Walmart Radio play log ====================== Every song observed on the public in-store Walmart Radio stream, logged by polling its metadata endpoint once a minute and recording each track change. Coverage 2024-05-28 to 2026-08-31 (825 days) Rows 234,289 plays Songs 2,012 titles by 1,041 artists Built 2026-08-31T22:31:46+00:00 Files ----- plays.csv one row per play, UTF-8, one header row plays.parquet the same rows, zstd-compressed, typed README.txt this file Columns ------- played_at_utc Time the track change was observed. UTC, second precision. CSV: ISO 8601 with an explicit Z (2026-08-13T12:34:56Z). Parquet: a UTC-aware timestamp. artist Artist string exactly as the stream reported it. song Title as reported, including any "(Feat. ...)" or remix credit. Song identity is the (artist, song) pair. Nothing is normalised or resolved against MusicBrainz or any other authority, so the same recording can appear under more than one string if the station's own metadata varies. Punctuation, capitalisation and featured-artist credits are reproduced, not cleaned. The timestamp is when the change was *observed*, within about a minute of when it happened. It is not the track's start time as the station would report it, and there is no duration column. The gap to the next row is the closest thing, and it is wrong wherever a gap intervenes. What is missing, and why it matters ----------------------------------- Play counts are lower bounds. 84% of the clock hours in the span contain something the logger recorded, whether a song or one of the live-show and advert markers described below. Songs alone account for 76%, and the markers cover a further 1,721 hours that would otherwise look unobserved. The remainder splits into two kinds of gap, and they are not the same thing: Logger downtime 27 outages, 1,284 hours in total, longest 877 hours. Nothing at all was recorded. This is us failing, and it is listed in full below. Station quiet 3,704 shorter blocks with no music, about 4,414 hours. The stream was up. It just was not reporting a song. Until 2026-04-18 the station ran three live talk shows on fixed daily slots, about 6.4 hours out of every 24, always the same Central-time hours, matching its own published schedule. The log carries no music inside them. Since 2026-04-18 it has run music through all 24 hours. If you compare two periods either side of that date, you are comparing two different stations. Do not bundle the two into one "downtime" figure. A note on the live shows. For the first two years the station ran three live talk shows on a fixed Central-time grid, and the early metadata source named them: 2,947 show blocks and 1,447 adverts are in the dataset because of it. The source changed on 2025-07-24 to one that reports only artist and title, so from that date the shows leave no trace at all, so their hours are genuinely unobserved rather than empty. Any comparison spanning that date is comparing two different loggers. Logger outages -------------- Whole days with no plays at all. Partial-day outages exist too and are not listed here. If precision matters, read gaps of six hours or more off the play sequence itself. 2024-07-06 to 2024-07-08 2024-08-06 to 2024-08-06 2024-09-10 to 2024-09-10 2024-09-12 to 2024-09-12 2024-09-25 to 2024-09-25 2024-10-07 to 2024-10-09 2025-06-18 to 2025-07-23 2026-03-24 to 2026-03-27 2026-06-27 to 2026-06-28 Provenance and reuse -------------------- An independent, non-commercial record of a public broadcast. Not affiliated with, sponsored by, or endorsed by Walmart. The log is factual observation, so use it for whatever you like. A link back is appreciated. Site https://scubdon.github.io/wmradio/ Source https://github.com/scubdon/wmradio Corrections https://github.com/scubdon/wmradio/issues