F2C Labs
Custom Web & Software Development
2 weeks ago
1,431 words
2026-11-05

The Offshore Pilot Project: How to Test a Software Vendor Before You Commit

How to structure a paid offshore pilot project: fixed scope, acceptance criteria, weekly demos on running software, what to evaluate, red flags, and what a pilot can never prove.

Liêu Vĩnh Toàn
Liêu Vĩnh Toàn

Founder & CEO

ArchitectureOptimization

An offshore development relationship can fail even when nobody lied. One way it happens: the buyer makes a large, hard-to-reverse commitment based on a sales deck, a rate card, and two reference calls the vendor arranged. Everything in that package was produced or curated by the vendor. What it cannot show is how the vendor actually works, and watching the vendor work on a small real engagement gives you direct evidence before a long commitment. That gap is exactly what an offshore pilot project is for.

A pilot is a small, paid, time-boxed engagement, usually two to four weeks, with a narrow but real scope. Done properly, it converts the hiring decision from a bet on marketing materials into a judgment based on evidence you collected yourself. This guide covers how to structure a pilot, what to measure while it runs, the red flags that should end it early, and, just as important, what a pilot can never prove.

Why committing up front is a gamble

When you sign a six or twelve month engagement after a few calls, you are pricing in three assumptions: that the team you met in the sales process is the team that will do the work, that communication will stay as responsive after signature as it was before, and that the estimate you were given reflects how this vendor actually estimates. None of those assumptions can be checked from a proposal. Our guide on how to hire an offshore development team covers screening vendors before this stage, but even after good screening, the residual uncertainty is large, and the price of discovering it late is a sunk quarter and a codebase you may not fully control.

The contract does not remove this risk either. A good contract determines what happens when things go wrong; the offshore development contract checklist walks through those clauses. A pilot answers an earlier question on a scale small enough to afford the answer: do things go wrong at all.

What a pilot project is, and what it is not

A pilot is a real piece of work: a feature, an integration, a prototype with users. It is delivered by the same developers who would staff the larger engagement, into a repository you own, under the same legal terms you intend for the main contract. And it is paid. A vendor working for free is not running a pilot; they are running an extended sales demo, and both sides know it, so neither behaves the way they will once real money and real deadlines are involved.

A pilot is also not a shrunken version of the final product, and not a throwaway test task from a hiring site. The scope should be small in size but real in every other dimension: real requirements, real acceptance criteria, real users if you can manage it, real deployment at the end.

Two developers reviewing code together on a laptop
Two developers reviewing code together on a laptopSource: Technical & Research Documentation

How to structure the pilot

Four decisions, all made before any code is written, determine whether the pilot produces evidence or just activity.

  • Fix the scope and write the acceptance criteria first. Two to four weeks, one clearly bounded deliverable, and a written definition of done that both sides agreed to before work started. If the acceptance criteria can be renegotiated after the demo, they were never criteria.
  • Insist on the actual team. The developers on the pilot should be named, and they should be the same people proposed for the main engagement. If the vendor reserves the right to swap staff between the pilot and the contract, the pilot proves nothing about who you will actually get.
  • Set the demo rhythm before the project starts. A weekly review of running software, not a status deck. Scrum.org's guidance on the Sprint Review is a useful template: the point of the review is to inspect the actual increment and adapt the plan, not to present progress. A vendor who demos working software every week is a vendor you can manage from twelve time zones away.
  • Put the legal groundwork in place from day one. IP assignment, confidentiality, repository ownership, and access rights should follow the same clauses you would use for the full contract. A pilot signed on a handshake teaches you nothing about how the vendor handles real terms.

What to evaluate while the pilot runs

The deliverable matters, but it is not the main output of the pilot. The main output is a set of observations about how this vendor works. Five of them predict the large engagement better than the final demo does.

  • How they deliver bad news. At some point in four weeks, something will be late or harder than expected. Did you hear about it early, in writing, with a revised estimate, or did you discover it yourself?
  • Whether estimates move. An estimate that never changes across four weeks of real work is not proof of accuracy on its own, because precision alone does not establish estimate quality. Ask what assumptions and contingency sit behind the number, and watch how and when they revise.
  • Written communication quality. Offshore work runs on writing: tickets, decisions, questions. Are decisions confirmed in writing without being asked? Are questions precise, or does each one need three round trips?
  • Code you can inspect. Have your own senior developer, or a hired reviewer, read the code mid-pilot. You are not judging elegance; you are checking whether a stranger can understand it, whether it is tested, and whether the vendor is comfortable being read.
  • What happens when you push back. Reject one thing during the pilot, reasonably, and watch. A partner fixes it without drama or pushes back with reasons. A vendor who agrees instantly with everything will also agree instantly with things they cannot deliver.
A developer working at a desk with a monitor and a laptop showing code
A developer working at a desk with a monitor and a laptop showing codeSource: Technical & Research Documentation

Red flags that should end the pilot early

Some signals are strong enough to stop spending immediately: demos of slides instead of running software; a weekly demo quietly rescheduled or replaced by a written report; new faces appearing on calls without explanation; no questions asked in the first week, which suggests the requirements were never read closely; and a fixed price quoted for the main engagement before the pilot has even surfaced the unknowns. Any one of these is cheaper to learn in week two of a pilot than in month four of a contract.

What a pilot cannot tell you

Be honest about the limits. A pilot cannot show you how the vendor behaves in month nine, when the novelty is gone and the maintenance tickets are boring. It cannot prove team stability over a year, or how the vendor scales from three developers to eight. Treat the pilot as the first filter, not as proof. A practical way to protect a long engagement is to keep the pilot's habits for the whole relationship: weekly demos on running software, written decisions, inspectable code. If you are still weighing where to run the engagement, our buyer's guide to outsourcing software development to Vietnam covers the market context.

Frequently asked questions

How long should an offshore pilot project be?

Two to four weeks is a workable default. Shorter than two weeks and you are measuring the vendor's onboarding, not their working style. Longer than four and you have quietly started the real engagement without ever deciding to.

Should a pilot be paid?

Yes. A paid pilot pushes both sides much closer to how they will behave in the real contract, and it buys you ownership of the output, including the code and the findings. Expect to pay the vendor's normal rates for a small scope, and treat any discount as what it is: a sales cost.

What if the pilot fails?

Then the pilot worked. A pilot that reveals weak communication or missed demos in week three has saved you a six month contract. Budget for the possibility of running two pilots with two vendors and picking one; the losing pilot is the cheapest due diligence you will ever buy.

Can a pilot replace reference checks?

No, and it does not need to. References tell you how the vendor finished with other clients; the pilot shows you how they will start with you. Use both, in that order: references to shortlist, pilot to decide.

References

Liêu Vĩnh Toàn
Liêu Vĩnh ToànTechnical Expert

Founder & CEO

Software engineer specializing in AI-integrated product development on Google Cloud Platform (GCP). Leads overall product lifecycle from distributed backend architecture to production deployments, accelerating business go-to-market.

1-on-1 Consultation

Related Dispatches

View all articles →