The dataset
Every number on every page here comes out of five small tables. This page says what is in them, what each row rests on, and what is still empty — including the parts that stop us publishing a reviewed release.
What it contains
| Table | Rows | What it holds |
|---|---|---|
countries | 243 | Nominal supply voltage, mains frequency and the socket types in current use, per country and territory. |
plug_types | 15 | The IEC letter types, pin count, and whether the type is earthed. |
compatibility | 225 | Which plug physically enters which socket — 28 pairs mate — and, for each, whether the earth connection survives. 13 pairs fit and drop the earth. |
device_classes | 28 | Voltage tolerance, load type and wattage band per class of device, which is what turns a lookup into a verdict. |
regions | 0 | Child rows for countries whose supply varies geographically. Empty — see below. |
What a country row rests on
Country rows are graded by comparing two general references rather than merging them. The grade is derived from that comparison, never typed by hand, and where the two disagree the page publishes the disagreement instead of picking a winner.
The references are Wikipedia and worldstandards.eu.
| Grade | Countries | What it means |
|---|---|---|
verified | 1 | checked against a source |
concordant | 149 | two general references agree |
partial | 59 | references list slightly different sockets |
single_source | 15 | one general reference |
conflict | 16 | references disagree — unsettled |
unknown | 3 | not established |
Only verified means a person read the sources and signed the row off. Everything else is a statement about two references, not about the world.
What is not in it
Listed because an empty column is a finding, and because these are the columns the site would most like to have:
socket_types_legacy— what is still in the walls. Filled on 1 of 243 rows.variation_type— whether a country is uniform, regionally split, mixed or mid-transition. Settled on 1 of 243 rows, which is why no country page carries variation prose: 243 identical sentences saying we have not checked would be worse than silence.regions— 0 rows. Brazil’s 127/220 V split is not machine-readable here; it exists only as a disagreement between the two references.source_urlandsource_quote— the first-party citation behind a row. Filled on 1 of 243.sources— 0 rows. This is the table that would carry a quote, a page reference and a retrieval date per fact.
These are gaps in a research pass, not in a template. Filling them to make the pages look better informed is the one thing this project is built not to do.
Using it
The tables ship as one file, data.json, which is what the tool on this site runs against — so it cannot fall behind the pages. Roughly 37 KB of JSON.
Licensed CC BY 4.0. Attribution: plugverdict.com
, with a link. The nominal supply values are reference-derived and are credited above; the compatibility and device-class tables are ours.
Why there is no CSV release yet
A separate, stricter gate governs publishing the tables as a dataset in their own right, and it does not pass. Country rows have to be signed off by a person; the two consequential tables — where a wrong answer destroys a device rather than embarrassing us — need two sources on different domains per column and a signature.
| Gate | Passing |
|---|---|
| Country rows signed off | 1 of 243 |
compatibility rows signed off | 0 of 225 |
device_classes rows signed off | 0 of 28 |
| Sources recorded | 0 |
The distance is reported on every build rather than suppressed. Publishing the file the site already serves is one thing; issuing it as a reviewed release before anybody has reviewed it would be another.
Corrections
Live somewhere we have got wrong, or wire buildings for a living? That is the evidence this dataset is thinnest on. info@plugverdict.com.