CLIENT STORY
The client is a private investment firm managing $250–400M across public markets, private equity, real estate, and direct investments. A team of two to three people consolidates positions, transactions, capital calls, and distributions from custodians, GP portals, accounting systems, and spreadsheets. With new systems coming online, the firm wanted a data foundation flexible enough to outlast any single tool or vendor.
| Service: | Data Engineering |
|---|---|
| Industry: | Finance |
| Region: | USA |
Read summarized version with
We do have historical data present in a single source of truth. I just don't think it's the source of truth we need going forward. There's a lot of back and forth. I want to create that internally. — Head of the Reporting Function
Most of the formulas were embedded within Power BI. So if you want to change the system, you have to start all over. — Head of the Reporting Function
The Challenge: KPI Logic Locked Inside the BI Tool and SaaS Products
Capital statements, cash calls, and distributions arrived through separate GP portals in different formats, often without APIs, so the two-to-three-person reporting team logged into each portal, downloaded files, and entered figures into spreadsheets every reporting period – the single most time-consuming part of their month. The same position or valuation did not always match across custodians, GP portals, and internal spreadsheets. Resolving each mismatch meant investigating it by hand, without any record of how the team had settled the issue before.
Power BI calculated more than 40 portfolio KPIs from spreadsheets maintained outside the BI tool. No independent data layer held those definitions. Rename a column or restructure a file, and a report could fail without warning. A move to another BI platform would then force the firm to rebuild every formula.
Off-the-shelf portfolio and reporting platforms created a second constraint. The firm was paying for packaged SaaS products but evolving faster than those products could add the features, integrations, and reporting logic it needed. Every new requirement depended on a vendor roadmap rather than the firm’s own priorities.
With a new general ledger rollout underway, additional portfolio tools under evaluation, and a significant liquidity event approaching by year-end – an exit followed by capital deployment – the firm needed its reporting logic documented, validated, and under its own control before that decision point, not after.
A Client-Controlled Data Platform Built Beyond SaaS Constraints
GroupBWT built a governed data platform in the client’s own cloud environment, separating the firm’s data, KPI logic, and reporting workflows from the limitations of any single SaaS or BI product. KPI definitions are documented and versioned, data passes validation before it reaches a report, and every figure stays traceable to its source. The client retains control over its data, code, and core business logic while individual source systems, portfolio applications, and BI tools can change independently.
Unifying the Investment Data Pipeline. GroupBWT connected custodian feeds, accounting systems, portfolio tools, spreadsheets, and GP-portal documents through one ingestion process – standard connectors for sources with existing integrations and custom extraction pipelines for GP statements and other documents, with each extracted record still linked to its original source.
Using AI Where It Helps – Not Where Financial Logic Must Stay Deterministic. GroupBWT kept portfolio calculations and KPI logic inside the governed data layer rather than delegating financial calculations to an AI model. AI is used where unstructured or variable data makes deterministic processing harder: extracting information from PDF fund statements and helping identify unusual or inconsistent records for review. The resulting structured, governed data foundation is also designed to support future natural-language access to portfolio data without moving the underlying calculations into the AI layer.
Validating Data Before It Reaches a Report. Incoming data passes through raw, validated, and reporting layers. The raw layer keeps original records unchanged. The validated layer standardizes formats, maps entities across systems, and checks discrepancies before the reporting layer prepares positions, transactions, valuations, and approved KPI calculations for business use.
Moving KPI Logic Out of Power BI. GroupBWT took more than 40 portfolio KPIs out of dashboard formulas and put them in the governed data layer. Each definition is now documented, versioned, and reusable. Dashboards receive prepared figures rather than calculating them. The firm can therefore replace its BI tool without rebuilding the core KPI calculations.
Making Every Number Traceable. From a portfolio KPI, the reporting team can drill down to an individual position and then to its source record. Finding where a number came from no longer requires a manual search.
Tech stack: Databricks, Delta Lake, Fivetran, Power BI
In the silver layer we run data quality jobs. The data is saved only after it passes that verification.
40+ Portfolio KPIs Moved Out of Power BI
- By deploying the core data and reporting logic in the client’s own cloud environment, GroupBWT reduced dependence on packaged SaaS products and gave the firm direct control over its data, code, KPI definitions, and future platform development.
- By moving more than 40 KPI calculations into the governed data layer, GroupBWT let the firm change BI tools without recreating the core calculations behind its reporting.
- By connecting custodian, GP-portal, accounting, and portfolio data into a single pipeline, GroupBWT reduced the manual downloading and reconciliation work required every reporting period.
- By keeping financial calculations deterministic while using AI for unstructured document processing and anomaly detection, GroupBWT introduced AI without allowing model-generated outputs to become the source of truth for portfolio calculations.
- By adding validation checks ahead of the reporting layer, GroupBWT flags missing or inconsistent source data before it reaches an investment report.
- By keeping every reported figure traceable to its original record, GroupBWT gave the reporting team drill-down from a portfolio KPI to its source record.
- The governed data foundation also creates a structured, traceable base for future natural-language access to portfolio information without moving core financial logic into the AI layer.
Looking to Move KPI Logic Out of Your BI Tool?
We design data platforms that keep your KPI logic out of BI-tool formulas, so changing dashboards or data providers doesn't mean rebuilding the core calculations behind your reporting.
You have an idea?
We handle all the rest.
How can we help you?
