How to Measure GCC ROI Beyond Cost Savings
Most capability centres are still judged by a metric that was designed to justify their creation, not to manage their performance. Cost per full time equivalent tells a finance team what it saved on payroll. It says almost nothing about whether the centre shipped anything that mattered.
That gap is the single most common reason a GCC mandate stalls in year two. The savings case lands, the board acknowledges it, and then the questions quietly change. Leadership starts asking what the centre actually owns, how much faster the roadmap moves because it exists, and what would break if it went away tomorrow. Cost per FTE cannot answer any of those questions, and a leadership team that keeps hearing it will eventually conclude the centre has nothing else to say.
This guide sets out a practical way to measure GCC return on investment across four value layers. For each layer you get named metrics, the baseline you need to capture before the team exists, the owner who should report it, and the failure mode to watch for.
What this guide covers
- Why cost per FTE is a setup metric rather than a management metric
- The four value layers a mature capability centre should report against
- A scorecard with named metrics, baselines and owners for every layer
- The 30, 90, 180 and 365 day cadence that keeps the numbers honest
- Five measurement mistakes that reliably destroy credibility with a board
Why cost per FTE became the default, and why it fails
Cost per FTE won because it is easy. It needs one number from payroll and one number from a domestic benchmark, and it produces a percentage that anyone can read. In the business case phase that simplicity is genuinely useful. It is the number that gets a mandate approved.
The problem is that it is a ratio of inputs. It measures what you pay for a seat, not what comes out of it. A centre can improve its cost per FTE every quarter by hiring more junior people, and a centre can worsen its cost per FTE by hiring the senior architect who finally unblocks the platform migration. If the only metric on the board slide is cost per FTE, the second decision looks like a failure and the first looks like good management. That is precisely backwards.
There is a second, subtler failure. Cost per FTE is a comparison against a counterfactual that ages badly. It answers the question you asked in year zero, which was whether to build offshore at all. By year two nobody is relitigating that decision. The live question is whether this centre, with these people, working on these problems, is the best use of the budget it consumes. A savings percentage cannot adjudicate that.
The four value layers of a mature capability centre
A capability centre generates value in four distinct layers. They accumulate in order, each one takes longer to appear than the last, and each one needs a different kind of evidence. Reporting only on the first layer is what makes a centre look like a cost line rather than a capability.
What you pay for equivalent capability. Appears immediately, plateaus within four quarters, and is the easiest layer to game. Necessary to report, never sufficient.
How much more the organisation can attempt, and how much faster it gets there. Appears from roughly month four and is the first layer that reflects real delivery.
What the centre knows how to do that nobody else in the company does, and what it owns outright. Appears from month nine and compounds.
The moves the company can now make because the capability exists. Slowest to appear, hardest to quantify, and the layer that actually protects the budget.
The practical implication is that your reporting should change shape over time. A first year review dominated by layer one is appropriate. A third year review dominated by layer one means something has gone wrong, either in the centre or in how it is being asked to account for itself.
Layer one: measuring cost efficiency without lying to yourself
Cost efficiency is worth measuring properly, which most organisations do not do. The usual error is comparing a fully loaded offshore cost against an unloaded domestic salary, which flatters the result and eventually gets found out. The honest comparison is fully loaded against fully loaded, and it includes the costs that only appear once the centre is running.
Capture the baseline before the first hire. Once the team exists, nobody can reconstruct what the work used to cost, and every subsequent comparison becomes an argument rather than a measurement.
| Cost component | Frequently omitted | Why it matters |
|---|---|---|
| Salary and statutory benefits | No | The number everyone already tracks |
| Entity, compliance and statutory filings | Yes | Recurs annually and scales with headcount |
| Workspace, infrastructure and device refresh | Sometimes | Material at small scale, where it cannot be amortised |
| Recruitment and replacement cost | Yes | Attrition makes this a running cost, not a one off |
| Onshore management overhead | Almost always | The time your existing leaders spend running the centre |
| Ramp cost before productive output | Almost always | Real spend against zero delivery for one to two quarters |
| Knowledge transfer drag on the source team | Almost always | The sending team slows down while it teaches |
Omitting the last three is the most common reason a savings case that looked strong in year one fails to reconcile in year two.
