Full stack is a real and useful capability and an unreal expectation when stretched. One person owning a feature end to end is genuinely faster than two people coordinating, up to a point. Past that point, depth in data modelling and depth in interface work diverge and the same person cannot credibly be excellent at both.
Where the model is clearly right
Early products, small teams, and systems where features are narrow enough for one person to own end to end. The saving is not in salary but in coordination: no handover, no waiting, no misunderstanding at the boundary. For a product finding its shape, that speed is worth more than specialist depth, and a full stack agency or consultant is often the most efficient arrangement available.
Where it starts to cost you
As a system grows, the hard problems become specialised: data modelling that survives growth, performance under load, accessibility done properly, security that stands examination. Generalists handle each competently and rarely excellently. The usual pattern is a full stack team with access to specialists for the parts where excellence matters, rather than either extreme. Be explicit about which half of your product needs the depth.
What to test, given nobody is expert in everything
Ask which half a candidate is genuinely stronger in and listen for an honest answer, since a claim of equal expert depth in both is itself the warning. Then test the weaker half, because that is where your risk sits. A good full stack engineer knows precisely where their competence ends and says so, which is considerably more valuable than confidence across the board.
A consultant is not the same purchase as an agency
A full stack consultant is one person: fast, cheap relative to a team, and a single point of failure with no cover and no review. That is acceptable for bounded work and risky for anything a business depends on. If you engage one, insist on your own repository and accounts, documentation as a deliverable and an express written copyright assignment, since under 17 USC 101 commissioned software is not automatically a work made for hire.
Questions people ask about full stack development agency
Is full stack good enough for a serious product?
For early products and narrow features, yes, and the saving is in coordination rather than salary. As the system grows, data modelling, performance, accessibility and security become specialised, and generalists handle them competently rather than excellently.
How do we test a full stack candidate?
Ask which half they are genuinely stronger in, then test the weaker half, because that is where your risk is. A claim of equal expert depth in both is itself the warning sign.
What are the risks of a single consultant?
No cover, no review and a single point of failure. Acceptable for bounded work, risky for anything the business depends on. Insist on your own repository and accounts, documentation as a deliverable and an express copyright assignment.