A keyword tool will give you fifty topics in an afternoon. It will not tell you which of them you have any business writing, and that is the part deciding whether a post is worth the day it costs. Most content plans fail here rather than at the writing.
There is a cheaper source of topics, and you have been generating it for as long as you have had customers. The problem is that nobody writes it down.
- The best topics are questions somebody already asked you, in their words rather than yours.
- A topic is ready when you can answer it from having done the thing, not from having read about it.
- Cadence is a capacity question. Set it by your worst month.
- Keyword research is a good editor and a poor author. Use it second.
- Judging a post by its first fortnight will make you rewrite things that were fine.
Why the keyword list runs out
Starting from search volume produces a list of topics ranked by how many people want the answer, which sounds like exactly the right ranking. It leaves out the only two questions that matter: whether you can answer it better than what is already there, and whether you will still care by paragraph nine.
The result is recognisable. Four posts written with enthusiasm, then a gap, then a post that reads like it was assembled rather than written, then nothing. The plan did not fail because the topics were wrong. It failed because there was no reason behind any single one of them beyond a number in a column.
Search tools also describe demand that already exists. The questions worth answering are often ones people cannot phrase yet, which is why they are searching for something adjacent and not finding it. Those never look good in a volume estimate.
The questions you have already answered
Open a document and go back through the last three months of anything where somebody asked you something. Support email. The confused part of a client call. The message from a colleague that started with a quick question. The forum thread you replied to at length because the accepted answer was wrong.
Copy the question in the words it was asked in. Not your tidied version, theirs. Beside it, write what you actually said back.
That second column is the useful one. If your answer was two lines, the topic is thin and probably belongs inside another post. If you wrote six paragraphs, caveated two of them and then sent a follow up because you forgot something, you have a post and most of its structure. The shape of your reply is the shape of the article.
| Where the question came from | What it tells you | What it becomes |
|---|---|---|
| Asked more than once | It is not one person being confused | A post, and probably a documentation page too |
| You answered at length | The real answer is longer than the obvious one | The main article, with your reply as the outline |
| You had to correct an assumption first | People arrive holding the wrong model | The opening section, which is usually the whole reason it works |
| You searched for it yourself and found nothing | Demand exists and the answer does not | The piece with the least competition you will ever write |
Keep the phrasing people use. When somebody writes that their exported file lost the headings, that is the sentence to work from. Translating it into the industry term for the same thing is how a post ends up matching nothing anybody types.
What they asked: why does the text come out in one block when I copy it out of the PDF?
What the keyword tool suggests: PDF text extraction best practices.
Only one of those is a question somebody has at eleven at night with a deadline.
The test for a topic worth writing
One question decides it. Can you answer this from having done it?
If the answer would be assembled from three articles you read this morning, skip it. Not for reasons of principle, but because that version already exists and yours will be the fourth copy of it. Nothing in it will be worth linking to.
What makes a piece worth reading is the part that only comes from having been there. The setting that is wrong by default. The step that fails on large files. The thing everybody recommends that stops working once the folder holds two hundred items. You cannot research those, which is precisely why they are valuable.
This test disqualifies a lot of topics you would like to cover, and that is the point. A short list of things you can answer properly is a content strategy. A long list of things you could write about is a document.
Cadence is a capacity question
Decide how often to publish by looking at your worst month rather than your best week. The month with the deadline, the illness and the thing that broke on a Friday is the one setting your real rate, because that is the month the schedule has to survive.
Publishing monthly for two years leaves you with twenty four pieces and a habit. Publishing weekly for six weeks leaves you with six pieces and a blog that visibly stopped, which says something about the business you did not mean to say.
If you want a sanity check on what a plan actually costs before committing to it, the arithmetic we use for tools applies here too: count the hours it really takes, including the two you lose to finding a decent image.
What order to publish in
Not most popular first. Publish the one you can answer best first, because the early pieces are where you work out what your writing sounds like, and you want that discovery happening on a topic you know cold.
