Open data
The whole catalogue, downloadable in full: every grade, every owner check-in count, every source count. No login, no rate limit, no email gate — a dataset behind a form is a lead magnet, not a dataset.
Download the full CSV ↓ 81 product-eras · 104 source attributions
Field dictionary
Written out rather than generated from the header row: a column name is not a
definition, and score in particular needs saying out loud.
slug
stable record id; the address of the record is /product/<slug>
brand
brand name at the time of this era, not the current owner
name
product name
era
the production period this row grades — the unit of the catalogue
category
aisle
origin
where this era was made, as stated on the product or its terms
price_usd
typical street price when the row was graded; for vintage rows, a used-market estimate
expected_years
median expected service life
cost_per_year
price ÷ expected years — the headline number
score
0–100 composite, recomputed from this row's own evidence on every read; rubric on the record page
score_confidence
high | medium | low, driven by how many distinct sources back the row
warranty_grade
A–F on terms × how they are honoured; blank for secondhand
health_grade
A–F; contact route × duration × substances
env_make
footprint to manufacture
env_use
footprint in use
status
rising | steady | falling | vintage
alive_reports
owner check-ins reporting the item still in service
fail_reports
owner check-ins reporting a failure
distinct_sources
how many independent documents back this row
published
false means the grades are seeded and unverified — shown with the caveat, not hidden
What the score is, and is not
score is recomputed from each row's own evidence every time it is
read — it is never stored and served back. That means a grade change moves the
number everywhere at once, and it also means two downloads taken either side of
a correction will legitimately differ. If you are replicating something, take
the date with the file.
published = false means the grades were seeded from research notes
and nobody has re-read the terms since. Those rows ship in the file rather than
being hidden, because a dataset that quietly omits its weakest rows is telling
you something untrue about its strongest ones.
Licence
CC BY 4.0 — attribute "Buy It Once contributors". Build failure-rate charts, price trackers, replications; that is the point. If your work changes a grade, say so and it gets logged with credit on the corrections page.