The comparison almost everybody makes first is an hourly rate here against an hourly rate there. It is the wrong comparison and it flatters the decision in both directions, because it omits four costs on one side and the counterfactual on the other.
What is on the invoice
A monthly figure per person, covering salary, statutory contributions, the thirteenth month, the seat, equipment, supervision and margin. Whether those components are itemised differs by provider and the entry on the three models explains why the itemisation matters.
That figure is the smallest part of the analysis and the only part most businesses model.
The four costs that are not on it
Your own time. Somebody onshore briefs, reviews and answers questions. Budget a day a week in the first month and half a day a week thereafter, indefinitely. At a senior person's cost that is not a rounding error, and if nobody has the time, the engagement is in trouble before it starts.
Writing things down. Procedures held in somebody's head have to be written before they can be handed over. The work is genuinely valuable and it arrives entirely before any benefit does, which makes the first months worse than the model suggests and every subsequent month better.
The ramp. A person is not fully productive on day one, and in a role with any context at all they are not fully productive for two or three months. Modelling full output from week one overstates the first year by a substantial margin.
Turnover. People leave. Replacement means recruitment time, a gap, and a second onboarding, and the second is cheaper than the first only if the first produced documentation. The entry on attrition covers the rates to plan against.
Two smaller ones worth naming
Tooling licences for an additional user, which are trivial individually and not always trivial across a team. And currency: a contract priced in one currency and paid from another moves with the exchange rate, and over a multi-year engagement that movement can exceed the margin you were saving.
The term nobody includes
The counterfactual. The saving is not the difference between an offshore person and a local person; it is the difference between the offshore person and whatever you would otherwise have done.
Frequently what you would otherwise have done is nothing, in which case the comparison is between a cost and zero and the case has to be made on output rather than on saving.
Sometimes what you would otherwise have done is automate the task, in which case a person is being hired to do something a script could do, permanently, and the saving is negative over any reasonable horizon. That case is worth checking before any conversation with any provider.
A model that survives contact
Take twenty-four months. On one side, put the invoice, your management time at a real internal cost, the documentation effort, the ramp expressed as reduced output in the first quarter, and one replacement cycle. On the other, put the counterfactual honestly.
Twenty-four months matters because a twelve-month model is dominated by the setup costs and makes almost every engagement look poor, while a sixty-month model assumes a stability nobody has.
What the answer usually is
For a well-defined role with steady volume, still substantially positive. The industry exists because the arithmetic works when the role is right.
For a poorly defined role, negative, and the model will show it if the ramp and the management time are in there. That is the value of building it: not to justify the decision but to identify which roles survive it.
Break-even
Typically some months in, not immediately, and later than most businesses expect. A business that needs the saving this quarter is solving a cash problem with a hiring decision, and that combination has a poor record wherever the person sits.
Comparing providers on price
Two quotes for the same role can differ by a third and describe different things. Check what the salary band assumed, whether supervision is included or billed, what happens at the annual review, and whether the seat and equipment are inside the figure.
A cheaper quote that assumes a lower salary band is not cheaper; it is a different candidate pool, and the entry on attrition explains what that costs eighteen months later.
The internal cost of your own time
Most businesses model their own management time at zero, because it is not invoiced. Put a real number on it: the cost of the person doing the briefing, at whatever that person costs.
Doing this changes conclusions. A role saving a modest amount per month, consuming half a day a week of an expensive person, can be roughly neutral, and knowing that before rather than after is the point of the exercise.
What improves the arithmetic most
Not a cheaper provider. Better documentation, which shortens the ramp and reduces both the question load and the cost of the eventual replacement.
It is the only input that improves every term in the model at once, and it is the one businesses defer because it produces nothing immediately.