Ponto

Inspiration

The starting point was a daily task: a company owner brings together sales reports from multiple stores to check the overall business. I wanted a simple way to turn separate tables into a report, while letting the person using it decide what to aggregate.

What it does

Ponto is a PC browser application. Users add CSV or value-only Excel files, explicitly map different column names, and select grouping columns and metrics. Supported operations are sum, row count, average, minimum, and maximum. Users can edit output headings, review the result, and download Excel, CSV, or PDF. They can save aggregation conditions in the browser and reuse them with new files, checking changed columns before recalculating.

Excel output additionally includes per-file accepted and excluded row counts and metric values, the effective calculation conditions, and original-to-mapped column names and selected workbook sheets. These records help explain the basis of the result; they do not detect missing stores, duplicate transactions, or accounting-period errors. CSV and PDF contain the result table.

How it was built

I directed the product decisions and iterated on the interface with a coding agent. Codex assisted with implementation and verification. The Devpost Learn Skill Pack and the scope, PRD, specification, and checklist documents were used to structure the work. The implementation uses TypeScript, Vite, ExcelJS, Papa Parse, Decimal.js, and pdf-lib. Calculations and file exports run locally in the browser.

Challenges and observations

The project had to handle inconsistent column names without assuming that similarly named fields mean the same thing. Another challenge was keeping the interface simple while allowing different aggregation fields. The implemented workflow requires review when conditions change and confirmation before export. Automated tests cover calculations, invalid inputs, saved conditions, and output formats.

Current status and limits

The demonstration uses fictional data and shows the working manual aggregation workflow, three export formats, saving reviewed conditions, and verification records rendered from the downloaded workbook. Live runtime AI is not configured or verified; actual microphone recognition is also unverified. An optional AI proposal endpoint and review interface are implemented, but no real AI performance is claimed. Supported inputs are UTF-8 CSV and value-only XLSX within the documented limits. Human time savings and advantages over competing products have not yet been measured.

Controlled verification

A fictional five-store dataset used three CSV files and two value-only XLSX files, with different column names and integer-yen amounts. Ponto produced 3,933,030 yen, 945 accepted transactions, and 3,176 items, excluding 55 cancelled rows. Every store and overall total matched a separate SUMIF/COUNTIF reference workbook evaluated with Artifact Tool. This was not a comparison against the owner's existing business workbook or native Microsoft Excel. Human time savings and competitive superiority remain unmeasured.