The Definitive Guide to GCC ROI Calculators in 2026-2027, How to Build a Business Case Your Board Will Approve

Jai Shri Krishna. Every India Global Capability Centre (GCC) decision eventually reduces to a single slide. A board asks what the centre costs, what it saves, when it pays back, and what happens if the assumptions are wrong. The Return on Investment (ROI) calculator is the instrument that answers those four questions. Used well, it is the most persuasive artefact in the entire approval pack. Used badly, it is the reason a sound programme is deferred by two quarters while Finance re-runs the numbers. This guide sets out how to build, interrogate and defend a Global Capability Centre ROI model in 2026-2027, and how to use the seventeen decision tools published at /calculators to do it in an afternoon rather than a fiscal quarter.

TL;DR

A credible GCC ROI model has five layers: a fully loaded cost stack, a transition and one time cost schedule, a productivity and attrition adjustment, a tax and transfer pricing overlay, and a scenario grid. Most first drafts contain only the first layer, which is why they overstate savings by 18 to 30 per cent and understate payback by four to seven months. A well built model for a 250 seat centre in Pune or Hyderabad typically shows 52 to 64 per cent steady state cost reduction against a United States or United Kingdom baseline, cash payback between month 14 and month 22, and a five year net present value that survives a 20 per cent adverse shock on every major assumption. Build the model for defensibility, not for the largest possible savings number.

Table of contents

• 1. Why most GCC ROI models fail their first Finance review
• 2. The five layers of a defensible model
• 3. Layer one, the fully loaded cost stack per seat
• 4. Layer two, one time and transition costs
• 5. Layer three, productivity, ramp and attrition
• 6. Layer four, tax, transfer pricing and the cost plus mark up
• 7. Layer five, scenarios, sensitivities and the shock test
• 8. A tour of the seventeen calculators and what each one answers
• 9. Worked example, a 250 seat finance and technology centre
• 10. City sensitivity, what changes between Pune, Mumbai, Bangalore and Hyderabad
• 11. Tier 2 cities and the hub and spoke ROI case
• 12. Entry model economics, direct build against managed against Build Operate and Transfer
• 13. Capital expenditure against operating expenditure and what the Chief Financial Officer actually optimises
• 14. Currency, wage inflation and the five year drift
• 15. Sector nuances, accounting, banking, pharmaceutical and life sciences, engineering and semiconductors
• 16. What generative Artificial Intelligence does to the denominator
• 17. The board pack, seven slides that get an approval
• 18. Twelve modelling errors we see every quarter
• 19. Governance, who owns the model after approval
• 20. Frequently asked questions
• 21. Next steps and references

1. Why most GCC ROI models fail their first Finance review

The failure is rarely arithmetic. It is almost always a failure of completeness and provenance. A Group Financial Planning and Analysis (FP&A) team reviewing an India business case asks three questions in sequence. Where did each input come from. What is excluded. What happens if the input is wrong by twenty per cent. A model that cannot answer all three within one meeting is sent back, and the programme loses a quarter.

The most common structural defect is comparing the wrong two numbers. Teams compare an Indian salary to a home country salary. That comparison is meaningless. The correct comparison is between the fully loaded, fully absorbed annual cost of delivering a defined scope of work in the home country, including the share of management, real estate, technology, recruitment, benefits, payroll taxes and unproductive time, and the equivalent fully loaded and fully absorbed annual cost of delivering the same scope from India, including the mark up paid to the Indian entity under transfer pricing rules, the cost of the retained onshore governance layer, and the cost of the transition itself amortised over the model horizon.

The second defect is silence on the retained organisation. Work does not vanish from the home country the moment it lands in India. A vendor manager, a process owner, a control owner and a quality reviewer usually remain, at least for the first two years. If the model shows a one hundred per cent reduction in home country effort for a migrated process, the model is wrong. A realistic retained factor is 8 to 15 per cent of the migrated Full-Time Equivalent (FTE) count in year one, tapering to 4 to 8 per cent by year three for stable transactional work, and remaining higher for judgement heavy regulatory work.

The third defect is a ramp curve drawn as a straight line. Hiring, notice periods, knowledge transfer and stabilisation do not behave linearly. A ninety day average notice period in India means the gap between an accepted offer and a productive desk is routinely four to five months for experienced hires. A model that assumes month one hiring and month two productivity will always show a payback that the programme cannot deliver, and the credibility loss when actuals arrive is expensive.

The fourth defect is treating attrition as a human resources statistic rather than a cost line. Every regretted exit carries a replacement recruitment fee, a re-training cost, a productivity gap of roughly ten to sixteen weeks, and a quality cost that shows up as rework or control exceptions. At an eighteen per cent attrition rate on a 250 seat centre, that is forty five replacements a year and a real annual cost that belongs inside the model, not in a footnote. Quantify it with the attrition cost tool at /calculators.

