You interviewed three good engineers. Eight months later two of them are gone, replaced by people who are perfectly competent and know nothing about your system. Nobody breached the contract. The vendor apologised, the invoices did not change, and your delivery slowed by a third for a quarter while the new arrivals learned what the old ones knew.
This is the most common way an offshore engagement disappoints, and it is rarely malice. It is usually structural, which is good news: structures can be inspected before you sign.
Why it happens
- The engineers were never yours. In a resourcing-pool model, people are allocated to accounts and reallocated when a larger account needs them. Your team is an assignment, and assignments change without anyone consulting you.
- The vendor buys talent rather than growing it. A firm that hires exclusively on the open market competes for the same engineers as every other firm in the same city, every year. Retention becomes a bidding contest they periodically lose, and you absorb the result.
- The margin depends on it. If the commercial model is a blended rate, replacing a senior with a mid-level engineer improves the vendor’s economics and leaves your invoice untouched. Nothing in the contract prevents it and everything in the incentives encourages it.
- Nobody was named. If the agreement specifies a number of engineers rather than named people, then swapping them is not a breach. It is the deal you signed.
What it actually costs
The visible cost is onboarding time. The larger cost is the slow erosion of judgement about your system — the accumulated sense of which parts are fragile, which shortcuts were deliberate, and what happened last time somebody touched the billing code. That judgement is the difference between a team that ships confidently and one that ships carefully, and it takes far longer to rebuild than to read the codebase.
Churn also compounds. Each departure makes the next one likelier, because the people left behind carry more context, feel more pressure, and become more valuable elsewhere. A team can be one resignation away from being effectively new.
What to check before signing
Retention promises are cheap. These are the things underneath them that are hard to fake.
- Are the engineers named in the contract? Named people, not a headcount. This single clause converts a swap from a routine operational decision into a conversation with you.
- Are they employees? Subcontracted engineers answer to a different employer with different priorities, and they leave on that employer’s schedule.
- Is the rate per person or blended? Per named engineer removes the incentive to quietly downgrade the mix.
- Where do their engineers come from? A vendor that trains its own people has a supply that does not depend entirely on out-bidding competitors. Ask how many of their current seniors started there junior. The answer is either a real pipeline or a shrug.
- What happens operationally when someone does leave? Everyone loses people eventually. The question is whether there is an overlap period, a documented handover and a named successor, or whether you find out by seeing a new face in standup.
The structural version of an answer
Our own position is that continuity is an architecture problem rather than a promise. Everyone on a Hazesoft team is a Hazesoft employee — no subcontracting. The fee is per named engineer, so there is no commercial upside in swapping anyone for a cheaper profile, and nobody is swapped mid-engagement. Clients interview the tech lead before signing, which only works if that person is still there afterwards.
The supply side matters just as much and gets discussed less. We run paid internships from our Kathmandu studio and grow seniors internally, which means our bench does not depend solely on winning a salary auction every year. It is an unglamorous answer to a question most vendors respond to with a retention statistic, and it is the reason we can talk about the same engineers owning the same system across years rather than quarters.
If your current team has turned over more than once and you want to work out whether the next arrangement would be different, take a technical call — an hour with an engineer, not a sales pitch.