Seven questions to ask before you subcontract a build

Agencies sell development more often than they staff it. A brand studio wins a client, the client wants a portal alongside the site, and the studio now has three weeks to find somebody who can build it without embarrassing them. The decision usually gets made quickly and on instinct, which is how it goes wrong.

These are the questions worth asking, in the order they matter. We are on the other side of this conversation regularly, so treat the list as interested rather than neutral — but every question here is one we would want asked of us.

1. Will you approach my client, ever?

The fear that stops most agencies subcontracting is that the developer will eventually deal with the client directly and cut them out. It is a reasonable fear and it happens.

Ask for it in writing, not as reassurance. A partner who will put non-solicitation in the agreement has told you something; one who says "we would never do that" and leaves it verbal has told you something else. Include the period after the project ends, which is when it actually happens.

2. Who writes the code — the person I am talking to?

You are buying delivery you will not supervise. The gap between the person who is persuasive on a call and the person who commits at 2am is where most disappointment lives.

Ask plainly whether the people in the conversation are the people building it. A larger partner will say no, and that can be fine — but then ask who, and ask to speak to them before you commit, not after.

3. What happens when you are busy?

This is the question almost nobody asks, and it is the one that damages agencies most. You have already given your client a date. Capacity that is promised and then missed is materially worse than capacity that was never offered, because you cannot go back and re-plan a quarter you have already committed.

What you want to hear is that you will be told early and honestly, including being told no. A partner who has never turned work away is either much larger than they appear or is about to disappoint somebody, and there is no third option.

4. Can we start with something small and fixed?

A first engagement should be small, scoped and fixed-price. Not because the partner needs proving in the abstract, but because it converts an unbounded risk into a known one, and both sides find out how the other actually works while the stakes are low.

If a partner will only start with a retainer or a large phase one, that is information. It may mean they are busy enough to choose. It may mean the model requires you to be locked in before you can evaluate it.

5. Whose conventions does the code follow?

You will inherit this. If the work comes back in a house style nobody at your agency recognises, with its own tooling and its own opinions, every future change goes back to the same partner whether you want that or not.

Agree conventions, repository layout and documentation format before the first commit. A partner who asks about your conventions unprompted has done this before.

6. What is the handover, precisely?

"We hand everything over" is not an answer. Ask what a handover contains: source, deployment configuration, environment variables, third-party accounts created for the project, the runbook for the parts that are not obvious.

Ask specifically about anything that depends on a paid third-party service, and ask before you start rather than at the end. A dependency discovered at handover is a conversation with your client that you should never have to have.

7. What does the timezone actually mean day to day?

Offshore is not one thing. A partner nine and a half hours behind you and a partner an hour and a half ahead are different products, whatever the hourly rate says.

The practical question is not how many hours overlap but whether a question asked at the end of your day has an answer waiting at the start of the next one. That depends less on geography than on whether the partner writes things down. Ask what the daily handover looks like in writing.

The two nobody asks

Numbers three and six. Everyone asks about price and stack; almost nobody asks what happens when the partner is full, or what exactly lands in their hands at the end. Those are the two that turn into an awkward call with a client, and they are both answerable in five minutes before anything is signed.