2. The five layers of a defensible model

A model that survives scrutiny is built in layers, each of which can be defended independently. Present it in this order, because it is the order in which Finance will interrogate it.

• Layer one, the fully loaded cost stack per seat, expressed in a single reporting currency and reconciled to a published salary benchmark.
• Layer two, one time and transition costs, including entity formation, facility fit out, technology, recruitment, travel, knowledge transfer and parallel run.
• Layer three, productivity, ramp and attrition adjustments that convert a headcount plan into an effective delivered capacity plan.
• Layer four, the tax and transfer pricing overlay, which converts the Indian cost base into a charge to the parent and captures the corporate income tax cost of the mark up.
• Layer five, the scenario grid, which stress tests each material assumption independently and jointly and reports the range rather than the point estimate.

Each layer has a single named owner. Layer one belongs to the location partner or the reward team. Layer two belongs to the programme director. Layer three belongs to the operations lead. Layer four belongs to the group tax director. Layer five belongs to Financial Planning and Analysis. When a board asks a question, it should be obvious who answers.

3. Layer one, the fully loaded cost stack per seat

The fully loaded cost per seat is the sum of eleven components. Missing any one of them is the fastest way to lose an argument with Finance. The percentages below are typical for a Tier 1 Indian city in Financial Year 2026-2027 and should be replaced with your own quotes before board submission.

• Base salary, benchmarked to the seventy fifth percentile for the first two hiring waves and the median thereafter. This is 58 to 66 per cent of the stack.
• Variable pay and annual bonus, typically 8 to 15 per cent of base for individual contributors and 15 to 25 per cent for leadership.
• Statutory employer contributions, including Provident Fund, gratuity accrual, Employees State Insurance where applicable and professional tax. Budget 13 to 16 per cent of base.
• Health, life and accident insurance for the employee and dependants, an increasingly decisive retention lever. Budget 3 to 5 per cent of base.
• Real estate, fit out amortisation, facility management, power backup and security. In a Grade A managed environment this runs at roughly 8 to 14 per cent of the stack depending on city and seat density.
• Technology, comprising end user compute, licences, connectivity, virtual desktop infrastructure where required, and the security stack. Budget 6 to 10 per cent.
• Recruitment and onboarding, amortised over expected tenure rather than expensed in year one.
• Training, certification and language or domain academies, 1 to 3 per cent.
• Transport, meals, wellbeing and engagement, 2 to 4 per cent, higher where a night shift operates.
• The India management and enabling layer, comprising the centre head, human resources, finance, compliance, information security and administration. At scale this is 9 to 13 per cent of the direct cost base and it is the line most often omitted entirely from first drafts.
• Compliance, statutory audit, secretarial, transfer pricing documentation and legal, a modest but non zero line that grows with entity complexity.

Express the result as a single fully loaded annual cost per seat, then divide into a monthly figure for cash flow modelling. For 2026-2027, indicative fully loaded annual costs per seat for a mid level professional with four to eight years of experience sit broadly in these bands, in United States dollars, and vary materially by function and seniority. Validate role by role against the published benchmarks at /india-gcc-salary-benchmarks.

• Pune, 24,000 to 34,000 for finance and accounting, 30,000 to 46,000 for technology and engineering.
• Hyderabad, 24,000 to 35,000 for finance and accounting, 31,000 to 48,000 for technology, higher for silicon design.
• Mumbai, 29,000 to 42,000 for finance, banking and capital markets roles, with real estate the principal driver of the premium.
• Bangalore, 27,000 to 38,000 for finance, 36,000 to 58,000 for senior product engineering, Artificial Intelligence and platform roles.
• Tier 2 hubs including Ahmedabad, Kochi, Coimbatore, Indore, Jaipur, Bhubaneswar and Surat, 30 to 45 per cent below Bangalore for comparable mid level roles.

Against these, a comparable fully loaded seat in New York, London, Zurich, Sydney, Singapore or Dublin ranges from 95,000 to 210,000 United States dollars depending on city and function. That is the arithmetic behind the headline arbitrage. It is also why the arbitrage is not the interesting part of the model. The interesting part is what remains after transition, retained organisation, attrition, mark up and tax.

4. Layer two, one time and transition costs

One time costs are where optimistic models quietly lose four to eight months of payback. Build the schedule explicitly, month by month, and keep it separate from the run rate so that the board can see the investment envelope in isolation.

• Entity formation, comprising incorporation under the Companies Act 2013, Permanent Account Number and Tax Deduction Account Number registration, Goods and Services Tax registration, bank account opening, Foreign Exchange Management Act filings and Software Technology Parks of India registration where relevant.
• Facility, comprising security deposit, fit out, furniture, access control, and the period of double rent that overlaps the previous arrangement.
• Technology build, comprising network, security tooling, endpoint fleet, identity and access management integration and any data residency work required by the parent's regulator.
• Recruitment for the founding cohort, including search fees for the centre head and the first leadership layer, which are the two hires that determine the trajectory of the entire centre.
• Knowledge transfer, comprising travel, subject matter expert time in the home country, documentation, shadow and reverse shadow phases, and the parallel run during which both locations perform the work.
• Programme management, legal, tax advisory, transfer pricing study and the initial internal audit readiness work.
• Change management and communication in the home country, a line that is almost always underfunded and almost always the source of the largest execution risk.

