The Poll That Lied to You
You ran a poll on your channel with one question: what kind of content do you want?
The answer came back clear: seventy-eight percent chose "deep, detailed content".
You were delighted, because that's what you wanted to make anyway.
So you spent three weeks on a forty-minute video, carefully built, with research and sources and detail.
And you published it.
Half the viewers left at minute three.
The following week you published a short video you filmed in an hour, about something simple nobody had asked for.
People finished it, shared it, and asked for more of the same.
Nobody lied to you in that poll. People describe what they wish they wanted, and watch what they actually want.
Why Guessing Is Expensive
Every video you make costs you hours. Guessing means spending those hours on a subject that might interest nobody.
And worse, a wrong guess repeats: you assume the problem is execution so you improve it, while the problem is the subject itself.
The fix is simpler than it looks: people tell you what they want if you ask, and their behaviour tells you even when they don't speak.
Way One: Ask Directly
The simplest method and the most neglected.
At the end of a video: "what subject do you want me to cover next?" One question in the last ten seconds.
In a written post: ask something specific rather than general. "What's the hardest part of filming for you?" beats "what do you want?"
A specific question gets specific answers. A general question gets "anything" and "whatever you think".
Way Two: Polls
Most platforms offer a poll with options. They're useful because they cost the participant no effort, so more people take part.
Use them to choose between defined alternatives: "which video first: A, B or C?"
Don't use them for open questions, as in the story we opened with. The options you write constrain the answer to what you thought of, and what they want may be outside your list.
Way Three: Recurring Questions
This is the strongest signal of all, because a repeated question is a proven request rather than an opinion.
Log every question that recurs in your comments or messages. If three people asked it, there are thirty who didn't.
And how to capture those questions while replying is covered in: how to reply to comments
Way Four: Read Behaviour, Not Words
And this is the most important point in the article.
What people say they want isn't always what they watch. They may ask for "deep, detailed content" and then leave it at minute three, while finishing a simple video nobody requested.
That isn't dishonesty on their part.
So look at: which of your videos people finished. Which one someone came back to a second time. And which one people shared without being asked.
These are honest answers that need no question. And how to read them is covered in: after publishing: how to read a video's performance and feed it back into your ideas.
And the practical rule: when what they said contradicts what they did, believe what they did.
Way Five: Talk to One Person
Ten minutes with one subscriber gives you what a hundred-vote poll can't.
Pick someone who comments regularly and ask: what made you follow? Which video didn't work for you and why? What are you looking for and not finding?
Open answers reveal things that never occurred to you, while a poll confirms or denies only what occurred to you.
And What Not to Do
Don't act on every request. Some requests come from one person and concern only them. Act on what repeats, not on what's asked once.
Don't let your audience steer your channel entirely. You know your subject better than they do, and they may ask for something that pulls you away from your identity. Listen, take what's useful, then decide yourself.
And don't ask before you have an audience. Early on, experiment first and ask later. A question in a video three people watch gives you nothing.
And Collect the Answers in One Place
Scattered requests across comments, messages and polls get lost if they aren't gathered.
Log every recurring request in your idea backlog the moment you see it, along with how many people asked.
A month later you'll have a list ordered by actual demand rather than your guesswork. And that's the best content plan you can get.