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

In brief

Cumulative figures for the stated period, not a recent sample. Since: August 20, 2026 · As of: October 16, 2026

The 82% is 1838 of 2249. Of the 2541 attested attempts, 292 weren’t a test of the decoder and are set aside: a model or part number pasted into the serial box, a mistyped entry the visitor re-fixed, no brand selected, or a string that doesn’t fit the chosen brand’s serial format and reads as a wrong-brand or non-serial entry. Of the 2249 that remain: 840 resolved to a single year, 986 to two or more candidate years (the ambiguity this page is about), 12 failed when first logged but resolve on the decoder as it stands today, and 411 did not resolve because we don’t support that brand or serial format yet. That last group is counted against us, not excluded.

We remove repeated input combinations and marked QA probes, then match attempts to hours in which GA4 recorded a decode event. That is a traffic-quality proxy, not proof that each record came from a different person or a real browser; consent choices and ad blockers can leave legitimate attempts unmatched. The 2949 logged records contain 2660 distinct input combinations before this hour matching. We re-run historical failures through the current decoder. A recovered candidate year does not prove the year is correct.

This October refresh corrects the audit classifier to use the water-heater and HVAC registries and the recorded Kenmore builder. Some supported-brand failures were previously classified as unselected brands. The current percentage is not directly comparable with the earlier figure: both the corpus and the reporting method changed. Since then, we added a limited AEG washer/dryer format from AEG’s UK FAQ; its one-digit year remains decade-ambiguous, and the code rejects the separate nine-digit PNC rather than treating it as a serial.

Separately, our decoding logic is checked against 220 automated test cases covering known real-world serial formats (220/220 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.

Common questions

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.