For a 250 seat centre entering through a managed model, the one time envelope typically lands between 1.1 and 1.9 million United States dollars. For a direct build with an owned entity and a self managed facility, the envelope is usually 2.2 to 3.6 million United States dollars and the elapsed time to a first productive team is eight to fourteen weeks longer. Model both. The comparison is the single most useful slide in the entry model discussion, and the entry model comparison tool at /calculators produces it directly.

5. Layer three, productivity, ramp and attrition

Headcount is an input. Delivered capacity is the output that matters. Three adjustments convert one into the other.

Ramp

A realistic ramp for a knowledge process is a four to six week hiring cycle for junior roles and eight to fourteen weeks for experienced roles, followed by a ninety day notice period for a material share of experienced hires, followed by six to twelve weeks of training, shadowing and reverse shadowing before independent production. Plan for a new joiner to reach seventy per cent productivity at week ten and full productivity between week sixteen and week twenty two for judgement heavy work. For transactional work, those milestones compress to week six and week twelve.

Utilisation and shrinkage

Leave, public holidays, training, town halls and unplanned absence consume 14 to 18 per cent of gross available hours in India. A model built on 2,080 annual hours per seat overstates capacity. Use 1,700 to 1,780 productive hours for planning, and lower for centres running a strict five day office policy with long commutes.

Attrition

The Financial Year 2026 planning ranges are 14 to 19 per cent for Pune and Hyderabad, 18 to 24 per cent for Bangalore, 16 to 21 per cent for Mumbai, and 11 to 14 per cent for the Tier 2 hubs. Each regretted exit costs between 0.4 and 0.9 times the annual fully loaded cost of the seat once recruitment, ramp, lost productivity and rework are counted. On a 250 seat Bangalore centre at twenty per cent attrition, that is a recurring annual cost of roughly 1.3 to 2.4 million United States dollars, which is large enough to change a city decision on its own.

The practical consequence is that a centre in a lower attrition city can be materially cheaper on a five year total cost of ownership basis even when its salary bill is similar. Model attrition explicitly rather than burying it inside a contingency percentage, because attrition is the assumption most likely to be challenged and the one where a specific, defensible number earns the most credibility.

6. Layer four, tax, transfer pricing and the cost plus mark up

An India captive that serves only its parent is, in almost all cases, remunerated on a cost plus basis. The Indian entity recovers its full operating cost base and charges a mark up on that base. The mark up is taxable in India at the applicable corporate rate, and the total charge, cost plus mark up plus applicable indirect tax treatment, is what actually appears in the parent's profit and loss statement. Many first draft models compare the Indian cost base to the home country cost and forget the mark up entirely, which overstates savings by roughly the mark up percentage multiplied by the cost base.

• For software development and Information Technology Enabled Services, the safe harbour mark up commonly applied sits at 17 to 18 per cent on operating cost, with variation by turnover band and service category.
• For Knowledge Process Outsourcing with higher value analytical content, the applicable band is higher and depends on the employee cost to operating cost ratio.
• For contract research and development services, distinct safe harbour categories apply and the documentation burden is heavier.
• Where a group prefers not to use safe harbour, an Advance Pricing Agreement provides multi year certainty and is worth evaluating for centres above roughly 400 seats or where the charge exceeds a level that would attract routine scrutiny.

The corporate tax cost on the mark up is real but modest against the arbitrage. On a fifteen million dollar annual Indian cost base at an eighteen per cent mark up, the mark up is 2.7 million dollars and the Indian corporate tax on it, at the concessional rate available to many new manufacturing and certain service structures or at the standard rate otherwise, is a low single digit percentage of the total programme cost. Present it explicitly. A tax line that appears in the model builds more confidence than a tax line that a reviewer discovers. The technical background is set out at /transfer-pricing-safe-harbour-brief and the entity design choices at /holding-structure-india-entry.

Two further tax items belong in a complete model. First, Goods and Services Tax on export of services is generally zero rated, but input tax credit accumulation and the refund cycle create a working capital cost that should be modelled as a timing item rather than ignored. Second, equalisation levies, withholding on cross border payments for software and services procured by the Indian entity, and the group's own Pillar Two effective tax rate calculations can each move the answer for larger groups. Involve the group tax director in the model at the start, not at the approval meeting.

7. Layer five, scenarios, sensitivities and the shock test

A point estimate invites a debate about the point. A range invites a decision. Present three scenarios and one shock test.

