The first two years of offshore are a productivity loan

Offshore's brochure figure — 40 to 50 percent labour savings — is gross. The net arrives years later, after you repay it. Budget a roughly 20 percent drop in application-development efficiency across the first two years, 6 to 10 percent in management overhead, and a three-to-twelve-month knowledge-transfer window that saves nothing. Realistic steady-state savings on commoditised work land at 10 to 15 percent. You borrow productivity now and repay it, with interest, before the discount appears.

What does the first offshore contract cost up front?

The pitch compares two labour rates and stops there. The real profit-and-loss opens with a debit the rate card never mentions. Per CIO's long-running analysis of offshore economics, an IT organisation moving work offshore sees roughly a 20 percent decline in application-development efficiency across the first two years — friction from cultural and environmental distance, not headcount. Layer on 3 to 27 percent in productivity lags and another 6 to 10 percent simply to manage the contract. The handover is not free either: it takes three months to a full year to move the work, and that window saves nothing while you fly your own people out to teach the new team your own applications.

Net the debit against the advertised discount and the arithmetic gets uncomfortable. The brochure promises 40 to 50 percent; CIO's ROI work puts 10 to 15 percent as the realistic saving on highly commoditised work, with the headline figure reachable only in optimal circumstances. Here is the opening balance, before anyone writes a line of code:

Why does the dip last two years, not two weeks?

Because it is not a ramp cost you pay once; it is structural friction the model is built to sustain. Offshore delivery economics only close when 70 to 80 percent of the team sits offshore — practitioners quoted by CIO are blunt that the numbers only work at roughly 80 to 85 percent offshore. Which means your domain context — which SKU rules apply in which market, why the returnables reconciliation runs the way it does, which downstream job breaks silently when the nightly load is late — lives at the far end of a rotation schedule and a timezone gap. Every clarification becomes a round trip. In flow terms the work spends most of its life waiting for an answer rather than being worked, the exact pattern where a feature spends 80 percent of its life waiting in a queue. The efficiency loss is that queue, and it does not shrink with familiarity, because the model keeps refilling the far seats.

Where do the hidden costs hide on the balance sheet?

The debit above is the visible part. The expensive surprises are the items that never reach the business case. CIO's ROI analysis catalogues them without mercy: one retailer overstated its business case by $24 million because it kept half the work for a fifth of the staff it believed it was moving. A leading European software provider reported it takes 40 trips a year to manage its offshore product-testing programme — travel and coordination no rate card itemises. One pharmaceutical company held about 20 percent of its staff for six months past go-live, adding $1.5 million; overambitious headcount estimates alone can shave 10 to 20 percent off projected savings.

Those forty trips are the tell. Coordination is not a rounding error in distributed delivery; it is the dominant cost. We have measured what a single recurring meeting does to a calendar — one weekly executive meeting consumed 300,000 hours a year of downstream time — and an offshore governance layer is that dynamic institutionalised across a timezone gap and a contract boundary. Worse, the 6 to 10 percent you added to manage the engagement is time your organisation spends not shipping, stacked on top of the 42 percent of engineering time that already never touches a feature.

Who repays the loan, and when does it reset?

The metaphor breaks in one important way: a real loan amortises to zero. This one revolves. The steady-state 10 to 15 percent assumes the offshore team eventually holds enough context to work at parity minus the discount. But the delivery model that made the economics work — junior-heavy pyramids, a high offshore ratio — is also a rotation model. When the people who finally learned your empties logic roll onto the next statement of work, the context leaves with them, and the two-year efficiency debit begins accruing again on the same contract. There is no line item that reads “re-onboarding.” What arrives instead is the slow return of the precise friction the business case swore was a one-time cost. You are not paying down a balance; you are servicing a revolving one, and the interest rate is your own domain knowledge walking out of the room.

None of this argues that distance is the enemy; separating the work from the context is. The durable fix is not a thicker transfer deck or a heavier governance layer — both just service the interest. It is to keep the people who hold your domain logic on the same small team that ships it, and to write the reasoning into code and one event bus rather than a rotation schedule, which is the premise behind how we keep domain context inside the running system. The brewers who treat context retention as an architecture decision rather than a staffing line will stop paying two-year interest on work they have already done — and a year from now, when the next efficiency dip lands, they will notice it landed on someone else's contract.