Working with us
How to choose a software development partner
A practical framework for evaluating development agencies and contractors, the questions that actually predict a good outcome, the warning signs worth walking away from, and how to structure the first engagement.
Key takeaways
- Portfolios are marketing. Ask instead what they turned down recently and why, the answer tells you whether they have judgement or just capacity.
- Insist on knowing exactly who will do the work. Bait-and-switch between the sales team and the delivery team is the most common failure in agency engagements.
- Structure the first engagement small: a paid, scoped piece of work with a real deliverable, before committing to anything long.
- Ownership of code, infrastructure and accounts should be unambiguous and in writing before you start.
- The strongest signal is a partner who says no to something, or tells you a cheaper way to get what you want.
Most guides on this subject tell you to check portfolios, read testimonials and compare quotes. All three are close to useless. Portfolios show the successes, testimonials are collected from the happy end of the distribution, and quotes measure how confidently someone will guess at a number.
Here is what actually predicts a good outcome.
Ask what they turned down
The single most informative question: what work have you declined recently, and why?
An agency with judgement has a real answer, usually with mild regret attached, a project that was underfunded for its ambition, a client who wanted something they could have bought off the shelf, a technology bet they thought was wrong. An agency with only capacity to sell will struggle, because the honest answer is "nothing".
You are hiring judgement. This question tests it directly, and it is hard to prepare for.
Find out who is actually writing the code
The most common way agency engagements go wrong has nothing to do with technical skill. You are sold by senior people and delivered to by whoever is available.
Ask directly:
- Who specifically will work on this? Can I speak to them before signing?
- Are they employed by you or subcontracted?
- What else are they working on during this period?
- What happens if they leave mid-engagement?
Reasonable answers exist for all of these. What matters is that they are answered specifically rather than deflected into reassurance about "the team".
Ask to see a handover, not a demo
Anyone can demo a working application. Ask instead to see a repository they handed to a client, and, if the client agrees, to speak to the engineer who inherited it.
That conversation tells you everything: whether the tests were real, whether the documentation was written as they went or hastily assembled at the end, whether the team could be productive in a week or spent a month reverse-engineering.
Handover quality is the part of the work that is invisible while you are buying and expensive forever afterwards.
The questions that surface risk
"What would make this project fail?" Anyone who says nothing has not thought about it. A good answer names two or three specific risks, unprompted, including ones that are your fault rather than theirs.
"What's the cheapest version of this that would still be useful?" You are testing whether they will help you spend less. Note that this is against their immediate interest, which is exactly why the answer is informative.
"How do you handle scope changes?" You want a process, a written change with a price attached, decided explicitly. Both "we're flexible" and "the contract is the contract" are wrong answers.
"What are you not good at?" Specific limitations are a strong signal. "We're generalists" is not an answer.
Structure the engagement to limit damage
However good the evaluation, you are still guessing until you have worked together. So make being wrong cheap.
| Stage | Scope | What you learn |
|---|---|---|
| Scoping engagement | 1–2 weeks, paid | How they think, how they write, whether they push back |
| First milestone | 3–6 weeks, fixed scope | How they deliver, communicate and handle problems |
| Ongoing | Milestone by milestone | Whether it keeps working |
A paid scoping engagement is the highest-value thing you can buy from a prospective partner. You get a specification and an architecture, useful regardless of who builds it, and you find out how they work before the money gets serious.
Be wary of anyone who wants to skip straight to a six-month contract, and of anyone who offers to do the scoping free. Free scoping is sales, and it is priced accordingly: the output is optimistic.
Settle ownership in writing
Before anything starts:
- Code in your repository, your licence, from the first commit
- Infrastructure in your cloud accounts, not theirs
- Domains, DNS and third-party accounts registered to you
- Anything proprietary they intend to use, named explicitly, with the terms for leaving
The failure mode is not malice. It is default arrangements that were convenient at the start and become leverage at the end. A partner who has done this before will have clear answers ready.
Warning signs
- A quote produced without asking substantive questions
- Estimates with no range and no stated assumptions
- Reluctance to name the engineers
- A proprietary platform you would have to keep paying for
- Enthusiastic agreement with every idea you float
- Pressure to sign before a small engagement can be run
- No opinion about what to cut
None of these are individually fatal. Two or three together usually are.
What to expect from a good one
They will narrow your scope. They will raise problems early, when it is awkward. They will be specific about what they do not know. They will hand you something your own team can own, and they will not be needed afterwards.
That last point is worth dwelling on. A partner whose commercial model depends on you being unable to leave will make decisions accordingly, and you will not notice which ones until much later.
Frequently asked questions
What should I ask a software development agency before hiring them?
Ask who specifically will write the code and whether you can speak to them. Ask what they recently declined to build and why. Ask to see a codebase they handed over, and to talk to the team who took it on. Ask how they handle a change of scope mid-project, and what happens if they are late. The answers to those five questions predict outcomes far better than a portfolio does.
How do I know if a development agency is any good?
Look for evidence of judgement rather than evidence of output. Good partners narrow scope rather than expanding it, raise risks early even when it is inconvenient, and are specific about what they are not good at. An agency that agrees enthusiastically with everything you propose is selling, not advising.
Should I hire an agency or freelancers?
Freelancers are usually better value for a well-defined, self-contained piece of work where you can provide technical direction yourself. An agency is worth the premium when you need a team, when you need continuity if someone leaves, or when you need someone else to own the architecture. If you cannot technically evaluate the work yourself, the agency's accountability is a large part of what you are buying.
How much should a first engagement cost?
Small enough that being wrong is survivable. A scoping engagement or a single well-bounded milestone, typically a few weeks, lets you see how a team actually works before you commit to a quarter of spend. Any partner unwilling to start small is telling you something.
Who should own the code an agency writes?
You should, unambiguously, from the first commit. That means your repository, your cloud accounts, your domain, your licence. Be specific in writing about anything they retain rights to, and be wary of proprietary frameworks or hosted components that would make leaving expensive.
References & further reading
- [1]
- [2]Martin Fowler, Fixed Price Contracts
martinfowler.com
- [3]Martin Fowler, Conway's Law and team organisation
martinfowler.com
Related resources
Hiring engineers vs outsourcing: how to decide
In-house team, agency, contractors or a mix, a decision framework based on what you are actually optimising for, and the hidden costs on both sides that founders discover late.
How much does it cost to build an MVP?
An honest breakdown of what a first product version costs in 2026, what actually drives the number up or down, and how to tell a realistic quote from an optimistic one.
Technical due diligence: what to actually check
A practical checklist for assessing a codebase and engineering team before an acquisition, an investment, or taking over a product, including what has changed now that much of the code was AI-generated.
Turn the trend into shipped software
I contract as a senior fullstack and data engineer, building the production systems your customers and your reporting depend on.