Niche Research

Customer Research Before You Write a Single Chapter

By AI Creation Forge · 7 min read

The Direct Answer

Before you write a chapter, collect the questions your reader already asks, in their words, and sort those questions into jobs. The guide's outline is that sorted list. Research is finished when the same jobs keep showing up and you can name who the guide is not for. It is not finished when you have read everything.

Picture Noah at a cafe on Saturday at 8 a.m. with a notes app open and a cooling coffee. He has bookmarked forty threads about "starting a dog-walking route." He has not written a sentence of the guide. He is about to do the only research task that changes the outline: copy the questions people actually asked, strip his own opinions out, and group the questions until a chapter list appears.

Copy questions, not opinions

Open two or three public places where the reader already talks. A forum thread, a product review, a question box on a video. For one hour, copy questions and complaints into a note. Copy them as written, including the awkward phrasing. "How do I tell a client I will not do holiday boarding" is more useful than Noah's paraphrase "set boundaries."

Do not collect advice from other creators yet. Advice is someone else's outline. You are trying to hear the buyer. When a review says "I wish the book had included a script for the first client email," that sentence is a possible section. When a review says "great read," it is not.

Skip any urge to count the posts and call the count a market size. A pile of repeated questions is a signal. A number you invented from a glance at a screen is not a source. If you want a first pass on whether people already pay for this topic, use the niche calculator and then do the manual check described in how to find a profitable niche. The calculator is a prompt. The copied questions are the research.

You might be thinking you should interview people before you read anything public. A conversation is useful later, when you know which questions to ask. Starting with public questions keeps you from interviewing your friends about a guide they would never buy.

Sort the questions into jobs

When you have a few dozen lines, stop collecting. Read them and write a short label for the job hiding under each one. Noah's labels might be: set a service area, price a walk, write the first client note, handle a cancellation, decide what you will not offer. Five labels is a good sign. Thirty labels means you have not sorted. Merge labels that are the same job in different words.

Under each label, keep two or three of the original questions so the chapter still sounds like the reader. The heading in the guide can be cleaner than the raw question. The example inside the chapter should use the reader's situation, not a generic small-business story.

Throw out questions that are really a different product. "Which GPS collar should I buy" may be a gear review, not a guide about running a walking route. If you cannot teach it from work you understand and sources you can open, it does not become a chapter. Mark it as out of scope in one line so you do not wander back to it at midnight.

Write the person you will not serve

Research includes a refusal. Look at the questions that would pull the guide into a second book. Noah sees questions about overnight boarding, pet first aid, and hiring a staff of walkers. He writes at the top of the note: this guide is for one person walking dogs in a neighborhood they already live in. It is not a boarding manual, not a veterinary guide, and not a hiring handbook.

That refusal is what you will repeat on the sales page. It is also how you stop the draft from doubling. A question can be sincere and still be the wrong question for this file. You can point readers to a better source without pretending you covered it.

If your notes only contain your own questions, you have not done this step. Your questions matter, and they are not a substitute for the reader's. Add public lines before you outline. The Forge Method answer is a short map of the wider workflow if you want to see where research sits next to the outline and the page. It does not replace the notes.

Turn jobs into a chapter list, then stop

Order the jobs the way the reader will hit them. Noah's reader meets a neighbor before they need a cancellation policy, so the first client note comes before the cancellation chapter. Pricing might come before or after the first email, depending on the questions. Let the notes decide. Do not use a template order from a business book if the questions arrive in a different sequence.

Each job becomes one chapter sentence: after this chapter, the reader can ___. Take those sentences into the one-afternoon outline. Research that never becomes chapter sentences is a hobby. Chapter sentences that ignore the notes are a guess.

Stop when new threads repeat labels you already have. One more hour of reading will feel productive and will not change the list. Write the date you stopped on the note. If you reopen research later, you will know you are adding a fact, not restarting the guide.

Check the facts the chapters will depend on. If a chapter will mention a city permit, a payment app's fee, or a platform rule, put the official page on a proof list. Do not draft that paragraph from memory. Customer research tells you the question. It does not supply the answer to a rule you have not read.

Use one conversation to test the list

After the sort, talk to one person who matches the reader. Show them the chapter sentences, not a pitch. Ask which job they would need this week, and which sentence they do not understand. Noah asks a neighbor who already walks two dogs for friends. She says the cancellation sentence is clear and the service-area sentence is not, because she does not know whether "neighborhood" means her building or her whole borough. He rewrites that sentence before he drafts.

Do not turn that conversation into a testimonial. One neighbor is not a study. Thank them. Fix the sentence. If you later want a quote on the sales page, it has to be their words, their honest experience, and any connection has to be disclosed. Do not write the quote for them in the cafe.

When the chapter sentences match the questions and the refusal is written down, you are allowed to draft. Open the file to chapter one of the job list, not to a history of dog walking. The reader did not ask for the history. They asked how to send the first note and how to say no.

Before you write that chapter, look at your notes and answer this: which question, copied in someone else's words, does the chapter exist to answer? If you cannot point to one, you are not done with research. You are decorating.

FAQ

How many questions do I need? Enough that the same jobs repeat and you can see what to leave out. A dozen specific questions that cluster are more useful than a hundred vague ones.

Can I research in my own comments or inbox? Yes, if those people are the buyers. Still look at one public place they did not write for you. Your inbox agrees with you too easily.

What if the questions disagree? That is the point of the refusal. Pick the cluster you will serve and write the other cluster on the "not for" line. Do not average them into a guide for everyone.

Should I pay for a keyword tool first? You can, after you have the jobs. A keyword list without the reader's sentences produces chapters titled like search queries. The niche article explains how to use search as a check, not as the outline.

Do I quote the threads in the guide? Paraphrase the situation and do not paste someone's post into your PDF. You are learning the job, not republishing the thread.

When do I talk about price? After you know the job. Price is easier to judge when you can say what the file helps someone do. The pricing guide comes after that sentence exists.