Choosing a software development agency is a high-leverage decision with a low-information market. Every agency has a polished portfolio, a process diagram, and a case study that may or may not resemble your problem. The differences that matter show up in the second month: who writes the code, how decisions get made, and what happens when the first estimate is wrong.
These are the questions we would ask if we were buying. They are also the questions we are comfortable being asked.
Ten questions that matter
1. Who actually writes the code?
Ask for the names, roles, and time allocation of the people on your project. If the sales team is the only team you meet before signing, assume the delivery team will be whoever is available. A senior team you meet is worth more than a logo wall.
2. Can we meet the technical lead?
The technical lead is the person who will make or break the project. Meet them before you sign, and ask how they handle disagreement about architecture, scope, and timelines.
3. Show me something you built and still maintain.
Anyone can show a launch. Ask what happened eighteen months later: how it is monitored, who responds to incidents, and what they would rebuild. Maintenance reveals engineering character.
4. How is discovery priced?
A free "discovery" is usually a sales call with a proposal attached. Real discovery — interviews, technical spikes, a scoped plan — costs money. Be suspicious of a detailed fixed quote for a complex product produced without discovery.
5. What is your definition of done?
Does done mean "merged" or "deployed, tested, documented, and monitored"? Ask for the checklist. A team with a written definition of done is a team that has shipped enough to know where quality leaks.
6. Who owns the code and the IP?
You should own the repository, the code, the design files, and the infrastructure from day one. Confirm that third-party licences and contractor IP are assigned to you in writing. Ask for repository access before the project starts, not after it ends.
7. How do you communicate progress?
Look for a weekly demo of working software, a visible backlog, and a written status update that names risks. "We'll send a report monthly" is a warning sign for a six-week project.
8. How do you handle AI and data privacy?
If the work involves AI features or customer data, ask where data is processed, which vendors are involved, whether prompts and outputs are logged, and how the team evaluates quality. Vague answers here become compliance problems later.
9. What happens if the budget runs out mid-build?
Every project has a moment where scope meets reality. Ask what gets cut first and how the team decides. A good partner has a prioritization framework; a bad one asks for more money and blames the requirements.
10. What is the plan if you disappear?
The bus factor question. Ask for documentation, architecture decision records, and a handover process. If the answer is "you won't need to worry about that", worry.
Red flags
- A detailed estimate before discovery. It is either a guess or a sales tactic.
- No named team. Staff augmentation dressed up as an agency engagement.
- "We can build anything." A team without a point of view will not push back when you are about to make an expensive mistake.
- No code review or CI in the process. Quality is not a phase at the end.
- Reluctance to share repositories or documentation. Lock-in by opacity is still lock-in.
- Case studies with no outcomes. A launched app is not a result. Ask what changed for the business.
How to compare proposals
Three proposals rarely describe the same thing. Normalise them before comparing price.
| Dimension | What to compare | Why it matters |
|---|---|---|
| Team | Named roles, seniority, allocation | A cheaper team with no senior lead is not cheaper. |
| Scope | What is explicitly excluded | The exclusions reveal the real cost. |
| Discovery | Paid or unpaid, deliverables | Free discovery usually means no discovery. |
| Timeline | Milestones with working software | Demos are harder to fake than status reports. |
| Ownership | Code, IP, infrastructure, data | You should not need permission to use your own product. |
| Aftercare | Warranty, SLA, handover | Launch is the start of the expensive part. |
What happens after launch
The best signal of an agency's quality is not the launch party. It is the month after: how fast incidents are resolved, how clearly changes are communicated, and whether the team is still improving the product. Ask for the post-launch plan in writing — warranty period, monitoring, support hours, and the handover checklist.
If you want to see how we structure that, read about our services or book a discovery call. We will tell you if we are not the right fit, and we will explain why.
Frequently asked questions
How much does it cost to hire a software development agency?
It depends on scope and geography, but a useful discovery phase typically costs $10,000–$30,000, and a production web application commonly starts around $75,000–$150,000. Fixed-price quotes below that for a custom product usually hide a large amount of scope.
Should I choose a fixed-price or time-and-materials contract?
Fixed price works when the scope is genuinely known and unlikely to change. Time and materials works better for evolving products. Many successful engagements use a fixed-price discovery followed by time and materials for the build, with a clear weekly budget cap.
What should be in a software development contract?
At minimum: ownership of code and IP, a definition of done, payment milestones tied to deliverables, confidentiality and data-processing terms, a warranty period after launch, and a clear exit plan including repository access and handover documentation.
How do I evaluate an agency's technical quality?
Ask to see a codebase or a recent pull request, not just a portfolio. Look for tests, clear commit history, documentation, and evidence that the team maintains something it built. Ask how they handle code review, CI, and security.