• Conservative case, hiring twenty per cent slower than plan, attrition three percentage points above the city norm, wage inflation at the upper end, retained onshore organisation at the upper end, and a six month slip in the migration of the final wave.
• Planning case, the numbers you actually believe, with every input sourced and dated.
• Upside case, faster hiring in a favourable market, attrition at the lower end of the city band, and the productivity benefit of automation and generative Artificial Intelligence realised on schedule.
• Shock test, in which every one of the five most material assumptions moves twenty per cent against you simultaneously. If the programme still shows a positive five year net present value under the shock test, the board can approve without a long argument about any single input.

Run a tornado analysis to rank sensitivity. For most centres, the ranking is remarkably stable. Wage inflation and attrition dominate, followed by the pace of migration, then the retained onshore organisation, then real estate, then the mark up, then currency. Real estate and the mark up attract disproportionate debate and almost never change the answer. Say so in the pack, and the discussion moves to the assumptions that matter.

8. A tour of the seventeen calculators and what each one answers

The tools at /calculators are built to be used in sequence during a single working session. Each answers one question and hands its output to the next. All of them export to a Portable Document Format board pack, support fifty reporting currencies and default to United States dollars.

• Best Fit GCC City, which region, industry and priority weighting point to which Indian city. Start here if the location is not yet fixed. Deep dive at /gcc-city-finder.
• Cost savings calculator, the headline arbitrage between the home country baseline and the selected Indian city for a given headcount and role mix.
• Fully loaded cost per seat builder, the layer one stack described above, component by component.
• Five year total cost of ownership, comparing home country delivery, third party vendor delivery and a captive centre on a like for like basis.
• Payback and break even calculator, which converts the one time envelope and the monthly run rate delta into a cash payback month.
• Net present value and internal rate of return model, for groups whose capital allocation committee requires a discounted measure.
• Attrition cost calculator, converting an attrition percentage into an annual cash and capacity cost.
• Ramp and hiring plan simulator, which turns a target headcount and a start date into a month by month productive capacity curve.
• Seat and space planner, which converts headcount, shift pattern and desk sharing ratio into square feet, fit out capital and monthly facility cost.
• Entry model comparison, direct build against managed centre against Build Operate and Transfer, on cost, speed, risk and control.
• Transfer pricing mark up estimator, which shows the charge to the parent and the Indian tax cost at a chosen mark up.
• Capital expenditure against operating expenditure comparator, for boards optimising for balance sheet treatment rather than absolute cost.
• Currency and wage inflation drift model, which projects the arbitrage forward five years under chosen rupee and wage assumptions.
• Vendor to captive transition calculator, which finds the headcount at which a captive overtakes a managed service contract.
• Tier 2 hub and spoke model, which compares a single Tier 1 centre with a Tier 1 plus Tier 2 split.
• Productivity and automation adjustment, which applies a generative Artificial Intelligence uplift to a baseline headcount requirement.
• Sector benchmark comparator, which places your modelled cost per seat against observed ranges for accounting, banking, pharmaceutical and life sciences, engineering, semiconductors and technology centres.

Used in that order, the sequence takes about ninety minutes and produces a defensible first draft business case. Send the exported pack with your enquiry at /enquire and a senior partner will return a challenged version, with our own benchmarks substituted for the defaults, within one business day.

9. Worked example, a 250 seat finance and technology centre

Consider a mid market group headquartered in the United States with revenue of 1.8 billion dollars. It plans a 250 seat centre in Pune, split 140 finance and accounting seats, 80 technology seats and 30 seats across data, human resources operations and centre management. The home country baseline for the same scope is 210 Full-Time Equivalents, because the Indian design absorbs some additional volume and a retained onshore layer of 22 people remains.

• Home country fully loaded cost, 210 seats at an average of 128,000 dollars, is 26.9 million dollars a year.
• India fully loaded cost, 250 seats at a blended 31,500 dollars, is 7.9 million dollars a year.
• Retained onshore organisation, 22 seats at 141,000 dollars, is 3.1 million dollars a year.
• Transfer pricing mark up at 18 per cent on the Indian operating cost base adds 1.4 million dollars.
• Steady state total cost of the India model is therefore 12.4 million dollars against a 26.9 million dollar baseline, a reduction of 53.9 per cent, or 14.5 million dollars a year.
• One time transition envelope through a managed entry is 1.6 million dollars, spread across months one to fourteen.
• Attrition cost at sixteen per cent in Pune, at an average replacement cost of 0.55 times the fully loaded seat cost, is 2.2 million dollars a year and is already included in the India fully loaded figure above.
• Cash payback, accounting for the ramp curve in which only 30 per cent of seats are productive by month six and 88 per cent by month fourteen, falls in month 17.
• Five year net present value at a 10 per cent discount rate is approximately 47 million dollars.
• Under the shock test, with wage inflation at 11 per cent, attrition at 21 per cent, hiring twenty per cent slower and the retained organisation at 30 people, payback moves to month 23 and the five year net present value falls to approximately 33 million dollars. The programme remains clearly positive, which is the answer the board needs.