Layer two: capacity and velocity
This is the first layer that reflects delivery rather than procurement, and it is where most of the defensible value sits in years one and two. The question is not whether work is cheaper. It is whether the organisation can now attempt things it previously had to decline, and whether it reaches them sooner.
Velocity metrics are only meaningful against a pre existing baseline, and they must be measured at the level of the whole system rather than the offshore team alone. A centre that doubles its own throughput while doubling the review burden on the onshore team has not created velocity, it has moved a queue.
Where velocity gains typically show up first
Indicative relative sensitivity in the first four quarters, not measured research findings. Set your own weightings against your own baseline before the team is hired.
Layer three: capability and owned intellectual property
Layer three is the difference between a centre that executes work and a centre that holds knowledge. It is the layer that determines whether you have built an asset or rented a queue, and it is almost never measured because it does not fall out of a finance system.
The measurement is structural rather than numerical. Ask which systems the centre owns end to end, which decisions it makes without escalation, and what would have to be rebuilt elsewhere if the centre closed. If the honest answer to the last question is nothing, the centre is a staffing arrangement wearing a capability centre label.
- Number of systems or roadmap areas owned end to end, with no onshore approval gate
- Architectural and design decisions made in the centre and accepted without rework
- Patents, internal platforms, tooling and reusable components created in the centre
- Documented domain expertise that exists only in the centre
- Proportion of incidents resolved without escalating onshore
- Number of onshore engineers who now depend on the centre for context rather than the reverse
Layer four: strategic optionality
Optionality is the value of the moves you can now make. A company with a mature capability centre can enter an adjacent product line, take on a regulated customer segment, or run a continuous release model without a hiring cycle standing between the decision and the execution. That option has real worth even in the quarters when it is not exercised.
It cannot be measured as a number, and attempting to invent one destroys the credibility of the rest of the scorecard. Report it as a short, specific list of decisions that were available because the capability existed, and decisions that were previously declined for capacity reasons and are now live. Three concrete examples will land harder with a board than any modelled figure.
A GCC ROI scorecard you can actually run
The scorecard below is deliberately small. Every metric has a named owner and a baseline that must be captured before the first hire. If you cannot capture the baseline, remove the metric rather than estimating it, because an estimated baseline will be challenged the first time the number is inconvenient.
| Layer | Metric | Baseline needed | Owner |
|---|---|---|---|
| Cost efficiency | Fully loaded cost per delivered roadmap item | Prior year cost per equivalent item | Finance |
| Cost efficiency | Total cost of ownership including ramp and overhead | Prior year fully loaded run rate | Finance |
| Capacity | Roadmap items attempted per quarter | Trailing four quarters | Product |
| Velocity | Median cycle time, specification to production | Trailing four quarters | Engineering |
| Quality | Defect escape rate and time to restore | Trailing four quarters | Engineering |
| Capability | Systems owned end to end with no approval gate | Zero at start | Centre lead |
| Capability | Escalation rate to onshore | First 90 days of operation | Centre lead |
| Retention | Regretted attrition among senior engineers | Company wide rate | People |
| Optionality | Decisions now available, listed explicitly | Decisions declined for capacity in prior year | Executive sponsor |
The measurement cadence that keeps the numbers honest
Measurement fails more often through timing than through metric selection. Reporting delivery metrics during ramp makes a healthy centre look broken; reporting only cost metrics after year one makes a valuable centre look ordinary. Move the emphasis deliberately.
Capture every baseline
Trailing four quarters of throughput, cycle time, defect rate and fully loaded cost. This is the only moment these numbers are uncontested. Write them down and circulate them.
