At some point building this I noticed that the board with the most votes on it was also the board I understood the least. Not because the numbers were wrong. Because numbers answer a different question than the one I needed answered.
Votes and conversations are not the same instrument
A vote is breadth. It tells you demand exists, that some number of people wanted a thing badly enough to click. That is genuinely useful. But it is all the vote knows how to say.
A conversation is depth. It tells you what the thing is for, what the person was doing right before they hit the wall, what they will do with the solution once they have it. These are not two sizes of the same signal. They are different instruments that happen to share a topic.
The mistake I kept making early on: treating the vote as a cheaper, faster, compressed version of the conversation. It is not a compressed version. It is a different kind of evidence entirely, and the number it produces is silent about almost everything that matters for deciding what to actually build.
What you hear and what you miss
Ask someone what they want and they answer as the person they intend to become. The disciplined version of them who uses advanced reporting, sets up automation, keeps the inbox at zero. That version is real enough that they mean what they say. But it is not the version that was cursing at the screen on a Tuesday night.
Ask instead about the last time they actually hit the problem: what were they trying to do, what broke, what workaround did they end up using? You get something truer. The real deadline. The spreadsheet they were copying data into. The reason they went looking in the first place.
A vote for "dark mode" might really be "I'm doing expense reports at eleven at night and the screen is burning a hole in my eyes." That sentence points somewhere. The checkbox labeled dark mode doesn't carry it.
The vote recorded that a person wanted something and dropped the reason. The reason is the only part that tells you what to actually build.
I build vote counting software and I still tell you this
Here is the honest part. Under every request on the board there is a comment thread. The score that ranks the board weights a comment at roughly two thirds of a vote (the formula: votes times three, plus comments times two, plus views times a tenth). I tuned it that way on purpose. A written sentence is the closest thing to a conversation that still scales. The product is quietly trying to pull your attention away from the column of numbers and toward the column of words.
But I am not going to oversell my own comment thread. A comment is a postcard from the conversation you should be having. Better than a tally. Still not the ten minutes. The person who left you a rich, specific comment is almost always the person who would have told you something you couldn't have guessed if you had just kept them talking. And the people who never commented at all are a silence worth understanding on its own.
Count to decide where to point. Talk to decide what to build.
The answer is not to stop counting. Votes are good for one specific thing: figuring out which problem is worth ten minutes of a real person's time. Use them for that. Look at the board, find the cluster, pick up the phone.
Then put the count down.
The mistake is letting the easy evidence stand in for the hard evidence, because the easy evidence fits in a column and the hard evidence requires a calendar invite. A hundred votes can move a feature up a list. One conversation can tell you the list was sorted by the wrong question.
Keep counting. Just don't mistake the count for the knowing.