SuperMart: Recover Receipts from HTML and Scanned Pages

The challenge

Try two ways to turn receipts into data: read the HTML directly or use OCR on the page images. Compare the results, then answer a few questions about sales and popular products.

Start here

Open the SharePoint folder linked above. SuperMartData.html contains the receipt text. transactionPages.zip is a roughly 1.46 GB archive with 5,698 JPEG page images. transactions_2025-09-06.json is the structured reference; transactions_2025-09-06.csv has malformed quoting around its nested items. SummaryData.pdf is a one-page summary and sample preview.

Open the HTML in a browser and inspect its receipt/pre elements. View a few JPEG members without expanding the entire archive. The first page contains more than one receipt; the summary PDF says it shows six previews rather than the complete collection.

Try it

1. Extract transaction and item records directly from the HTML while preserving receipt boundaries.

2. Design an OCR route for the JPEG pages and compare it on a reviewed sample.

3. Repair the CSV export and validate joins to the JSON reference.

4. Reproduce revenue, customer count, item totals, and product rankings.

What to build

Build normalized tables, a source-to-record mapping, a comparison of HTML extraction and OCR on a small sample, and the analysis answers. Include how many pages you tried and how many receipts you recovered.