Report readiness, not output
Hiring against plan, compliance and entity milestones, environment and access readiness, onboarding completion. Delivery metrics in this window are noise and will be used against you.
Introduce delivery, hold cost flat
First throughput and cycle time comparisons against baseline. Keep reporting cost, but stop leading with it. Begin tracking escalation rate as the early capability signal.
Lead with capability
Systems owned end to end, escalation rate trend, retention of senior engineers. Cost becomes an appendix item. Introduce the optionality list at the annual review.
Five measurement mistakes that destroy board credibility
Each of these is recoverable in isolation. Together they are the standard path by which a centre that is genuinely working loses the argument for its own budget.
- Comparing fully loaded offshore cost against unloaded domestic salary. It inflates the saving and it will be caught.
- Reporting delivery metrics during the ramp window, before the team has had a chance to produce anything.
- Measuring the offshore team in isolation rather than the whole delivery system, which hides queues that simply moved onshore.
- Inventing a number for strategic optionality instead of listing concrete decisions it unlocked.
- Never retiring layer one from the top of the report, which trains the board to keep evaluating the centre as a cost line.
The measurement model is a design decision, not a reporting decision,Which metrics you can report in year two is determined by which baselines you captured in month zero and how the centre was scoped. A capability centre designed only to reduce cost will only ever be able to report cost.
Frequently asked questions
How long before a GCC shows return beyond cost savings?
Cost efficiency is visible almost immediately. Capacity and velocity gains typically become measurable from around month four to six, once ramp is complete. Capability and owned IP generally take nine to twelve months to become demonstrable, and strategic optionality usually needs a full year or more before you can point to specific decisions it enabled.
Should we stop reporting cost per FTE entirely?
No. It remains a legitimate input metric and boards will continue to ask for it. The change is positional. It should move from the headline to the appendix once the centre is past its first year, and it should never be the metric that determines whether the centre is judged successful.
What if we never captured a baseline before hiring?
Reconstruct what you can from version control, ticketing and incident systems, which are harder to dispute than recollection, and be explicit about what is reconstructed rather than measured. Then set a clean baseline from the current quarter forward and report against that. A transparent short baseline is more credible than a generous long one.
How do you measure IP ownership in a services context?
Count systems owned end to end without an onshore approval gate, decisions made without escalation, and internal tooling or platforms created in the centre and adopted elsewhere. If none of those are true, the arrangement is delivering capacity rather than capability, which is a legitimate model but should be labelled accurately.
Is a smaller centre harder to justify on ROI?
It is harder to justify on layer one alone, because fixed costs such as entity, compliance and workspace cannot be amortised across a large headcount. It is often easier to justify on layers two through four, because a small senior team tends to own more per person and escalate less. This is the core argument for the Nano GCC model.
Who should own the GCC scorecard?
Split it. Finance owns the cost layer, product and engineering own capacity, velocity and quality, the centre lead owns the capability metrics, and the executive sponsor owns the optionality list. A scorecard owned entirely by finance will drift back to cost per FTE within two quarters.
How often should the scorecard be reviewed?
Monthly at the operating level for the delivery and capability metrics, quarterly at the executive level, and annually at board level with the optionality list attached. Reviewing everything at board cadence hides problems for too long; reviewing everything monthly produces noise.
What is the single strongest signal that a GCC is working?
Falling escalation rate alongside rising ownership. When the centre resolves more without asking, and takes on more that nobody reviews, capability is genuinely transferring. Almost every other positive metric can be produced without that being true.
Sources & further reading
- NASSCOM — https://nasscom.in/
- Deloitte — https://www.deloitte.com/
- McKinsey & Company — https://www.mckinsey.com/
- Everest Group — https://www.everestgrp.com/
Build a capability centre that reports on value, not just cost
Hexominds designs Nano GCCs with the measurement model built in from day one, so your first board review has more than a payroll saving to show.