What Is a Trial Sprint? A 2-Week Way to De-Risk a New Dev Partner

Choosing a development partner is one of the stranger purchasing decisions an agency makes. The stakes are high: the partner will work under your brand, inside your deadlines, in front of your client. And the evidence available is close to zero. A sales call tells you how good they are at sales calls. A portfolio tells you what their best projects looked like after the bad ones were quietly removed. References tell you what their happiest customer thinks. Everyone in the room knows this, and most agencies sign anyway, because the deadline is real and the alternatives look the same.

There is a better instrument, and it is not complicated. Give the prospective partner two weeks of real, paid work and judge what actually happens. We call it a trial sprint. This article is about how to run one properly, with any partner, so that at the end you have a verdict instead of a vibe.

What a trial sprint is

A trial sprint is a short, paid, fixed-price engagement with a real deliverable from your actual backlog and a definition of done agreed before it starts. Two weeks is a natural length: long enough for real work to exist, short enough that a failure costs you a fortnight instead of a quarter.

Each word in that definition is doing work. Paid, because free samples select for partners desperate enough to work free, and because unpaid work gets toy effort on both sides; you want to see what their invoiced standard looks like, since that is what you will be buying. Fixed price, because the first thing you are testing is whether they can look at a bounded task and commit to a number. Real, from your backlog, because an invented exercise measures how well they do exercises. And defined done, in writing, before day one, because "we ran out of time but learned a lot" is the outcome you are paying two weeks to avoid.

Just as important: the work happens inside your process, not next to it. Their commits land in your repository, their tickets move on your board, they stand up in your standup. This is not ceremony. Process fit is most of what you are testing, because in our experience it is process friction, not raw skill, that kills white-label relationships by month three.

How to scope one that produces a verdict

The scoping mistake that ruins most trials is ambition. Pick something a small team can genuinely finish in two weeks with room for review and one round of revisions. If the natural candidate is bigger than that, cut it down to its first shippable slice rather than running the trial on a fragment that never has to survive contact with production. A bounded feature, a well-defined integration, a contained module of a larger build: these work. "Make a start on the replatforming" does not, because a start cannot fail, and a test that cannot fail is not a test.

Write the definition of done together, and keep it observable: the feature deployed to staging, tests passing, a demo run by them, your developer able to explain the code afterwards. That last one matters more than it looks. Code your team cannot read is code your team cannot own, and the entire point of white-label capacity is that the work remains yours.

What you are actually measuring

Not brilliance. Two weeks will not tell you whether a partner is brilliant, and you do not need brilliant; you need reliable inside your workflow. What two weeks will tell you, with surprising accuracy, is friction.

Watch when the questions arrive. Good partners front-load them: the awkward, slightly embarrassing questions about scope and edge cases show up in the first days, when they are cheap. Weak partners go quiet, make assumptions, and present you with the assumptions rendered in code on day nine. Watch what happens to the first blocker, because there will be one, and a partner who surfaces it the same day is showing you what month six looks like. Watch the estimates: not whether they were right, but whether you found out they were wrong from them or from the calendar. And at the end, have your own developer read the code the way they would read a colleague's. Tidy, boring, documented code from a stranger is one of the strongest signals this trade offers.

What a fair trial guarantees on paper

A trial sprint de-risks nothing if the paperwork underneath it is vague, so a fair one puts four things in writing. The price is fixed before the work starts. The code is yours whether or not you continue, assigned in writing, since as we covered in what do you actually own if your developer disappears, paying an invoice does not by itself move the copyright. An NDA is signed before the partner sees your backlog or your client's name. And walking away costs nothing beyond the trial price: no commitment, no penalty, no awkward clause that converts the trial into a retainer while nobody was reading.

Since we sell exactly this, here are our published terms as one example of the shape: two weeks, fixed price, a real deliverable from your backlog, and you keep the code either way. If we are not a fit, you have lost nothing. The rate for the ongoing relationship is public too, €35 to €55 per hour depending on stack and seniority mix, cancellable with 30 days' notice, and we run the trial the way we run the relationship, inside your workflow: your Git, your Jira, your Slack, your standups, weekly demos. Any partner worth testing should be willing to put their version of those four guarantees in writing; the ones who resist are answering your question early, which is at least efficient.

It is worth saying plainly that the test runs both ways. A serious partner is also deciding whether your agency is workable: whether the backlog item was really scoped, whether someone on your side answers questions inside a day, whether anyone is available to review the work. A partner who asks you hard questions before agreeing to a trial is not being difficult. They are doing the thing you are hiring them to do, just earlier than you expected.

When a trial sprint is the wrong tool

Keeping our rule about honest answers: sometimes you should skip it. If you need one small, well-understood task done once, a trial is overhead; hire a freelancer and be done. If the work is due in less time than a trial takes, the trial is a luxury; you are going to have to choose on thinner evidence, so weight references from people you actually know and read our guide to choosing a white-label partner for what to check in a hurry. And if nobody on your side can review the output, fix that first, because no trial can protect an agency that cannot evaluate what it receives.

But if what you are really deciding is whether to trust a partner with a workstream, a client, or the overflow that your hire-or-capacity arithmetic says you should not staff for, then two paid weeks is the cheapest good evidence money buys. A bad partner costs you a fortnight and a fixed price. A good one just delivered something from your backlog while auditioning.

If you have a backlog item that would make a fair test, tell us about it or book a partner call from the partners page, and we will tell you honestly whether it is scoped like a trial or like a trap, including the version of that answer where you should not run one with us at all.