Requirement matching, with evidence

From a photo on the job site to a cited match in your inventory.

Froga reads a customer's requirement list — however messy it arrives — and matches it against your inventory. Every match links back to the exact line it came from. If it can't find evidence, it says so, instead of guessing.

Source · photo
2x Faro Xenón 35W (D2S)
1x lámpara xenón D1S
4 Bombilla halógena H7
traced
GOOD
Xenon Lamp D2S 35W
SKU XL-100  ·  score 1.00
Evidence: exact code match — 35W, D2S
The problem

Someone is still reading PDFs by hand.

Tender and order requirements arrive as scanned tables, phone photos, and spreadsheets with no shared format. Right now, a person cross-checks every line against inventory by eye — slow, easy to miss, and hard to prove correct after the fact.

01 / SOURCE
Whatever arrives

PDF tables, CSV/XLSX exports, or a photo taken on-site. No format is off-limits, none is assumed clean.

02 / RISK
No audit trail

A manual match has no record of why it was made. When something's wrong, nobody can trace it back to the line that caused it.

03 / COST
Hours, every time

The same fact-hunting work repeats on every tender, with no memory carried over from the last one.

How it works

Three steps. Every one of them checkable.

Froga isn't a chat window. It's a fixed pipeline — the same three steps run every time, in order, and each one leaves something you can inspect.

01

Upload

Drop in the requirement list and your inventory export — PDF, CSV, XLSX, or a photo of a printed page. Froga parses tables where it can and reads text via OCR where it can't.
02

Match

A three-tier engine checks each line: exact part-code first, then fuzzy name similarity, then semantic meaning. Every requirement is tagged GOOD, MAYBE, or NO — never a silent guess.
03

Cite & teach

Every match shows its evidence in plain language. Correct a miss once — teach it a synonym — and the fix applies instantly, and every time after.
Why not just prompt an AI chatbot

Copilot helps people think. Froga decides whether a match is allowed.

A general assistant can summarize a document well. It was never built to enforce a rule, block a bad match, or fail closed when evidence is missing — because none of that is what it optimizes for.

General AI assistant

Optimizes for fluency
  • Reads and summarizes documents well
  • Answers even when it isn't sure
  • No fixed rule it's forced to follow
  • Nothing to audit after the fact

Froga

Optimizes for accountability
  • Every match cites the exact source line
  • No evidence found → says so, not a guess
  • Same three-tier rule, run every time
  • A record for every decision, GOOD or NO
3Match tiers — code, fuzzy text, semantic meaning, checked in order
89%Baseline match accuracy on mixed-language test inventory
100%Accuracy after four corrections taught to the synonym table
<5sTarget time from upload to a full match report
Where Froga stands today

A working prototype, built and running.

Froga is live and handles PDF, CSV, XLSX, and photographed requirement lists end to end, including bilingual Spanish/English inventories. It's currently a founder-built prototype seeking its first pilot partners in procurement and tendering.

Hybrid matching engine OCR for photos & scans Editable synonym table Cited, exportable reports

See it trace a real list, right now.

Upload a requirement file — even a rough photo — and watch Froga match it against inventory with a citation for every line.

Open the live app →