Automating a global asset manager's data pipeline to streamline decisions

A global asset manager now refreshes enterprise-value metrics across its portfolio in minutes rather than the four to five hours the task previously demanded by hand, after QuantSpark rebuilt a sprawling legacy spreadsheet as an automated Python data pipeline. The change removed a long-standing operational bottleneck, corrected historical records in full, and freed several working weeks a year that had previously gone on manual maintenance rather than investment analysis.

Before the rebuild, the firm's enterprise-value calculations lived in a single spreadsheet that had grown, sheet by sheet, into a tool of more than 20 tabs covering roughly 160 companies across roughly 260 time periods. Years of incremental additions had buried the underlying logic inside cell references and manual overrides that were difficult for anyone outside the original builders to follow.

Every refresh was a manual exercise: pulling new data, re-running formulas, checking for breakages, and reconciling anything that looked wrong. The process was slow, prone to error, and impossible to run in a single pass, so figures were frequently out of date by the time they reached decision-makers. Maintaining the tool consumed several working weeks of analyst time each year, time that could otherwise have gone into investment advice rather than spreadsheet housekeeping. The fragility of the system also meant that any request for a fresh cut of the data, or a change to the underlying logic, carried real operational risk.

QuantSpark's approach kept the emphasis on fidelity to the original logic rather than a wholesale redesign, so the client could trust the new tool from day one.

The team traced how data flowed through every sheet of the legacy tool and documented years of accumulated calculation logic before writing any new code. That logic was reimplemented as one Python pipeline running inside the client's own coding environment, replacing the sprawl of interlinked spreadsheet tabs with a single, auditable codebase. The pipeline draws data directly from the firm's cloud data warehouse through SQL, removing the manual pull and paste step that previously started every refresh. The spreadsheet's formulas were rebuilt in Pandas, preserving the original calculation logic while making it far easier to read, test and change. A rigorous regression and reconciliation process then checked the new pipeline's outputs against the legacy spreadsheet, matching results to within one per cent on all key metrics; the residual differences were traced to known discrepancies between the two source systems rather than to errors in the rebuild. Logging and version control were added throughout, so every run and every change to the pipeline is visible and traceable, and the client's own team can maintain it without QuantSpark in the loop.

In production, the pipeline runs as a straight line from source to decision: data is pulled from the cloud warehouse via SQL, passed through the Pandas calculation engine that mirrors the original spreadsheet logic, checked against the reconciliation baseline, and written to the client's internal database, with each step logged and version controlled.

The rebuild used a categorical, mainstream stack: Python and Pandas for the calculation engine, SQL for data extraction from the client's cloud data warehouse, and a version-control system to track changes, alongside the client's own internal database as the destination for outputs. No proprietary or named third-party platform is referenced in the public record.

The headline change is time: a refresh that took four to five hours by hand now takes minutes. That single shift carries several knock-on effects. Historical records, previously locked into whatever state the spreadsheet last left them in, have been corrected in full. Automated daily refreshes are now practical, where before they were not, which lets the firm model its metrics with greater sensitivity to new data. Version control gives full visibility over every change to the pipeline, closing off a class of silent errors that spreadsheets are prone to. The legacy tool itself has been retired, and the several working weeks a year it used to consume have been freed for analysis rather than maintenance.

The return on this work is best understood as time reclaimed rather than a modelled pound figure: the method is to compare the manual refresh time against the automated one, multiply by however often the update needs to run, and set that against the analyst weeks previously lost to upkeep. QuantSpark has not attached a monetary value to this in the public record, and none should be inferred beyond the time-saving figures the client itself reported.