Case Study - Self-Service Analytics
DataLens BI is a browser-based self-service analytics workspace I built to help business users inspect, clean and analyse CSV and Excel data before making decisions from it, without requiring every question to become a new BI-development request.
A programme manager, finance lead or operations team may receive several Excel exports a few hours before a performance review. The files contain useful information, but nobody yet knows whether the data is complete, duplicated, correctly typed or ready for analysis.
The first business question is not what the dashboard shows. It is whether the data can be trusted enough to make a decision from it.
Everyday analytics still begins with spreadsheets: finance exports, HRIS extracts, service logs, field-collection files and CSV downloads from operational systems.
I framed the product around one practical question:
The goal was not to replace Power BI, enterprise warehouses or professional analysts. It was to solve the space between having a file and understanding what should happen next.
In analytics work, stakeholders often begin with "Can you look at this Excel file?" Before a permanent reporting product is designed, an analyst needs to understand fields, completeness, duplicates, measures, trends, relationships and unusual values.
DataLens makes that exploratory stage easier to perform and easier to explain.
I designed and developed the platform end-to-end.
Load a CSV or Excel file directly in the browser.
Understand columns, types, completeness, duplicates and outliers.
Apply controlled transformations while preserving the raw source.
Explore distributions, segments, relationships and trends.
Download filtered data, cleaned data, pivot output or an analytical report.
DataLens does not use machine learning just to sound intelligent. The analysis is built around transparent statistical methods so the same data and selections produce the same result.
The insight layer converts calculated results into plain-language observations, such as revenue movement, highest average cost by segment or strongest positive relationship between selected measures.
Because these statements come from deterministic calculations, the explanatory layer remains reproducible and avoids unsupported AI-style conclusions.
For a monthly programme review, a manager can check data quality, remove duplicate records, standardise formatting, filter by country, programme and month, then identify whether a delivery issue is concentrated in specific locations.
DataLens is not an enterprise data warehouse, governed semantic model, master-data platform, causal-inference engine, machine-learning platform or secure document-management system.
I designed it for rapid exploratory analytics on local files. Recurring logic should still move into governed datasets, semantic models and controlled BI products.
The public version processes the selected CSV or Excel file inside the user's browser. The source file is not uploaded to an application server for processing.
This reduces server-side data exposure, while still requiring each organisation to follow its own information-security, classification and data-handling controls.
A dashboard assumes the data has already been prepared. DataLens starts earlier and demonstrates the analytical thought process: understand the data, test quality, prepare it, explore patterns, explain the pattern and decide what deeper work is required.
An enterprise version would add authenticated users, organisation workspaces, governed dataset connections, saved sessions, role-based permissions, audit logs, reusable transformation recipes, telemetry, monitoring and integration with Power BI or enterprise semantic models.
DataLens BI helps users turn unfamiliar spreadsheets into understandable analytical evidence before committing time to a full reporting or data-engineering solution.