Who picks the music? Two months of bar data
Published on
For two months I watched what happens in a bar in my home town when the guests get to pick the music. Alongside it, I looked at what happens when nobody is there who wants to pick anything. The two numbers sit so far apart that together they say more about Juqed than either does alone.
Where guests are present, 88 per cent of the songs played come from them. Where there are none, the Auto-DJ plays 98 per cent. Same software, same setup – the only difference is whether somebody reaches for their phone.
The bar: 392 songs, 363 of them requested
The venue runs Juqed Go with Juqed Gold, and the Auto-DJ was set to genres – rock, country, metal, plus the 80s, 90s and 2000s. That is the safety line: if nobody requests anything, it carries on in that direction so the music never stops.
It rarely had to. Over two months 392 songs played, and 363 of them came from guests. The Auto-DJ stepped in a couple of dozen times; the landlord intervened seven times. That works out at 92.6 per cent guest requests.
| Metric | Value |
|---|---|
| Period | 23 May – 23 July 2026 (two months) |
| Songs played | 392 |
| of those guest-requested | 363 (92.6%) |
| Votes cast | 527 |
| Guests taking part | 36 |
| Votes per guest | 14.6 |
| Songs per guest | 10.9 |
| Auto-DJ source | Genre-based (rock, country, metal, 80s–2000s) |
The figure that surprised me most is not the song share but 14.6 votes per guest. Someone who scans the QR code does not request one song and put the phone away. They stay with it and vote on what other people suggest. The music request turns into a game that carries the evening.
Now the counter-test
In the same period, two Juqed Web events ran without a single guest: one computer, one browser window, music from morning to night. No requests, no votes. The Auto-DJ handled 98 per cent of them on its own; the host started the rest.
That is not a failure, it is the second use case. Some venues do not want participation at all – they want music that fits, playing without anyone having to think about it. Both are the same app, configured differently.
Is the bar an outlier?
No – it sits above average, but not out of line. Taking together every event that had guests at all, 88.1 per cent of the songs played came from those guests. And votes per guest land almost exactly on the bar's figure. That looks less like a quirk of one venue and more like what people simply do once they have scanned the code.
| Metric | The bar | All events with guests | All events |
|---|---|---|---|
| Share of songs played that were guest-requested | 92.6% | 88.1% | about 44% |
| Votes per guest | 14.6 | 13.6 | – |
Why that last column is so much lower
Added up across every event, only about 44 per cent of the songs played came from guests – so the Auto-DJ plays more overall. At first that reads as a contradiction of everything above, but it is precisely the point: the unattended long-runners count in full, and they run around the clock, while a party night is over after six hours. A single web event across two days puts more songs on the clock than three weddings.
So the last column does not measure how popular having a say is. It measures how much music plays quietly along while nobody is watching.
What I took from it
Two things I had only assumed before.
The Auto-DJ is not a compromise, it is the base load. I built it as a gap-filler. In fact it is the entire product in half the cases – and in the other half it stays so far out of the way that it had to step in only a couple of dozen times in two months. Both of those are the same success.
Participation does not tail off. My worry was that guests would request something once and then drop out. 14.6 votes per person say the opposite. Voting asks less than requesting – you do not have to think of anything, only agree – and that is what keeps people in.
If you run a bar, a restaurant or a hotel yourself: the curated playlists and continuous operation are part of Juqed Gold, and the individual scenarios are set out under use cases.
How these numbers came about
Every figure comes from Juqed's own event log, as of 19 August 2026. A song counts as a "guest request" when it was submitted through a scanned QR code and actually played – requests that were submitted but never played are not included. These are operating data, not a study: they show what happened, not what would happen in a properly designed trial. The bar is a single venue, and 36 guests is a small number. I quote them anyway, because they held steady across two months and because the other events point the same way.