Why We Built fdsrch: 739,000 Products and the Allergy Problem Nobody Solved
The FDA maintains one of the most comprehensive food product databases in the world. We made it searchable by allergen in under a second. Here's why that mattered.
The problem with food allergen screening wasn’t the data. The FDA’s FoodData Central database contains ingredient lists for over 739,000 branded food products — every ingredient, every additive, every cross-contact warning for the products we buy off grocery store shelves.
The problem was access. FoodData Central wasn’t designed for consumers to query by allergen. It’s a research tool. Searching it for “products that don’t contain sesame, soy, or tree nuts” was technically possible but practically useless — the kind of task that takes ten minutes per product if you know what you’re doing.
We built fdsrch to solve that.
What fdsrch Does
fdsrch is a food allergen screening tool. You type in what you want — a food name, a product, an ingredient you’re avoiding — and it returns results in under a second, filtered through the ingredient lists of 739,000+ FDA-verified products.
More importantly, it supports exclusion filtering. You tell it what you need to avoid, and it shows you products that don’t contain those ingredients. Not just products that don’t list them as the main ingredient — products that don’t contain them in any form, including derivatives, processing aids, and shared equipment warnings.
For someone managing multiple food allergies, this changes the grocery shopping calculus entirely.
Why 739,000 Products Matters
The FDA’s Branded Foods dataset isn’t just large — it’s specific. Unlike general food databases that aggregate nutritional data, it contains the actual ingredient lists as they appear on product labels, updated as manufacturers change formulations.
When a food company reformulates a product and the updated label gets submitted to the FDA, that change appears in fdsrch. When an allergen that wasn’t in a product gets added — because of a supplier change, a production line update, or a new certification requirement — the database reflects it.
This isn’t perfect. There’s always lag between a label change and a database update. fdsrch is a screening tool, not a safety guarantee. But it’s the most comprehensive ingredient-level database available to consumers, and it’s free to use.
The Search Architecture
The core challenge in building fdsrch was speed. The FDA database is enormous. Doing a full-text search across 739,000 ingredient lists, filtering for the presence or absence of multiple ingredients, and returning ranked results — that needs to happen in under a second to be useful.
We solved this with a custom search index built on PostgreSQL’s full-text search capabilities, combined with pre-processed allergen flags for the most common exclusion criteria. The result: most queries return in 200-400ms. Complex multi-allergen exclusion queries sometimes take 600-800ms. On a product database this large, that’s fast.
What’s Next for fdsrch
fdsrch launched as a basic search-and-filter tool. The roadmap we’re building toward:
- Allergen profiles: save your exclusion list once, have it applied to every search automatically
- Fit score: a single number telling you how well a product matches your dietary needs
- Barcode scanning: point your phone at a product in the store and get an instant allergen check
- Public API: for developers building nutrition and allergy apps who need reliable ingredient data
The database is the hard part. Everything else is an interface problem — and those are solvable.