Offline Data Fusion Engine
A desktop application that lets enterprise users merge, map, transform and export data from up to twenty spreadsheets into a single standardised output — fully offline, with no cloud, no database and no login.
Combining and transforming spreadsheets from many sources into one standardised output — entirely on the analyst's machine, with nothing sent anywhere.
A desktop application with no server, no database and no authentication — data never leaves the machine, which removes the security objection entirely rather than mitigating it.
Two modes cover the real cases: appending files with similar structures, and joining files with different structures through explicit column mappings.
Fuzzy matching suggests source columns for each output column, so the analyst corrects suggestions instead of building every mapping from scratch.
Mappings are saved as reusable templates, turning a recurring multi-hour task into a repeatable one.
A validation layer surfaces unmatched rows, missing keys and incomplete mappings before export — catching silent errors while they are still cheap.
Analysts were combining spreadsheets from multiple systems by hand — copying, pasting, VLOOKUPs, and rebuilding the same mapping every reporting cycle. It was slow, and errors were invisible until someone acted on the output.
The obvious fix, a cloud tool, was unavailable: the data was commercially sensitive and could not leave the machine.
Existing offline options assumed technical users. This had to work for the analysts who actually do the task, without training.
A desktop application that lets enterprise users merge, map, transform and export data from up to twenty spreadsheets into a single standardised output — fully offline, with no cloud, no database and no login.
Where data is too sensitive for a cloud tool but the manual process is costing real hours, a local-first application is often the answer nobody proposed.
The constraint that shaped it
No cloud, no database, no login
The constraint that shaped everything: the data could not leave the machine, and the users were analysts rather than engineers. That ruled out the entire class of hosted tools and forced a genuinely offline desktop architecture — which turned out to be the reason it was adopted.
Preview before generate
Analysts see merged output before committing to an export. The failure mode of data tooling is a wrong file that looks right, so the check happens before the artefact exists.
Suggest, then let the human correct
Fuzzy matching proposes mappings rather than deciding them — fast when it is right, harmless when it is wrong.
Templates as the real payoff
The recurring value is not the first merge, it is the fiftieth. Saved mappings turn a repeated manual task into a repeatable one.
- Framed the problem with the analysts doing the work
- Designed the end-to-end experience, including the mapping builder
- Built the desktop application and transformation engine
How We Solved the Problem
Multi-file ingestion
Up to twenty spreadsheets loaded locally, with every sheet and column surfaced so the analyst can see the full shape of their source data.
Append or join
Vertical stacking for similar structures, or inner and left joins across files with different structures via configured key columns.
Mapping builder with suggestions
A reorderable output definition where fuzzy matching proposes the likely source column for each target column.
Validation and preview
Warnings for unmatched rows, missing keys and empty mappings, with a live preview of merged output before anything is generated.
Templates and export
Mappings saved as reusable templates; output written as a formatted, auto-sized spreadsheet.