How we count a play
What Airtime actually measures, what it misses, and why we say so.
Every figure on this site comes from the same thing: a recording of what went out on air. Ten Zambian stations, around the clock, identified by sound. No returns from programmers, no self-reported logs, nobody's word for it.
Here is exactly how a number becomes a number.
A play is a minute on air
Short samples are taken from each station continuously and matched against a fingerprint database. A song counts as played when it has been heard for at least a minute, with its position in the track keeping pace with the clock throughout — so a record found at 1:10 in and again a minute later at 2:10 is running, and one found twice at 0:30 is not.
That gate exists because a few seconds of a record under a station ident or inside an advert is not an airing, and counting it would flatter whoever uses the most music beds.
A minute is not a round number chosen for neatness. It is the shortest airing this method can actually prove. One recognition tells you a record was present; two of them, a minute apart and advancing in step, tell you it was playing. Drop the gate below that and a single recognition would be enough, which an advert bed can produce. Raise it and we would start discarding real airings — edits, and records faded out under a presenter link. So: the shortest span we can evidence, and the longest we can demand without losing things that genuinely happened.
We sample, we do not listen
This is the part most worth understanding. Airtime does not listen to every second of every station. It samples twelve seconds at a time. When nothing is recognised it looks again three minutes later. When something is recognised it looks once more a minute later to confirm, and then skips ahead to roughly where that record should end rather than listening through it.
That means the site can tell you that a song was on air and when, but not how long it ran. A clean airing produces two confirmations a minute apart whether the record is three minutes or six. Anywhere we show a time, it is the moment the song started.
One airing, counted once
Two things that look like repetition are not, and we fold them together.
- A stream stuck on a loop is found over and over, always deep in the same track. A record cannot start again before it could have finished, and a run picked up well past the beginning did not start there. That is one airing, not forty.
- The same recording sometimes exists twice under different identifiers. Chef 187 appears under five separate artist profiles upstream; T-Sean under six; Yo Maps five. They are one artist here.
Two stations, or it does not chart
Counting a play right is not the same as ranking it right, and we got this wrong first time.
A record played once an hour by a single station reaches a national chart on that one station's rotation alone. We found exactly that, pulled all sixteen airings out of the raw recordings, and compared them against each other — identical, and identical again when we asked the fingerprint service a second time by hand. Sixteen real airings, one station, every hour at about eleven minutes past. A plug, not a hit.
A chart that cannot tell a clock wheel from a hit is measuring one programmer's decision and calling it a country.
So a record now has to be carried by at least two stations to appear on the national chart. Nothing is deleted: the plays are still counted, still shown on the song's own page and on that station's page, and if you filter the chart to a single station the rule lifts, because there the question means nothing. The record simply does not rank nationally on one station's say-so.
What we cannot see
Only commercially released, digitally distributed music can be identified by fingerprint. Unreleased tracks, freestyles, live sessions and a great deal of Zambian music that was never properly released are invisible to this method, however often they are played.
So when we say 8.8% of plays in the week to 27 October were by Zambian artists, the honest version is: at least 8.8%. The true figure is higher and we do not know by how much. We would rather print that sentence than a cleaner number we cannot stand behind.
8.8% of plays were by Zambian artists — and that is a floor, not a measurement 8,228 plays, ten stations, week to 27 October 2026
And one more limit
The archive is still being worked through, backwards from today. Coverage is not yet even across stations or across the clock, so totals move as it fills — the same week read 2,916 plays when we first wrote about it two days ago and reads 8,228 now, because the scan caught up, not because radio got busier. Shares are steadier than totals, which is why we lean on them.
Figures in our stories carry the date they were taken. If one looks wrong, tell us — that is how we find the errors.