Business Case

Cost Benchmarks and Sensitivity Analysis

Detailed cost model with line-item justification benchmarked against historical UK government IT programmes, including best/likely/worst case sensitivity analysis and 10-year TCO comparison.

This analysis addresses the Treasury concern: "Government IT programmes routinely exceed estimates by 3-5x. Why should this be different?" We benchmark against actual UK programmes and include appropriate contingencies.

1. UK Government IT Programme Benchmarks

Historical data from NAO reports on major UK government IT programmes provides essential benchmarking context for estimating sovereign cloud costs.

Major UK Government IT Programme Performance
Programme Original Estimate Final Cost Overrun Duration Source
NHS National Programme for IT (NPfIT) £2.3 billion £12.7 billion 452% 2002-2011 (terminated) NAO/PAC Reports
Universal Credit £2.0 billion £15.8 billion 690% 2013-2025 (ongoing) NAO 2024
UK Borders (e-Borders) £742 million £1.1 billion+ 148%+ 2007-2015 NAO Reports
Rural Payments Agency IT £75.8 million £350 million+ 362%+ Multiple failures NAO Reports
HMRC ASPIRE £2.8 billion (10yr) £8.5 billion 204% 2004-2017 NAO Reports

Key Finding: Major UK government IT programmes have historically overrun by 148% to 690% of original estimates. The median overrun is approximately 300% (3x original estimate). This informs our "worst case" scenario.

1.1 Why Overruns Occur

Factor Typical Impact ISCC Mitigation
Scope creep 50-100% cost increase Fixed capability catalogue; change control via TSC
Requirements instability 30-80% cost increase Mature open-source base; proven patterns from existing deployments
Integration complexity 40-100% cost increase Standard APIs; Kubernetes-native; containerised workloads
Supplier failure 20-50% cost increase Multi-supplier model; no single vendor dependency; open source fallback
Skills shortages 20-40% cost increase International talent pool; competitive compensation; training investment
Political changes Programme restart/delay Cross-party support for sovereignty; international treaty commitment

2. Detailed Cost Model: United Kingdom

Line-item cost breakdown for UK sovereign cloud implementation, with three scenarios: Best Case (efficient delivery, no major issues), Likely Case (moderate overruns, typical challenges), and Worst Case (significant overruns, major challenges).

UK Cost Model - 7 Year Programme (£ millions)
Cost Category Best Case Likely Case Worst Case Justification
Infrastructure (datacentres, hardware) £1,500 £2,200 £3,500 3-5 regional datacentres; compute/storage at scale
Platform development (CloudStack, services) £800 £1,200 £2,000 35 capability services; integration; testing
Application refactoring/migration £1,200 £2,500 £5,000 Legacy system modernisation; data migration; testing
Security and compliance £300 £500 £800 Certification; SOC; penetration testing; hardening
Training and organisational change £200 £400 £700 10,000+ staff; £40-70k per person equivalent
Programme management and governance £150 £300 £500 PMO; assurance; gateway reviews
Parallel running costs £400 £600 £1,000 Dual operation during transition; US cloud costs continue
International coordination (ISCC share) £100 £150 £200 UK contribution to shared programme (25%)
Subtotal (base) £4,650 £7,850 £13,700
Contingency (% of base) £700 (15%) £1,570 (20%) £4,110 (30%) Aligned with IPA guidance
TOTAL £5,350 £9,420 £17,810

Comparison to Previous Estimates: The Business Case summary showed £5-10 billion. This detailed analysis produces £5.4-17.8 billion range. The upper bound reflects realistic government IT overrun experience. We recommend budgeting to the Likely Case (£9.4 billion) with Treasury approval for contingency access up to Worst Case if needed.

3. Training Cost Deep Dive

The CRITICAL-GAPS document noted: "Training: €100-300M for 'tens of thousands of staff' = €2-10k per person (should be €40-60k)." Let's validate this.

Training Cost Benchmarks
Training Type Cost per Person Staff Count (UK) Total Cost
Deep reskilling (cloud architects, SREs) £40,000-60,000 2,000-3,000 £80-180 million
Technical upskilling (DevOps, developers) £15,000-25,000 5,000-8,000 £75-200 million
Awareness/familiarisation (users, managers) £2,000-5,000 20,000-30,000 £40-150 million
TOTAL (UK only) 27,000-41,000 £195-530 million

Updated estimate: £200-500 million for UK training (vs original €100-300M combined). This aligns with the £40-70k per person requirement for meaningful reskilling identified in the gap analysis.

4. Application Refactoring Reality Check

The gap analysis noted: "Application refactoring: Listed as €500M-2B, likely €5-15B based on NHS/Universal Credit comparisons."