After that, group them. Three related posts linking to each other are worth more than three unrelated ones, both to a reader deciding whether you know the area and to a search engine deciding the same thing. Related does not mean sequential. It means somebody finishing one has an obvious reason to open the next.
If a single project produced several of these questions, it will usually carry more than one piece on its own, which is a separate exercise: turning one project into several things covers how far that goes before it starts repeating.
How to tell whether it is working
Not by traffic in the first fortnight. A new piece spends its first weeks being read by people who already know you, which tells you about your existing audience and almost nothing about the topic.
Three signals are worth more than the visitor count early on. Whether anybody replied. Whether you have started sending the link to people instead of retyping the answer, which is the clearest sign the piece did its job. And whether the questions you get about that subject have changed shape, because a good piece moves the conversation on rather than ending it.
A post you send to somebody twice a month is doing more work than one that had a good first week and has been quiet since.
Give search a few months. Results move slowly, and every judgement made before then is made on the wrong data. Rewriting a headline three times in six weeks is a way of feeling busy while destroying the only signal you had.
What we got wrong
Our first plan was a spreadsheet of thirty topics with a search volume beside each one, sorted descending. We wrote the top four. Two of them were fine and two were assembled from other people’s articles, because we had committed to the row rather than to the answer.
The pieces that went on to do anything were not from that list at all. They came from questions we had already answered in email, which we only noticed months later while looking for something else in the sent folder.
The correction was small and slightly embarrassing. Keep a file, add the question the moment it is asked, in the words it was asked in. That file has been a better editorial calendar than the spreadsheet ever was, and it costs about a minute a week to maintain.
What to do this week
Do not plan twelve months. Open a file, go back through three months of sent messages, and pull out every question somebody asked you. Aim for fifteen. Beside each one, note how long your answer was.
Take the three longest. Apply the test: can you answer this from having done it? Write the one where the answer is most obviously yes.
Then keep the file open. The next question arrives this week, and the only thing standing between it and a post is somebody writing it down.
Frequently asked questions
How often should I publish?
As often as you can manage in your worst month rather than your best one. A schedule set during a quiet week becomes a source of guilt during a busy one, and an abandoned blog reads worse to a visitor than a slow one. Once a month, kept for two years, beats weekly for six weeks and then nothing.
Do I still need keyword research?
Yes, but later. It is a good editor and a poor author. Use it after you have the topic, to check the words real people use for the thing, to find phrasing you would not have guessed, and to see whether the question is asked in a form you did not consider. Starting there is what produces the list of fifty topics nobody asked for.
What if nobody ever asks me anything?
Then you are looking in the wrong place rather than sitting on an empty well. Questions arrive as complaints, as confusion in a call, as the thing someone gets wrong every time. Write down what you searched for last week and could not find a straight answer to. That is a question, and you were the one asking it.
Should I write about topics my competitors already cover?
Yes, if you can answer from having done the thing rather than from having read about it. An overlapping topic is evidence that the question is real. What makes yours worth reading is the specific detail, the failure you hit, and the part everybody leaves out because it is awkward to admit.
How long should a post be?
As long as the answer takes and no longer. Length targets are what people reach for when they cannot tell whether a piece is finished. A short piece that answers the question completely will outlast a long one padded to hit a number, and it is cheaper for you to write.
When do I know a topic did not work?
Give it a few months before you conclude anything, because search results move slowly and the first weeks tell you about your own audience rather than about the topic. If a piece is still flat after that, look at whether the question was one people actually ask before you rewrite the headline for the third time.
How this was put together: this is the process we use for this blog, including the spreadsheet that did not work. There are no traffic figures or conversion rates here, because ours would describe a small site in a narrow niche and would not transfer to yours. Anything specific enough to be useful, such as how long a given post took to rank, is specific enough to be misleading somewhere else.