Institute's annual study. That risk is structural, not bad luck. You reduce it before you sign, not after.

Before you talk to anyone: write a one-page brief

You do not need technical knowledge. You need a page that says:

• What the business problem is, in numbers (hours lost, orders missed, calls unanswered).
• Who will use the software, and for what.
• A realistic budget band, and what success looks like in six months.
• Your decision deadline.

Send the same brief to everyone you shortlist. A vague brief is how you get vague proposals, and then the real cost arrives later as change orders. The studios that ask follow-up questions about your brief are showing you something already: they read it.

The five areas to vet

1. People and proof. The team you meet should be the team that builds, or know exactly who will. Ask for two live products you can click right now.
2. Process and communication. Weekly demos of working software, a shared task board, and a named person accountable for your project.
3. Money and scope. Written scope with explicit exclusions, a change-order process, and payment tied to milestones.
4. Ownership and exit. Code, designs, and accounts in your name. A handover plan that does not require their permission.
5. References and red flags. Calls with recent clients they did not hand-pick, including one whose project hit trouble.

An agency needs to hold up on all five.

The questions, with the answers to listen for

"Who will actually work on this, and can we meet them?"

Good: you meet the lead developer, and their role is written in the proposal. Bad: "our senior team" with no names. Team substitution was the first red flag in every failed project the author of DesignRush's 2025 vetting guide audited.

"Show me something you shipped that I can open right now."

Good: two links, plus a sentence on what was hard. Bad: slides, mockups, and a portfolio behind an NDA. Working software is the only portfolio that cannot be faked.

"What is included, and what is explicitly out of scope?"

Good: a written list of exclusions. Bad: "everything you need." Without written exclusions, the gaps show up later as extra charges.

"How do change requests work?"

Good: written change orders with cost shown before work starts. Bad: "we're agile, we handle it." The process is your protection.

"Fixed price or time and materials?"

Both are legitimate. Fixed-price contracts typically carry a built-in risk premium of 15 to 30 percent that you pay even when nothing goes wrong, and every change becomes a formal negotiation. Time and materials, ideally with a not-to-exceed cap, usually fits a small business because you pay for work done, and changing direction costs less. If the scope is stable and you value certainty over flexibility, fixed price is fine; just know what you are buying.

"Who owns the code, designs, and the accounts?"

Good: you do, from day one, and it is written in the contract. Bad: any hesitation, any "we license it to you." Also confirm the app store accounts, domains, and cloud accounts are registered in your name. Vendor lock-in is the most expensive thing an agency can do to you, and it never appears in the quote.

"Walk me through the last project where something went wrong."

Good: a specific story with what they changed afterward. Bad: "nothing ever goes wrong." Every real studio has this story on file. The ones that claim otherwise have either not shipped much or are not telling you the truth.

"Can we talk to three recent clients?"

Good: yes, including one from a project with a rough patch. Bad: references withheld, or only the clients in their marketing. Source at least two of the three yourself so you are not hearing a curated set.

"What does support cost after launch?"

Good: a number, in writing. For custom software, maintenance and support typically run 15 to 25 percent of the build cost per year. Ask what that covers: bug fixes, security updates, hosting, and who responds when something breaks on a Friday.

For scale, know what good projects cost. Custom software projects reviewed on Clutch average $132,480 over about 13 months. US firms commonly bill $50 to $99 per hour. If a quote comes in at a fraction of the others for the same scope, ask what it leaves out.

Red flags that end the conversation

• The A-team pitches; a different, junior team delivers.
• Guaranteed dates with no discovery phase, or a detailed fixed price before understanding your business.
• No access to your own code repository during the project.
• Accounts registered in the agency's name.
• Demos that are never quite ready.
• Deadlines slip repeatedly with no root-cause explanation.
• A quote dramatically under everyone else's. At half the market rate, the difference usually comes back as lower quality or as charges added later.

One flag alone can have a good explanation. Two or three together are a pattern.

What good looks like

A partner worth signing asks about your business before pricing it, pushes back on bad ideas, shows something working every week or two, explains trade-offs in plain language, and makes leaving easy because they expect the work to speak for itself. That last one tells you the most.

FAQ

Should I choose a local agency or an offshore team?

Local buys time-zone overlap, contracts in your jurisdiction, and easier communication. Offshore widens the pool of studios you can choose from. Either can work. What matters more than location is the process: weekly demos, written scope, and your name on the accounts.

How do I check references properly?

Ask for three clients from the past year, source one or two yourself, and ask each the same two questions: what was the first scope change and what did it cost, and would you work with them again. Specific answers are a good sign. So is a reference who mentions a problem before you ask about one.

What if we disagree after launch?

That is what the contract is for. Before signing, confirm three things: a bug-fix warranty window after launch, a written support plan with response times, and clear termination terms that let another team pick up the work.

How long should a software project take?

For a basic website, most agencies (71 percent) deliver in under four weeks. Mid-level app builds typically run 6 to 9 months. PMI's 2025 research says half of projects fall short of full success; your defense is weekly demos, so slippage shows up in weeks, not at the deadline.

Next step

If you are shortlisting studios right now, send each one the same one-page brief, then run these questions. If Pixflint is on your list, good. We will answer all of them, including the ones that cost us the deal. Tell us about your project.