Notice what the example demonstrates. The headline salary arbitrage suggests savings above seventy per cent. The defensible number after the retained organisation, the mark up and the attrition cost is fifty four per cent. That gap is not a disappointment. It is the difference between a business case that survives year two and one that is quietly rewritten.

10. City sensitivity, what changes between Pune, Mumbai, Bangalore and Hyderabad

Run the same 250 seat scope through four cities and the model moves in predictable ways. Salary is the largest driver, followed by attrition, then real estate, then the speed at which the leadership layer can be hired.

• Pune. The lowest total cost of ownership of the Tier 1 cities for finance, accounting, engineering research and development and manufacturing linked work. Attrition of 14 to 19 per cent, deep Chartered Accountant and engineering supply, strong German and Japanese captive precedent, and a real estate cost roughly 25 to 35 per cent below the Mumbai Bandra Kurla Complex benchmark. Detail at /why-pune.
• Mumbai. The right answer where the work is banking, capital markets, insurance, treasury or regulator facing. Higher salary and real estate, offset by proximity to the Reserve Bank of India, the Securities and Exchange Board of India, the exchanges and the entire Indian financial ecosystem. Detail at /why-mumbai.
• Bangalore. The premium city, 15 to 22 per cent above Pune on fully loaded cost for comparable technology roles, with the highest attrition. It is the correct choice when the differentiator is senior product engineering, Artificial Intelligence research or platform depth that no other city can supply at speed. Detail at /why-bangalore.
• Hyderabad. The strongest combination of cost, pharmaceutical and life sciences depth, semiconductor design capability and infrastructure quality, with attrition close to Pune levels. Detail at /why-hyderabad.

The practical guidance is to choose the charter first, the city second and the entry model third. A model that optimises for the lowest cost per seat without regard to whether the city can supply the required talent depth will produce a very attractive spreadsheet and a very disappointing centre. Use /gcc-city-finder to test the fit before the cost model is finalised.

11. Tier 2 cities and the hub and spoke ROI case

The Tier 2 case has strengthened materially. Ahmedabad, Kochi, Coimbatore, Indore, Jaipur, Bhubaneswar and Surat now offer 30 to 45 per cent lower fully loaded cost than Bangalore for comparable mid level roles, attrition in the eleven to fourteen per cent band, and Grade A managed space at a fraction of Tier 1 rents. What they do not yet offer at scale is a deep senior leadership pool or the ability to hire fifty experienced specialists in ninety days.

That asymmetry produces a specific design. Place the leadership layer, the architecture and design functions and the regulator facing work in a Tier 1 hub. Place the volume delivery, the twenty four hour support coverage and the process heavy work in a Tier 2 spoke. A hub and spoke split of roughly 40 per cent Tier 1 and 60 per cent Tier 2 typically improves five year total cost of ownership by 12 to 19 per cent against a single Tier 1 centre of the same size, while reducing weighted attrition by three to five percentage points. Model it at /tier-2-hub-and-spoke and read the city profiles at /tier-2-gcc-cities-india.

12. Entry model economics, direct build against managed against Build Operate and Transfer

The entry model changes the shape of the cash flow more than it changes the destination. Over five years the three models converge within a few percentage points on total cost. Over the first eighteen months they differ enormously on risk, speed and capital.

• Direct build. The parent incorporates, leases, hires and operates. Highest control from day one, highest execution risk, longest time to a first productive team at twenty four to thirty six weeks, and the largest one time envelope. Appropriate where the group already has an Indian entity and India experienced leadership.
• Managed centre. The partner provides the entity, employment, premises, technology and enabling functions while the parent directs the work. Eight to twelve weeks to a first productive team, the lowest capital at entry, and the cost converted to a predictable operating expense. Appropriate for first centres, mid market entrants and teams below roughly 150 seats.
• Build Operate and Transfer. The partner builds and operates, then transfers the entity, people and assets to the parent at a pre agreed formula, typically between month eighteen and month thirty six. Combines a fast start with eventual full ownership. The two variables that decide whether it works are the transfer price formula and the employee continuity terms, both of which must be fixed in the original agreement rather than negotiated later.

In the model, represent the choice as three separate cash flow profiles rather than three cost per seat numbers. The board is choosing a risk profile. Compare them at /managed-gcc-vs-bot-vs-direct and /engagement-models.

13. Capital expenditure against operating expenditure and what the Chief Financial Officer actually optimises

Two groups with identical operations can reach opposite conclusions because they optimise different measures. A group under pressure on free cash flow and return on capital employed will prefer a managed model that converts fit out capital, security deposits and technology assets into a monthly operating charge. A group with abundant capital and a long horizon will prefer to own the entity and the assets, because the five year cost is lower and the balance sheet treatment is irrelevant to its investors.

