bandhand.io
Private listening pages: what the numbers actually tell you
We sent an EP to a dozen people on private listening pages and measured everything. Three findings that changed how we read that data, and one false signal that looks exactly like a click.
We built this for our own band first. Instead of sending industry people a link to a cloud folder or an attachment, everyone got their own page: their name at the top, four tracks, a player, nothing else. Every page records what happened on it. The open, the play, progress in quarters, and whether anyone reached the end.
Two weeks later we had a full set of data from a dozen of those pages. Half of what it told us was not what we expected.
Most people open and never press play
This is the first fact to make peace with before you read anything else into your stats. Across dozens of visits to the listening pages there was a handful of plays, and the number of tracks finished end to end fits on one hand.
That is not a failed campaign. That is the normal ratio, and it is better to know it than to be surprised by it. Someone gets a link, clicks because they are curious, sees four songs at three and a half minutes each, works out that this is a quarter of an hour of their life, and saves it for later. Later sometimes arrives and sometimes does not.
The practical conclusion is that an open on its own means nothing and there is no point building hope on it. If someone opened and did not play, that is not a signal saying no. It is a signal saying not right now. A second message to that person makes sense and is not pushy. A second message to someone who finished three tracks makes a completely different kind of sense, because you know the material held them.
Track one always wins
This one was a surprise, or rather a lesson in humility. We looked at the numbers and saw that one song had clearly more plays than the other three. For a moment we read that as the song people like most.
Then somebody asked what order the tracks were in on the page. That order, exactly. The song was first on the list, on every page, for every recipient. People hit play at the top, because that is how people hit play.
When we threw away the play count and looked only at tracks finished to the end, the picture flipped. The only two people who finished anything finished the same, completely different song. Not the one with the most clicks.
Two things follow. First, with four tracks on one page the play count measures position, not preference, and has to be read with that correction applied. Second, if you genuinely want to know which song holds up best, either set the order deliberately and stay aware of it while reading the results, or just ask. One sentence on the page saying tell me which one you would put out as a single gets you further than a week of staring at charts.
The fake click that looks real
This one is the sneakiest and deserves its own section.
You send someone a link in a messaging app. A few seconds later an open shows up in your stats. It looks perfect: you sent it, they clicked immediately, they are interested.
Except it was not them. The platform fetched the page itself to build the link preview, that small card with a title and an image. The bot hits the address instantly after sending and leaves exactly the same trace in your log as a human.
You can spot it by three things. It arrives seconds after the message, not fifteen minutes later. It comes from an address range belonging to that platform, not from a home or mobile provider. And it never starts the player, because bots do not listen to music.
We counted those visits backwards and found several sitting in our stats, including one from a few days earlier that had cheered us up for no reason whatsoever. We filtered them out, and the report now shows them in a separate section labelled plainly: this is not a person.
If you measure your own links, do the same. Otherwise every message you send hands you one free open, and you slowly start believing that people respond better than they do.
An address with a name in it is not a private address
The last thing is not about numbers but about what you expose along the way.
Every listening page has an address containing the recipient's name or handle. Ours carry a meta tag telling search engines not to index the page, which is obvious. What is less obvious is that the sitemap, the file your platform generates automatically and serves publicly, will happily list all of those addresses at once.
That is exactly what happened to us. Two freshly created pages landed in the sitemap because the system adds every new page there by default. For half an hour, anyone who opened that file could see a list of the people we were writing to.
Blocking indexing protects you from Google, not from a curious human. If you build pages like this yourself, check the sitemap as well as the meta tag, and check it after creating the page rather than instead.
What we took away
Three sentences we repeat to ourselves before every send.
An open is curiosity, not interest, and curiosity usually stops there. A track finished to the end is the only number in this set that really means something, which is why it is worth measuring progress rather than the play button. And if a number looks too good, check whether you are looking at a robot before you let yourself enjoy it.
All posts