How we date an appliance from its serial number — and where we still don’t know

In brief

Lifetime figures — this is the complete history, not a recent sample. Since: August 20, 2026 · As of: September 4, 2026

Of the 291 attested attempts: 121 resolved to a single year, 105 returned two or more candidate years (the ambiguity this page is about), and 65 could not be read at all (unsupported brand or malformed input). We only count attempts a real browser fired — a measured share of raw traffic hits our logging endpoint directly with no browser behind it, and including that traffic makes the decoder look worse than it is, in a one-directional way, so it’s excluded from the percentage above rather than used to flatter it.

Separately, our decoding logic is checked against 129 automated test cases covering known real-world serial formats (129/129 passing) — that’s a code-correctness check, not a count of real appliances; the figures above are the real ones.

Why the ambiguity exists

Every major manufacturer's date code is short by design — it has to fit alongside a plant code and a week number on a small rating-plate sticker. Whirlpool assigns a single letter to the year and reuses it roughly every 30 years. GE cycles every 12. Frigidaire, LG and Samsung encode only the final digit of the year, which cycles every decade. None of that is a defect in the appliance's own labeling — it's just not designed to be read decades later, out of context, by someone who doesn't already know roughly when they bought the thing.

The naive fix — assume the most recent cycle — is wrong often enough to matter, and wrong in a predictable direction: it always overstates how new an appliance is. We built this decoder specifically to avoid defaulting to that guess without saying so.

How we resolve it, when we can

For a specific model — not a brand in general, a specific model number — a manufacturer spec sheet, a dated press review, or a retailer's own archived listing history can pin down the window that model actually existed in. If that window rules out all but one of the candidate years the serial's cycle produces, we narrow to that year. If it only rules out some, we say so and still show the rest.

We do not treat a single visitor's self-reported correction as sufficient evidence on its own to change what the decoder tells the next person — a self-selected sample of one can easily be wrong (misremembering a purchase date, confusing it with an installation date). It's treated as a lead worth checking, not a fact worth shipping.

Resolved cases

Every row below started as a visitor-reported miss and was closed out with a dated, independently checkable source — not a guess, and not a visitor's memory alone.

Open cases — what we still don’t know

A page that only shows resolved cases isn’t a credible one. These are logged with the same rigor and are shown ambiguous on the live tool today, on purpose, until a real source turns up.