Model both explicitly, and show the earnings before interest, taxes, depreciation and amortisation impact alongside the cash impact. Lease accounting under Indian Accounting Standard 116 and the equivalent international standard means an owned long lease will bring a right of use asset and a lease liability onto the balance sheet, which sometimes surprises a treasurer late in the approval process. Detail at /opex-vs-capex-gcc.

14. Currency, wage inflation and the five year drift

Two forces work against the arbitrage over time and one works for it. Indian wage inflation for Global Capability Centre roles has run at 8 to 11 per cent for technology and 7 to 9 per cent for finance and accounting, materially above home country wage inflation of 3 to 5 per cent. Left unmodelled, that compression alone would erode a fifty five per cent arbitrage to roughly forty two per cent over five years.

Working the other way, the Indian rupee has depreciated against the United States dollar at a long run average of roughly 2.5 to 3.5 per cent a year, which offsets a meaningful part of the wage compression for dollar reporting parents. For euro, pound sterling, Swiss franc, Japanese yen and Australian dollar reporting parents, the offset differs and must be modelled in the reporting currency rather than assumed.

The third force is mix. As a centre matures it takes on more senior and more valuable work, which raises the average cost per seat while raising the value delivered far faster. A model that projects a constant role mix for five years will show cost creep and miss the value story entirely. Model the role mix shift explicitly, and report both cost per seat and cost per unit of output.

15. Sector nuances, accounting, banking, pharmaceutical and life sciences, engineering and semiconductors

Accounting and finance

The highest and most predictable arbitrage of any function, with the shortest ramp and the deepest talent supply. Pune and Hyderabad lead on cost per seat, and Chartered Accountant supply is the binding constraint at senior levels rather than at entry level. Controls, segregation of duties and internal financial controls documentation are the model's hidden cost lines. Sector detail at /finance-gcc-india and /accounting-gcc-pune.

Banking, financial services and insurance

Higher cost per seat, higher governance overhead, and a regulator facing retained organisation that stays larger for longer. Model outsourcing governance, business continuity, and audit rights as explicit cost lines rather than as assumptions. Mumbai is usually the answer. Detail at /hubs.

Pharmaceutical and life sciences

Good Practice quality systems, computer system validation and regulated data handling add 6 to 11 per cent to the cost stack against a comparable technology seat, and they lengthen the ramp by four to eight weeks. Hyderabad has the deepest supply, with Mumbai, Pune and Bangalore all credible. Detail at /pharma-gcc-india and /life-sciences-gcc-india.

Engineering research and development

Capital intensity is higher because laboratories, test rigs, licences for computer aided design and simulation tools and secure data rooms are real money. Software licence cost per engineer can exceed twenty per cent of salary in simulation heavy disciplines and must sit in the stack. Pune and Bangalore lead. Detail at /engineering-rd-gcc-india.

Semiconductors

The most licence intensive and the most seniority dependent model of all. Electronic design automation licences, emulation hardware and secure intellectual property handling can equal or exceed the salary line for a small advanced node team. The ROI case rests on capability access, not cost. Hyderabad, Bangalore and Pune are the credible locations. Detail at /semiconductor-gcc-pune.

Technology and software as a service

The widest cost band, because the answer depends entirely on seniority. A model that assumes a Bangalore staff engineer costs the same as a Pune mid level developer will be wrong by a factor approaching two. Detail at /product-gcc-india.

16. What generative Artificial Intelligence does to the denominator

Generative Artificial Intelligence has changed the capacity question more than the cost question. A 200 seat accounting centre in 2026 produces roughly the throughput that a 320 seat centre produced in 2022, because reconciliation, variance commentary, document extraction, first draft policy work, test script generation and code review have all been substantially assisted. The consequence for the ROI model is threefold.

• The headcount required for a given scope is lower, so the savings against a home country baseline are larger in absolute terms even though the cost per seat is unchanged.
• The role mix shifts upward, because the assisted work removes the most junior tasks first. Plan for a higher average seniority and therefore a higher average cost per seat, and say so explicitly rather than letting it appear as an overrun.
• The value case shifts from cost per seat to output per dollar. A board that approves on cost per seat alone will be measuring the wrong thing by year three.

Model the uplift conservatively. A 12 to 22 per cent effective capacity uplift on transactional work and 6 to 12 per cent on judgement heavy work is defensible for 2026-2027. Claims materially above that band invite challenge and are hard to evidence in a first year. Further reading at /new-finance-roles-ai-gcc and /ai-ml-gcc-india.

17. The board pack, seven slides that get an approval

• Slide one, the decision requested, in one sentence, with the capital envelope and the approval date required.
• Slide two, the charter. What the centre will own in year one, year three and year five, expressed as capabilities rather than headcount.
• Slide three, the location recommendation with the two rejected alternatives and the reason each was rejected.
• Slide four, the cost stack and the steady state saving, with the retained organisation and the transfer pricing mark up visible on the face of the slide.
• Slide five, the cash curve, showing the one time envelope, the ramp, the payback month and the five year net present value.
• Slide six, the scenario grid and the shock test, with the tornado chart ranking sensitivity.
• Slide seven, the risks and the mitigations, naming the owner of each, with the entry model shown as the principal risk mitigation instrument.

