For a long time my process for deciding what to make was essentially whether I felt like making it. That produced a lot of work, some of which was good, most of which existed for no particular reason.
The question I ask now is simple and it has cut my output by about half and improved almost everything.
The question
Who is going to be glad this exists, and what were they doing before they found it?
Both halves matter. The first is about audience, which everyone asks. The second is about the situation, which almost nobody does.
Why the second half is the useful one
If I can describe what someone was doing in the ten minutes before they arrived, I know something concrete about what the thing needs to be.
Somebody who was stuck on a specific problem, searching for an answer, needs the answer near the top and needs it precise. They are not going to enjoy a leisurely introduction.
Somebody who was browsing with nothing particular in mind needs a reason to stay in the first fifteen seconds and will tolerate, even prefer, a slower unfolding.
Somebody who follows the channel and clicked because it was mine has already decided to stay and needs neither.
These are three completely different pieces of work, and I was making all of them the same way.
What it eliminated
The ideas that fail this question fail in a recognisable pattern.
Topics I found interesting where I could not identify anybody with a use for the result. These are the hardest to give up because the interest is genuine, and interest is not the same as usefulness.
Things everyone else had already covered adequately, where the honest answer to who would be glad is nobody, since they would have found one of the twenty existing versions first.
Things I was making because it had been a while since I posted, which is not a reason and which produced the worst work I have made.
And overviews of large subjects, which sound useful and where the person arriving is always at some specific point in a subject rather than at the beginning of it.
What it produced instead
A bias toward narrower topics, which I resisted initially because the audience seemed too small.
That resistance was wrong and I can see why now. A narrow piece that exactly answers something has a small audience that finds it reliably and values it highly. A broad piece has a large potential audience that mostly finds something better.
The narrow ones also last. A specific answer to a specific problem keeps getting found for years. A general overview is superseded within months by someone else's general overview.
The failure mode of this approach
Being fair about it. Applied too strictly, this produces relentlessly utilitarian work with no personality.
Not everything needs to solve a problem. Some of the work I value most, from other people, exists because someone found something interesting and made a case for it, and there was no identifiable audience until it arrived.
What I do is allow a proportion. Roughly one in five of the things I make now do not pass the question and I make them anyway, because they are the ones that define what the channel is rather than what it does.
They perform worse on average and a few have done far better than anything else, which is what you would expect from higher variance.
Asking it before the research, not after
The timing turns out to matter.
Once I have spent six hours researching something, I am committed and the question gets answered generously. There is always somebody who would be glad, if you want there to be.
Asking before means answering it while I do not care about the outcome, which is when the answer is honest.
The written form helps too. If I cannot write the answer in two sentences, I do not have one, and vagueness at this stage becomes vagueness in the finished thing.
The version for anything, not just video
I have started applying it more widely and it holds up.
Before writing something, before starting a project, before agreeing to a piece of work. Who is going to be glad this exists, and what were they doing before they found it.
The number of things that do not survive it is uncomfortable, and it is more comfortable than finishing them and finding out afterwards.
What changed measurably
Output down by roughly half. Average performance up considerably, mostly because the weak entries stopped being made rather than because the strong ones improved.
And a thing I did not expect: the work is easier now. When you know who it is for and what they were doing, most of the structural decisions answer themselves. The difficulty I used to experience in the middle of a project was usually the difficulty of not knowing what the thing was.