Practitioner Case Study: Regulatory Capital Analytics Transformation | Belton Bridge Analytics
Practitioner Experience · Data Engineering & Applied ML

Regulatory Capital Analytics Transformation

How our founder led a three-year internal transformation at a leading investment bank, unlocking £40m+ in reserved capital, cutting reporting cycles in half, and embedding machine learning into financial decision-making.

£40m+
regulatory capital unlocked
50%
reduction in reporting cycle
10 days
removed from monthly close
6 entities
consolidated into one data fabric
14 BUs
served by unified reporting
InstitutionMajor Investment Bank, South Africa (name withheld)
Duration3 years
RoleSenior Data Analyst (employed directly by the institution)
DisciplinesData Engineering · Applied ML · Regulatory Capital Analytics
Portfolio£250m regulatory capital book

The challenge

During his time as a senior analyst within the regulatory capital team at a major South African investment bank, our founder inherited responsibility for a £250m book managed across multiple legal entities using Excel, manual processes, and disconnected reporting tools. Data was fragmented: different systems held variants of what should have been identical numbers, reconciliations were manual and prone to human error given the large processing requirments, and the monthly cycle consumed analyst capacity that should have been focused on insight generation.

More critically, the team had no way to identify where capital was being inefficiently deployed. The opportunity to optimise the book, and release tens of millions in reserved funds, was invisible in the data.

The opportunity: the data existed. The problem was that it was trapped in silos, produced too slowly, and presented in a format that informed rather than enabled decisions. Fixing the infrastructure would unlock both operational efficiency and strategic opportunity.
A note on context: This work was carried out by our founder while employed directly at the institution, not as an external consultant. That context matters: the depth of access, the multi-year commitment, and the institutional knowledge required to deliver this kind of transformation is only possible from the inside. It is shared here as a demonstration of practitioner expertise that informs how BBA approaches similar problems for clients today. The institution name is withheld by mutual agreement.

What was built

  • 1
    Data architecture redesignAudited the existing landscape across 6 legal entities and 14 business units. Designed and implemented a unified data architecture, eliminating silos and creating a single authoritative source of truth for capital reporting.
  • 2
    Pipeline automation (Python, SQL, R)Replaced manual extraction and transformation with automated pipelines. Cut the monthly reporting cycle from 20 business days to 10, a 50% improvement that freed analyst time for insight work.
  • 3
    Executive dashboard suite (Power BI)Built 10 executive-level dashboards replacing manual reporting packs sent by email. Dashboards served Exco, Risk Committee, and Capital Committee with live, governed data.
  • 4
    Machine learning in finance initiativeIntroduced and led an ML adoption programme within the finance team. Applied supervised learning for risk prediction and portfolio optimisation. Built anomaly detection models to flag data quality issues and unusual behavioural patterns.
  • 5
    Capital optimisation analyticsUsed ML-driven insights to identify where capital was inefficiently deployed, thus enabling a targeted programme that released £40m+ in reserved regulatory capital and R1bn+ via associated collateral restructuring.

Technologies used

PythonSQLRPower BIscikit-learnpandasNumPyETL pipelinesMurexOracle ERPSupervised MLAnomaly detection

Before and after

Before
  • 20-day monthly reporting cycle
  • Manual Excel reconciliations — 6 entities
  • Siloed data, different numbers across systems
  • Reports distributed by email as spreadsheets
  • Capital inefficiencies invisible in the data
  • No ML or predictive capability in the team
After
  • 10-day cycle — 50% faster
  • Automated reconciliations, audit-ready outputs
  • Single source of truth across all entities
  • 10 live Power BI dashboards for Exco and committees
  • £40m+ capital released through ML analytics
  • ML embedded — risk prediction and anomaly detection live

Outcomes

£40m+
regulatory capital unlocked through analytics-driven optimisation
50%
reduction in reporting cycle — 10 business days recovered per month
10
executive dashboards replacing manual email reporting packs
6 entities
unified into a single governed data architecture
R1bn+
reserved funds released via associated collateral restructuring
ML live
supervised learning and anomaly detection embedded in capital reporting

The strategic multiplier

The work had the primary benefit of improving reporting efficiency but, additionally, created the analytical foundation that future AI and ML initiatives depend on. Well-structured, well-governed data is the prerequisite for every data-driven initiative that follows. Organisations that invest in this now avoid re-doing it under pressure later and usually at greater expense.

What this enabled

  • Exco and committee decisions based on real-time, governed data
  • Capital optimisation projects impossible without ML visibility
  • Analyst capacity freed for value-add work
  • Audit-ready outputs with full data lineage at year-end
  • Reusable ML framework adopted across subsequent analytics work

Transferable to

  • Finance teams still running month-end on Excel
  • Capital or risk books lacking proper analytics infrastructure
  • Firms wanting ML capability without a dedicated data science team
  • Multi-entity data landscapes with inconsistent reporting
  • Teams where audit preparation is still manual and high-stress

Is your data working hard enough?

If your team is spending time on processes that should be automated, or there's capital sitting locked in your data, let's talk.

Book a 20-min clarity call Email us directly