Keep the model itself out of the pack and available in the appendix. A board that asks to see the model is asking whether it can trust the presenter, and the fastest way to answer is a clean, sourced, single tab summary with every assumption dated and attributed.

18. Twelve modelling errors we see every quarter

• Comparing Indian salary to home country salary rather than fully loaded cost to fully loaded cost.
• Omitting the India management and enabling layer, which is 9 to 13 per cent of the direct cost base.
• Assuming the home country organisation reduces by one hundred per cent of the migrated headcount.
• Drawing the ramp as a straight line and ignoring the ninety day notice period convention.
• Treating attrition as a human resources metric rather than a cost line.
• Forgetting the transfer pricing mark up entirely, which overstates savings by roughly the mark up percentage.
• Using 2,080 annual hours per seat rather than 1,700 to 1,780 productive hours.
• Excluding software and design tool licences in engineering and semiconductor models where they can rival the salary line.
• Holding role mix constant for five years, which hides both the cost creep and the value uplift.
• Modelling in Indian rupees and converting once at a spot rate, rather than modelling in the reporting currency with an explicit rate path.
• Ignoring the working capital cost of the Goods and Services Tax input credit refund cycle.
• Presenting a single point estimate rather than a range, which converts a decision meeting into an assumptions meeting.

19. Governance, who owns the model after approval

The model does not stop being useful when the board approves. It becomes the baseline against which the programme is measured. Appoint a single owner, refresh it monthly for the first year and quarterly thereafter, and report actual against modelled on four measures only: productive headcount against plan, fully loaded cost per seat against plan, cumulative cash against the approved envelope, and regretted attrition against the city band. Four measures, one page, every month. Centres that do this correct course in weeks. Centres that do not discover the variance at the annual review, by which time it is a year old.

Publish the assumption log alongside the model. Every input should carry a source, a date and an owner. When an assumption changes, the log records who changed it and why. This single discipline is the difference between a model that is trusted in year three and one that has quietly become a spreadsheet nobody opens.

20. Frequently asked questions

What savings should a first India Global Capability Centre expect?

For a 150 to 300 seat centre against a United States, United Kingdom, Western European or Australian baseline, a defensible steady state figure after the retained organisation and the transfer pricing mark up is 48 to 64 per cent. Against Singapore, Dublin or Tel Aviv baselines, expect 35 to 52 per cent.

How long is payback?

Most 100 to 250 seat centres reach cash payback between month 14 and month 22. A managed entry typically pays back three to six months earlier than a direct build because the one time envelope is smaller and the ramp starts sooner.

What discount rate should we use?

Use the group weighted average cost of capital. Most groups modelling India centres use 9 to 12 per cent. The choice rarely changes the decision because the cash profile is front loaded after the first eighteen months.

Should the model be built in Indian rupees or the reporting currency?

Build the cost base in Indian rupees, because that is where the costs are incurred and inflate, then convert with an explicit annual rate path into the reporting currency. Never convert once at a spot rate.

How do we handle the transfer pricing mark up in the savings figure?

Include it in the cost of the India model. The parent pays cost plus mark up, so the mark up is a real cost of the arrangement, not an accounting adjustment.

What attrition rate should we assume?

Use the city band rather than a group average. Pune and Hyderabad 14 to 19 per cent, Mumbai 16 to 21 per cent, Bangalore 18 to 24 per cent, Tier 2 hubs 11 to 14 per cent. Add two to four percentage points for the first eighteen months of a new centre.

How much does a seat cost fully loaded?

For a mid level professional in 2026-2027, broadly 24,000 to 34,000 United States dollars in Pune and Hyderabad for finance roles and 30,000 to 58,000 for technology roles across the Tier 1 cities depending on city and seniority. Build the number component by component rather than accepting a single quoted figure.

What one time investment should we budget?

Between 1.1 and 1.9 million United States dollars for a 250 seat managed entry, and 2.2 to 3.6 million for a direct build with an owned entity and self managed facility.

Does the model change for a Build Operate and Transfer structure?

Yes, in two places. The operating cost during the build and operate phase carries the partner fee, and the transfer event carries a defined price. Both belong in the cash flow. The five year total usually lands within a few percentage points of a direct build.

How do we model the retained onshore organisation?

As a percentage of migrated Full-Time Equivalents, at 8 to 15 per cent in year one, tapering to 4 to 8 per cent by year three for transactional work and remaining at 10 to 14 per cent for regulator facing work.

What productivity uplift can we claim from automation and Artificial Intelligence?

A defensible band for 2026-2027 is 12 to 22 per cent on transactional work and 6 to 12 per cent on judgement heavy work. Claim less in year one and revise upward with evidence.

