How this site is built
Where every number on this site comes from
Every storage time here is the United States Department of Agriculture's, reproduced without change from one published snapshot of the FoodKeeper data. Nothing is estimated, rounded to taste or filled in where the source is silent. What this site adds is arithmetic on those figures and a rule for reading them, and both are described below so you can check them against the source yourself.
The source, exactly
The data is published by the USDA and the U.S. Department of Health and Human Services as the FoodKeeper application, at foodsafety.gov/keep-food-safe/foodkeeper-app. The file this site is built from is that application’s own data feed on the same domain, foodkeeper_data_url_en.json, retrieved on 1 September 2026. That copy is 676,401 bytes and carries 661 source records; its SHA-256 digest begins b0abf1e20ba5. The build never fetches anything: the snapshot is committed, the digest is checked before every release, and a source that quietly changed under us would fail the check rather than rewrite the site.
From source records to pages
The FoodKeeper file holds 661 records. Three pairs of them carry the same name and differ elsewhere — in the heading the source files them under, or in the cells it fills — so none of the 661 is an exact copy of another. Records that share a name are one food with several states — five kinds of milk, three of leftovers — so they are merged into a single product, and every state keeps its own windows and its own subheading. That leaves 466 products. A product earns a page when the source gives it a storage time outside the freezer and at least 3 neighbors in its category to compare it against, and when its own prose is not too close to a page already published: 133 products were held back by that last rule alone. 302 pages are published.
What is computed here, and how
Six quantities on this site are ours, not the source’s. The freezer gain divides the freezer figure by the longest window the same food has without a freezer, in the same package state. The opening cost divides the sealed window by the window after opening, in the same place. Neither is ever divided by a figure shorter than a day, because below a day the source is printing a limit on standing in a warm room and not a storage time. Rank within a category and position across the corpus are counted over foods the source rates for the same place, with ties sharing a place. The spread inside a range and the spread between methods are counted separately: the first says how sure the source is, the second says how much the choice of method matters. The date on a food page is the range added to the day you type, computed in your browser.
Where the source answers in words
Not every cell in the file is a number. These strings are read as answers rather than dropped, because a page that says the USDA is silent where it in fact answered is wrong about the one thing it is for. 3 of these strings are printed as they read; “1 Yea” is a typo in the source and is printed as “Unreadable in the source” with the original beside it, because reading it as one year would be our number rather than theirs.
- Package use-by date36 cells
- Indefinitely13 cells
- When Ripe12 cells
- 1 Yea2 cells
Which windows are about safety
The source prints storage times and does not say which of them are safety limits. This site decides that with one rule, from what the source itself printed: a window shorter than a day is a limit on standing in a warm room and is treated as safety; a freezer window is quality, because food held at 0 °F stays safe while it is frozen; a pantry window is quality, because rating room temperature at all says the food is stable there; a refrigerator window is safety for the categories the source files as perishable and quality otherwise. The page that prints a window says which of the two it is. The two-hour rule for food left out is USDA and FSIS guidance rather than a line in this file, and it is labeled as theirs wherever it appears.
What this site cannot tell you
It rests on one source. The figures were not re-derived from FSIS publications, and where FoodKeeper is wrong this site is wrong with it — tell us and we will say so here. It knows nothing about your refrigerator, your power cuts or how the food was handled before you got it. It has no laboratory, no test kitchen and no food scientist on staff; the arithmetic is ours and the expertise is the source’s. Where the source is silent, this site prints that it is silent rather than guessing.
Citing this
KeepsUntil, BiLingoPlus LLC. Built from USDA FoodKeeper data retrieved 1 September 2026, digest b0abf1e20ba5. Page addresses are stable: a food keeps its address for as long as the source keeps the food.
KeepsUntil is an independent reference built on public USDA data. It is not affiliated with, endorsed by, or connected to the U.S. Department of Agriculture or the Food and Drug Administration. Whether a given window is a safety limit or a quality window is stated on the page that prints it.