Researching your niche without a budget

Research tools can help you find demand and organise a comparison. You still need to watch the existing answers closely enough to say what you could add.

Start with an afternoon of watching. Tools can help you search and summarise; use your own knowledge of the trade to check what they miss.

Why do this by hand?

Because you are looking for the quality of the answer, as well as the size of the audience.

High search volume can signal demand; it does not tell you whether the answers are good. Watch for explanations that are out of date, aimed at the wrong level or missing the part you know people struggle with.

Doing it by hand has a second effect that survives the afternoon. You end up with an accurate picture of what your part of the internet sounds like, which changes how you write titles for years afterwards.

Who goes on the list?

Ten. Not the ten biggest.

Take three or four who are clearly ahead of you and doing something close to what you do. Take three or four at roughly your size, because those tell you what is working right now. Take two from an adjacent trade that shares your audience without competing for it, since that is where most format ideas come from.

Ignore the enormous general channels. They are playing a different game with different economics and copying their formats is how a specialist channel makes itself generic.

The columns

One row per channel. Five columns and a notes field.

What they cover

The recurring subjects, in their words rather than yours. Copy the phrasing from titles. You are collecting vocabulary as much as topics, and the vocabulary matters, because it is what people type.

What they never cover

Read down a year of titles and note what is absent: money, failure, maintenance, what to do when it goes wrong. Each gap is a question to investigate, not proof of hidden demand.

What the comments ask for

Go to the three most watched videos on each channel and read the top comments, plus a few of the recent ones. You are looking for questions asked repeatedly, corrections, and any variation of but what about. Each one of those is a video request that has been sitting in public for a year.

Which videos beat their own average

Not which have the most views overall. Which outperformed that channel's own normal range. You can see this by eye: scroll the uploads and the outliers are obvious against their neighbours.

What the outliers have in common

The one that produces the finding. Look across every outlier you have collected, from all ten, and ask what shape they share. It is often not the subject. It is a format, a title construction, a promise, a level of specificity, or the fact that all of them answer a question.

What are you looking for?

A subject with obvious demand and a weak set of existing answers.

The demand shows up as repetition: several channels covering it, comments asking about it, outliers touching it. The weakness shows up when you watch the best available answer and can immediately name three things wrong with it. Too long. Assumes equipment nobody has. Never says what to do when the thing fails. Made by somebody who has clearly never done it more than once.

An untouched subject may be a gap, a new need or something with little demand. Check it against customer conversations, search patterns and a small test before building a series around it.

What do you do with the sheet?

Three outputs, then close it.

A list of subjects where you can beat the current best answer, ranked by how confident you are. A list of format observations, which is what the outliers taught you. A list of things you will not make, which comes from the absences that turned out to be absent for good reason.

This studio runs the same process on its own business, which is why it takes an afternoon rather than a fortnight: the method does not change between trades, and the judgement is the only expensive part.

Do it once a year, or whenever the channel starts feeling directionless. In between, the comments on your own videos do most of the same job for free, and they are about you rather than about somebody who happens to be nearby.

Sources and evidence

What is established. Research on sampling generally finds that self-selection can make a sample unrepresentative of the wider population it is drawn from, because the people who choose to take part, or in this case choose to comment, differ systematically from those who do not. A study looking specifically at social media users found that who participates and who posts on a platform is itself self-selective, evidence drawn from one platform but consistent with the wider methodological point. Applied here, that means comments and titles collected by hand describe what a vocal, engaged slice of an audience says and does, not the whole audience, which is why the method in this piece treats what it finds as a source of judgement to be corroborated across several channels rather than as a statistic to be trusted on its own.

Our method. An afternoon spent watching ten channels, chosen for size and relevance rather than reach, reading comments for repeated questions and scanning upload history for outliers, produces a shortlist of subjects worth making and a shortlist of subjects to leave alone. This studio runs the same process on its own channel for the same reason: the judgement is the expensive part, not the tool.

References.

• Smith et al. Understanding sources of bias in research. pmc.ncbi.nlm.nih.gov. Self-selection can make samples non-representative. General research-methods source.
• Vicente et al. Sampling Twitter users for social science research. link.springer.com. Social-media participation and posting are self-selective. Twitter/X-specific evidence with wider methodological relevance.