Application Migration Cost Factors
Application Type Migration Approach Cost per Application Typical Count (UK)
Cloud-native (containers, K8s) Lift and shift £50k-200k 500-1,000
Modernisable (VMs, standard DBs) Refactor to containers £200k-1M 1,000-2,000
Legacy (mainframe, proprietary) Replatform/rewrite £5M-50M+ 100-300
Critical national infrastructure Full replacement £50M-500M+ 10-30

Revised UK Application Migration Estimate:

  • Cloud-native: 750 apps × £125k avg = £94 million
  • Modernisable: 1,500 apps × £600k avg = £900 million
  • Legacy: 200 apps × £20M avg = £4 billion
  • Critical infrastructure: 20 systems × £200M avg = £4 billion

Total: £9 billion range (vs original £500M-2B)

This aligns with the gap analysis estimate of €5-15B (approximately £4-13B). Our Likely Case of £2.5B assumes prioritisation and not migrating all legacy systems in Phase 1.

5. 10-Year Total Cost of Ownership Comparison

Comparison of continued US cloud spend vs sovereign cloud investment over 10 years.

UK 10-Year TCO Comparison (£ billions)
Cost Component Continue US Cloud Sovereign Migration Difference
Year 1-7: Cloud infrastructure £7.0 £5.8 (parallel + sovereign) -£1.2
Year 8-10: Cloud infrastructure £3.3 £2.4 -£0.9
Migration investment £0 £9.4 (Likely Case) +£9.4
Risk cost (expected value) £8.0-15.0 £0.5 -£7.5 to -£14.5
TOTAL 10-YEAR TCO £18.3-25.3 £18.1 -£0.2 to -£7.2

Key Finding

Even in the Likely Case, sovereign cloud TCO over 10 years is lower than continued US cloud dependency when risk costs are included. The investment pays for itself in risk mitigation alone.

6. Sensitivity Analysis Summary

Monte Carlo simulation with 10,000 iterations, varying cost factors within defined ranges:

Percentile UK Total Cost (£B) Interpretation
10th (P10) £5.8 Best case scenario with efficient delivery
25th (P25) £7.2 Optimistic but realistic
50th (P50/Median) £9.1 Most likely outcome
75th (P75) £12.4 Significant challenges encountered
90th (P90) £16.2 Major overruns (historical norm for gov IT)
95th (P95) £19.5 Near worst case

Recommendation for Treasury: Budget to P50 (£9.1B) with delegated authority to access contingency up to P75 (£12.4B). Require Cabinet/Treasury approval for spend beyond P75. This provides 75% confidence of delivery within budget while maintaining controls for exceptional circumstances.

7. Combined Jurisdictions (Likely Case)

Multi-National Investment Summary - Likely Case
Jurisdiction Best Case Likely Case Worst Case
United Kingdom £5.4B £9.4B £17.8B
European Union €22B €42B €85B
Canada C$6B C$11B C$22B
Australia A$8B A$14B A$28B
COMBINED (USD equiv.) ~$45B ~$85B ~$170B

8. Exit Strategy and Stranded Capital Analysis

Every major investment requires defined exit criteria. The Cooperative proposal includes explicit NO-GO gates at the pilot phase to limit losses if the initiative proves infeasible.

8.1 Pilot Phase Exit (Phase 0)

NO-GO Scenario: Maximum Stranded Capital

Component Pilot Investment Recoverable Value Stranded
Infrastructure (hardware) €320M €80M (resale/reuse) €240M
Personnel €80M €0 (skills retained in workforce) €80M
Migration work €50M €0 (sunk) €50M
Exit costs €28M €0 €28M
TOTAL €478M €80M €398M

8.2 Risk-Adjusted Value of Pilot

Scenario Probability Outcome Expected Value
GO decision 70% De-risked €85B programme; pilot costs absorbed +€85B programme enabled
CONDITIONAL 20% Additional 6-month remediation; programme proceeds +€50M additional, programme proceeds
NO-GO decision 10% €398M stranded; lessons learned; avoided €170B worst case -€398M but avoided catastrophe

Expected Value Calculation:
(70% × €0) + (20% × -€50M) + (10% × -€398M) = -€50M expected pilot cost

This is the insurance premium for de-risking an €85B programme. The pilot costs ~0.6% of the full programme to validate feasibility before committing to scale.

8.3 Full Programme Exit (Phase 1+)

If the initiative is abandoned after Phase 0 approval, stranded capital scales with investment:

Exit Point Invested Recovery Rate Stranded Capital
End of Year 1 (Phase 1) €2.5B 25% €1.9B
End of Year 3 €12B 30% €8.4B
End of Year 5 €35B 40% €21B
Year 7+ (operational) €85B 50%+ €42B (but sovereign value realised)
Warning Abandonment after Phase 0 approval represents progressively larger sunk costs. This is why the pilot phase is critical—it's the last low-cost exit point. Treaty exit clauses require 24 months notice and stranded capital contribution.

8.4 Comparison: Exit Cost vs Risk Cost

Stranded capital from initiative failure must be compared against the expected cost of cloud service disruption if dependencies continue:

Scenario Expected Cost (10-year) Notes
Pilot NO-GO €398M Controlled loss; lessons learned
Full programme failure (Year 5) €21B Significant loss; Munich scenario
Service denial event (no migration) €50-150B GDP impact of government service disruption

Even a failed initiative with €21B stranded capital is preferable to a €50B+ service denial event. The pilot's €398M maximum loss is trivial compared to either outcome.

Full exit strategy details: See Pilot Programme Exit Strategy.

9. Sources

Back to Business Case
View Risk Register