Is a Tier 2 city genuinely cheaper once everything is counted?

Yes for delivery heavy work, with a 30 to 45 per cent lower cost per seat and lower attrition. No for senior specialist hiring at pace. The hub and spoke design captures the benefit without the constraint.

How should we compare a captive with a third party vendor?

On five year total cost of ownership including the vendor mark up, change request pricing, transition costs at both ends and the value of retained intellectual property and process knowledge. The cross over usually falls between 40 and 80 Full-Time Equivalents.

What is the effect of Pillar Two on the model?

For groups within scope, the India entity's effective tax rate on the cost plus mark up is generally at or above the fifteen per cent minimum, so the top up is usually immaterial. Confirm with the group tax director rather than assuming.

Do we need a Special Economic Zone or a Software Technology Parks of India registration to make the numbers work?

No. Software Technology Parks of India registration is useful administratively for most services centres. A Special Economic Zone is worth evaluating only at larger scale with a long lease horizon and rarely changes the ROI conclusion by itself.

How do we model shift premia for follow the sun coverage?

Add 10 to 25 per cent to base salary for permanent night shift roles, plus transport and meal cost, plus a two to four percentage point attrition penalty. Model it as a distinct role category rather than blending it into the average.

What real estate assumption should we use?

Model 60 to 80 square feet per seat for a modern hybrid layout with collaboration space, and confirm the local Grade A rent for the specific micro market rather than the city average, which can differ by forty per cent within one city.

How accurate are the published calculators?

They are planning grade, built on current benchmarks and designed to produce a defensible first draft. Before board submission, they should be challenged with role specific quotes, a live real estate quote and your own group tax position. We do that challenge as part of a first response to an enquiry.

Can the model be exported for a board pack?

Yes. Every calculator exports to a Portable Document Format pack with the assumptions visible, which is the form Finance teams prefer to review.

Which currency do the tools default to?

United States dollars, with fifty reporting currencies supported so that the model can be presented in the currency the board actually thinks in.

How often should the model be refreshed after approval?

Monthly for the first twelve months, then quarterly. Report four measures only: productive headcount, cost per seat, cumulative cash against the envelope and regretted attrition.

Who should own the model inside the group?

Financial Planning and Analysis owns the model, the programme director owns the transition schedule, the group tax director owns the mark up and the operations lead owns the productivity assumptions. One name against each layer.

What is the single most common reason a business case is rejected?

Not the size of the saving. It is the absence of a credible answer to what happens if the assumptions are wrong. The shock test solves this.

Can you review a model we have already built?

Yes. Send it through /enquire and we will return a challenged version with our benchmarks substituted, a list of the lines we would add and a view on the payback month, within one business day.

What is the fastest way to start?

Run the Best Fit GCC City tool, then the fully loaded cost builder, then the payback calculator, in that order. Ninety minutes gives you a first draft that is good enough to circulate internally.

21. Next steps

• Run the seventeen decision tools at /calculators and export the board pack.
• Test the location assumption at /gcc-city-finder before the cost model is finalised.
• Compare entry models at /engagement-models and /managed-gcc-vs-bot-vs-direct.
• Check role level pay assumptions at /india-gcc-salary-benchmarks.
• Read the week by week execution plan at /gcc-launch-30-60-90.
• Review the tax structure at /transfer-pricing-safe-harbour-brief and /holding-structure-india-entry.
• Send your scope, target headcount, target go live and preferred city to /enquire for a challenged model and a two page location note within one business day.

References and further reading

• National Association of Software and Service Companies (NASSCOM) Strategic Review: nasscom.in
• Reserve Bank of India statistics and Foreign Exchange Management Act Master Directions: www.rbi.org.in
• Ministry of Corporate Affairs, Government of India: www.mca.gov.in
• Central Board of Direct Taxes safe harbour rules and Income Tax Department: www.incometax.gov.in
• Goods and Services Tax Network: www.gst.gov.in
• Software Technology Parks of India: stpi.in
• Invest India, national investment promotion agency: www.investindia.gov.in
• Organisation for Economic Co-operation and Development transfer pricing guidelines and BEPS Pillar Two: www.oecd.org/tax
• International Monetary Fund World Economic Outlook for currency and inflation paths: www.imf.org/en/Publications/WEO
• World Bank open data: data.worldbank.org
• Ministry of Electronics and Information Technology, Digital Personal Data Protection Act 2023: www.meity.gov.in/data-protection-framework
• ChirayuGCC decision tools: /calculators

Closing read. The ROI calculator does not make the decision. It makes the decision defensible. Build the five layers, source every input, publish the range rather than the point, and the approval conversation moves from arithmetic to strategy, which is where it belongs. ChirayuGCC is the Integrated Partner that builds, runs and, where you want it, transfers your India centre under a single accountability line. Talk to us at /enquire. Jai Shri Krishna.