Offshore proposals get compared the way procurement knows how: monthly rate per engineer, sorted ascending. It is the most visible number in the deck and the least informative one in the decision. Two proposals 30 percent apart on rate can invert on real cost within two quarters, because the rate card only prices the part of the engagement you can see. We have taken over enough engagements that started with the cheapest bid to know what the missing rows in that spreadsheet cost.
You are not buying hours. You are buying shipped outcomes, and the two are priced very differently.
What a monthly rate actually contains
Two rates for the same nominal seniority can describe two different products. The difference is what the provider chose to include:
- Recruitment and bench: a governed partner amortises sourcing, vetting, and a warm bench into the rate. A placement shop prices the person in front of you and rebuilds from zero when they leave.
- The governance layer: weekly performance reviews, independent code-quality checks, and delivery oversight either exist inside the rate or do not exist at all. This is the single largest structural difference between two otherwise similar numbers.
- Replacement risk: who pays for the transition when an engineer leaves or underperforms. A replacement guarantee with the cost on the partner is real money priced into the rate; its absence is real money priced into your future.
- The employment envelope: equipment, benefits, statutory obligations, and compliance in the engineer's country. Rates that look impossibly low are usually externalising something from this row.
- Escalation terms: what the rate does at renewal. A low entry rate with unbounded annual escalation is a pricing strategy, not a price.
The costs that never appear on any rate card
The second half of the real cost sits on your side of the table, and no proposal will itemise it for you. Rework is the largest: the gap between a team that ships reviewed, tested code and one that ships plausible code surfaces months later as defects, refactoring, and the slow tax of working in a degraded codebase. Ramp time is the quietest: every engagement has a period where you pay full rate for partial output, and its length is set by the partner's onboarding discipline, not by the engineer's talent. Attrition is the most expensive per event: when context walks out the door, you pay for the handover, the re-ramp, and the mistakes the replacement makes in territory the predecessor knew. None of these vary with the rate. All of them vary with the operating model behind it.
Why the cheapest rate is often the most expensive engagement
A rate meaningfully below market is funded by removing something: the reviews, the bench, the mentorship, the replacement guarantee. The engagement then runs fine for a quarter, because new engineers on fresh context perform well without scaffolding. The bill arrives in quarters two and three, as unreviewed decisions compound and the first rotation resets context to zero. By the time the total cost of the cheap engagement is visible, it includes a rescue: re-establishing standards, re-documenting systems, and often re-running the vendor selection. The engagements that stay cheap are the ones where someone is accountable for quality while the work happens, which is precisely the row the discount removed. How that accountability is structured is its own decision, covered in staff augmentation versus a dedicated squad; whether the work should leave your building at all is a time-to-capacity question, not a salary question.
Comparing proposals on equal footing
Before comparing numbers, normalise what the numbers describe. Five questions do most of the work:
- Itemise the rate: which of recruitment, bench, governance reviews, code-quality checks, equipment, and statutory costs are inside it, in writing.
- Replacement terms: who pays for transition and handover when a person changes, and what counts as grounds for replacement.
- Ramp expectations: what the partner commits to for time-to-first-merged-work, and what onboarding structure backs the commitment.
- Exit terms: notice period, knowledge handover obligations, and what happens to documentation and access on the last day.
- Governance cadence: the reviews, reports, and escalation paths the partner runs, with names and frequencies, not adjectives.
The answers matter less than the form they arrive in. A partner that answers all five in writing, with numbers and names, is pricing an operating model. A partner that answers with reassurance is pricing hours and hoping. Either way, you have learned what the rate card was hiding before you have spent anything.
A cost model that predicts the invoice
- Model twelve months, not one: rate, plus your own management hours at their loaded cost, plus expected ramp weeks, plus a rework allowance scaled to the governance on offer, plus one replacement event.
- Score both proposals in that model. The 30 percent rate gap usually shrinks sharply; sometimes it changes sign.
- Pilot with one or two engineers on an offshore software engineering engagement before any squad-sized commitment, and treat the pilot's delivery data as the only bid that cannot be gamed.
- Measure quarterly on cost per shipped outcome: features delivered to production per dollar, not hours billed per dollar. It is the only metric both sides cannot argue with.
None of this argues against offshore economics, which are real and structural. It argues against buying them on the one number that predicts the least. The rate card tells you what an hour costs. The operating model behind it decides what a shipped outcome costs, and shipped outcomes are the thing you were buying all along.
