Most outsourcing arrangements hand you capacity. You describe a ticket, someone elsewhere implements it, and the result comes back for review. The model works until the interesting problems arrive — the ones where the right answer is to change the ticket.
Embedded, not adjacent
A forward deployed squad works inside your codebase, on your roadmap, against your standards. That means:
- Reading the same dashboards your team reads
- Being on the same incident rotation
- Reviewing each other’s pull requests, in both directions
The difference is not proximity for its own sake. It is that context does not have to be serialised into a ticket before work can start.
Accountable for outcomes
We measure engagements on whether the thing shipped and held up in production, not on story points burned down. That distinction shows up in mundane places:
// Not "the endpoint returns 200"
expect(response.status).toBe(200);
// But "the customer's balance is correct afterwards"
expect(await getBalance(customerId)).toEqual(expected);If a feature ships and then pages someone at 3am, the work was not done.
Why this is a smaller claim than it sounds
Plenty of teams operate this way internally and do not think it is remarkable. What is unusual is buying it from outside. That requires senior engineers who can hold context without supervision, which is why our squads are structured the way they are — a principal for architecture, seniors who have run production before, and dedicated QA rather than QA as an afterthought.
It is a more expensive model per head. It is usually cheaper per outcome.