How to research content ideas for a niche

The usual content calendar is a brainstorm with dates on it. The team lists what it wants to say, someone checks a keyword tool, and the list ships. Nobody can say why an idea is on it, so nobody can say why it failed.

A researched backlog is different in one way: every idea carries the links to where your audience asked for it. That makes the list something you can argue with.

Look where the niche asks, not where it searches

Keyword tools show you what people type into a search box. That is useful, but it is the least specific version of the question. The same person writes a much better question when they ask other people.

Look in the places your niche talks to itself:

Communities. Subreddits, Slack and Discord groups that are public, forums, niche Q&A sites. Read the questions and, just as much, the replies that say “I have the same problem”.
YouTube. Video titles show what gets made. Comments show what the video did not answer. The comment “but what about when…” is an idea.
Autocomplete. Type the seed terms into search and a video platform and write down the completions. Cheap, fast, and in the audience’s words.
Answer engines. Ask the core question and read the follow-up questions they suggest. Those are the questions people ask next.
Your own inbox. Sales call notes and support tickets, if you have them, beat every public source.

Count evidence, do not estimate it

For each question, save the verbatim phrasing and the URL. Then cluster the questions that ask the same thing in different words.

The evidence score is a count: how many distinct places asked this. Distinct matters. The same person asking twice in one thread counts once. A question that appears in three communities, two comment sections, and the autocomplete is a better bet than one that appears once, however clever it sounds.

A worked example. Say your niche is bookkeepers who serve restaurants. A research pass finds “how do I handle tip pooling in payroll” phrased six different ways across two subreddits, a Facebook group post that is public, and the comments under a popular payroll video. That is one idea with strong evidence. “The future of restaurant accounting” appears nowhere. It is not an idea yet; it is a topic.

Score gap and fit

Evidence says people want it. Two more scores say whether you should write it.

Gap. Read the current top answers, in search and in answer engines. Are they thin, out of date, written for the wrong audience, or missing the part people keep asking about in the comments? Write down what they miss in one sentence. If you cannot name what they miss, the gap is small.

Fit. Does answering this honestly lead anywhere near what you sell? The tip pooling question fits a payroll product perfectly. “Best restaurant POS” might have huge evidence and zero fit for a bookkeeper. Keep it, but label it a traffic idea, not a pipeline idea.

Rank by evidence first, then gap, then fit. Resist the urge to rank by which ideas you find most interesting.

Freeze the backlog

Once the list is scored, freeze it with a date. Do not keep editing it as you go.

A backlog you edit every week becomes a list of what you felt like writing. A frozen one can be compared with the next research pass, and the difference tells you what changed in the niche: which questions got louder, which got answered, and which are new.

Put it on a schedule

The niche-idea-backlog template runs this whole guide as a loop and fails any idea without an evidence link. The niche-content-research skill does a single pass on request. If your niche lives on video, pair it with the YouTube planning guide.

How to find buying signals in public data. The demand themes from that scan are ready-made backlog ideas.
How to write a stopping condition, for when this becomes a loop.