SeasonSort Week of Sep 4 · updated Fri

Where our prices come from: reading the USDA retail reports

Every price on this site comes from a public dataset that most people have never heard of, and understanding what it is — and more importantly what it is not — is the difference between using these numbers well and being misled by them.

The reports

The US Department of Agriculture’s Agricultural Marketing Service collects what American grocery retailers advertised each week and publishes it. Two of those reports drive this site:

They are published weekly, on a Friday, and they are genuinely national in scope: tens of thousands of store ads, aggregated. That breadth is the reason we use them rather than scraping a handful of retailer websites, which would be narrower, less consistent, and on much shakier legal ground.

The one thing to understand: advertised is not shelf

This is the single most important caveat on the site, so it is worth being blunt about it. These reports capture promoted prices. An item only appears in a given week if stores actually ran an ad for it.

Three consequences follow, and they all matter:

  1. The prices skew low. An advertised price is by definition a price someone chose to shout about. The everyday shelf price for the same item is usually higher, sometimes substantially.
  2. Items disappear. If nobody advertised spinach in your region this week, spinach has no price this week. That is not an error and it does not mean the shops have none. It is simply absent from the data, and we show it as unavailable rather than carrying last week’s number forward and pretending it is current.
  3. A price change can be a mix change. If the stores that advertised apples this week are different from the ones that advertised them last week, the average moves even if no individual price did. Large week-over-week swings on a thinly-advertised item are often this rather than a real market move.

So treat every figure here as a guide to relative cost and to timing — is chicken cheap right now, is this the week for peaches, which region has the best price — rather than as a quote for your particular shop.

How a report becomes a page

The reports are organised around commodities and pack sizes, not around the way people shop. Turning one into a usable page takes a few steps.

We keep a hand-maintained catalog of the items we track — currently 48 produce items and 45 meat, poultry, and egg items. Each catalog entry records which commodity in the USDA report it corresponds to, which package sizes count as a match, a reference nutrition lookup, and a typical serving weight.

That last detail is less trivial than it sounds. The report might price corn by the ear, in trays of four to five ears, and by the pound, and a naive substring match on “ear” will happily match the tray. Getting a clean per-unit price means the catalog has to be specific about which sizes it will accept, item by item.

Nutrition comes separately, from the USDA’s FoodData Central database, cached per hundred grams and scaled to the serving weight in our catalog. There is a similar matching problem there: a plain search for a food will cheerfully return a restaurant menu item or the wrong species, so each catalog entry carries a specific query and we pick the result whose description overlaps it most.

This week · Midwest, week of September 4, 2026

43 of our 48 tracked produce items had an advertised price in the Midwest this week. The rest were not advertised anywhere in the region, which is normal and varies week to week. You can see the full list on this week’s produce.

Regions

The reports break the country into regions, and we carry six of them: Midwest, Northeast, Southeast, South Central, Northwest, and Southwest. Prices genuinely differ between them — often by twenty or thirty percent for the same item in the same week — which is why every item page shows a regional comparison rather than a single national figure.

Seasonality is regional too. A blueberry that is at its peak in the Southeast in May is still weeks away in the Midwest, so our season windows shift for the warmer regions rather than applying one national calendar to everyone.

What we do not have

Being explicit about the gaps is more useful than pretending they are not there:

When it updates

The USDA publishes on Friday. An automated job pulls the new reports on Saturday morning, rebuilds every price, cost, and seasonality figure on the site, and publishes. If the fetch fails, the job fails loudly rather than quietly serving last week’s data as though it were current — the date at the top of every page is the week the underlying report covers, so you can always see how fresh the number in front of you is.

For how these prices turn into a cost per serving for a whole recipe, including where that calculation is least reliable, see how we calculate cost per serving.

Questions or a correction? Get in touch. More on our data sources in About.