Grocery Datasets in 2026: 12 Sources Compared

Grocery data is harder to buy than most data.

A grocery catalogue changes every week. Prices move. Pack sizes change. New SKUs appear and old ones vanish. And the same product has a different name in every country.

This guide compares 12 real sources. Free and paid. Global and India. So you can pick one today instead of testing ten.

Free nutrition data: Open Food Facts or USDA FoodData Central. Free price data: WFP Food Prices. Live retail catalogues: Crawl Feeds, which covers 15+ grocery retailers. India quick-commerce: Blinkit and JioMart datasets from specialist vendors. Computer vision: SKU110K or Freiburg Groceries.

What makes grocery data hard

Four things go wrong with most grocery datasets.

Prices go stale. A static CSV from 2023 describes a store that no longer exists. If you are tracking prices, you need weekly refreshes at minimum.

Pack sizes break comparisons. £2.50 for 500g is £5/kg. £3.80 for 750g is £5.07/kg. Without a unit price, you cannot compare anything.

Categories are inconsistent. One retailer files "Greek yoghurt" under Dairy. Another files it under Chilled Desserts. Merging two catalogues is a mapping job, not a merge.

Regional SKUs do not match. "Toor dal" in India and "split pigeon peas" in the UK are the same product with different names and different barcodes.

Any dataset you buy should fix at least two of these. If it fixes none, it is a spreadsheet, not data.

The 12 sources

1. Open Food Facts

A collaborative database of packaged food from around the world. Millions of products, keyed by barcode. You get ingredients, nutrition, labels and brand.

It is free and open-licensed. It is also community-maintained, so some records are thin and some are wrong. Great for text and nutrition work. Not a price source.

2. USDA FoodData Central

The US government nutrition database. Reliable, well documented, free. It covers nutrients per 100g, which makes it easy to join to other catalogues.

It is US-centric and contains almost no retail pricing. For recipe-side text, see our recipe dataset roundup.

3. WFP Food Prices (HDX)

Market-level food prices across 98 countries, some going back to the 1990s. Free, in CSV, with proper licensing.

This is the best free source for price trends. It is not SKU-level and it is not a retailer catalogue.

4. Open Grocery Database Project

About 100,000 grocery products with UPC codes, brands and names. Free to download.

The fields are narrow — no prices, no images, no categories. It is a lookup table, and a good one.

5. Kaggle BigBasket

A popular free sample of an Indian online grocery catalogue. Category, sub-category, brand, SKU name, size, MRP and a product link.

Useful for a prototype. Static, undated in most copies, and small compared with a live feed.

6. SKU110K

11,762 images containing 110,712 labelled products, built for detecting densely packed shelves. Free for research.

If you are training shelf-detection or checkout models, start here. Check the licence before commercial use.

7. Freiburg Groceries

5,000 images across 25 food classes. Small, clean, easy to load.

Good for a first classifier. Too small for a production model.

8. Grocer-Help

A 2026 research set: over 13,000 images, 349 grocery product classes, shot in eight stores across five Indian states. Built for real shelf conditions — clutter, occlusion, odd angles.

If your model must work in an Indian store, this is the closest thing to it. Research licence.

9. Crawl Feeds retailer datasets

Live catalogues from Kroger, Walmart, Target, Tesco, Asda, Aldi, Ocado, Sainsbury's, Safeway, Meijer, Food Lion, Albertsons, Coles and more. Product name, brand, category, pack size, price, unit price and availability.

This is the option when you need current prices across countries. See the full grocery catalogue, or jump straight to Kroger or Tesco.

10. Crawl Feeds Blinkit data

SKU-level quick-commerce products with prices, discounts and stock. India-focused, refreshed on a schedule.

Quick commerce moves faster than any other grocery channel. A quarterly refresh is useless here. See the Blinkit products dataset.

11. Bright Data grocery datasets

A large multi-source catalogue. The vendor states 22 datasets and 629M+ records, from about $250, priced per record.

Volume is the selling point. You will still need to map categories yourself.

12. Actowiz quick-commerce sets

India-focused sets covering Blinkit, Zepto, Swiggy Instamart and JioMart, sold individually at around 50–500 depending on refresh frequency.

Good for a single-platform study. Narrower than a full retailer feed.

The 12 fields a usable grocery dataset needs

Ask for these before you pay. If three or more are missing, walk away.

Free or paid?

The rule: free data is fine for a prototype. The moment you need current prices, cross-country coverage or a commercial licence, free stops working. That switch usually happens in week two.

How to choose in five minutes

1. Only need nutrition or ingredients? Open Food Facts. Stop there.
2. Only need price trends? WFP Food Prices.
3. Building an image model for shelves? SKU110K, then Grocer-Help if you need India.
4. Need current prices from real retailers? Crawl Feeds or Bright Data.
5. Need India quick commerce? Crawl Feeds Blinkit, or Actowiz for other platforms.
6. Not sure? Download one free set, run your first model, then buy the field you are missing.

Need current grocery prices across 15 retailers, with pack sizes and unit prices already parsed? Browse the Crawl Feeds grocery datasets — global retailers and Indian quick commerce in one schema.