Most first-time sellers research a problem by asking people whether they would buy something. That produces polite answers and no evidence. A better question is smaller and harder to dodge: where does this problem already leave a trace? A trace might be a workaround, a recurring task, a half-finished file, a question someone keeps asking, or a step people skip because it is tedious. Traces are things you can observe. Opinions are not.
This guide covers the research that happens before you build or describe an offer. It starts from a hunch and ends with a one-page research log you can use to decide whether to keep going, narrow the idea, or drop it. If you already have a problem you want to phrase clearly, how to define the problem your offer solves covers the wording. This is a planning guide, not legal advice, and privacy and consent rules vary by place and by situation.
Turn the hunch into a question you can be wrong about
“Small teams need better file organization” cannot be wrong, which means research cannot test it. Rewrite the hunch as a claim with a visible failure condition: “Small teams that receive approved files in several formats lose time deciding which version is current.” Now you know what to look for, and you know what would count against it.
Add two more lines before you begin. Write the specific situation you are studying, and write what you would need to see to drop the idea. Deciding the drop condition in advance is the best protection you have against reading every signal as encouragement.
Rank your sources by how much they can prove
Not all evidence carries the same weight. Behavior beats statements, and repeated patterns beat single stories. Use this ladder to decide how much any one source should move you.
| Source | What it can show | What it cannot show | Weight |
|---|---|---|---|
| Your own direct experience with the workflow | The steps, the friction, and where work stalls | Whether other people share the problem | Medium for the process, low for demand |
| Permission-cleared artifacts, such as a shared template or a sample file | How the work is actually done, including workarounds | Why it is done that way | High for behavior |
| Public discussions, such as forum questions and help requests | Recurring confusion and the words people use | How many people are affected, or whether they would pay | Medium for language, low for size |
| Consented conversations | Context, cost, and what has already been tried | What people will do, as opposed to what they say | Medium to high when they describe past events |
| Stated interest, such as “I’d definitely use that” | Politeness | Anything about demand | Very low |
Anchor your research in the top three rows. Use conversations to fill in context the artifacts cannot give you. Treat enthusiasm as noise.
Ask about the last time it happened
When you do talk to someone, ask about a specific recent occasion. “Tell me about the last time you had to find the right version of a file” produces a story with steps in it. “Is file organization a problem for you?” produces a yes. Follow the story with neutral questions: what happened just before, what you did next, what you tried, and how you knew it was finished.
Do not describe your idea during these conversations. The moment you do, the person starts responding to your idea and stops describing their own situation. If it helps to keep the questions neutral, ask them as a checklist you work through, and save any pitch for a separate, clearly labeled conversation.
Keep consent and privacy rules attached to every note
Research involves other people’s information, so the rules go in your notes from the start, not in an afterthought.
- Say why you are asking. Tell people it is research, and do not present it as a sales call.
- Get permission before recording or saving anything, including what they say and what they show you.
- Do not request anything you do not need. You are looking for process and friction, so skip names, account details, and confidential figures.
- Keep employer and client material out. If a source belongs to someone else, ask whether you may use it. If the answer is unclear, do not.
- Write notes so a stranger could not identify the person. Record the situation, not the individual.
Public posts are not a free-for-all either. A forum question is a good source of wording, but you should paraphrase it and leave out anything that identifies the writer.
Separate what you saw from what you think it means
This is the habit that separates research from confirmation. For every note, write two lines: the observation (what happened or what was said) and the interpretation (what you think it means). Keep them in separate columns. Observations can be checked by someone else. Interpretations are yours to defend.
Then look for repetition. One story is a lead. The same friction showing up in several unrelated sources is a pattern. Keep counting honest: “three of five sources mentioned a workaround” is useful, while a percentage built from five sources is not, and you should not present it as a market statistic.
A synthetic example: the “which version is current” problem
This is a fully synthetic example built to show the method, not a record of any real research. A seller suspects that small teams lose time deciding which file version is current.
| Source | Observation | Interpretation | Counts for or against |
|---|---|---|---|
| Own walk-through of a shared folder set | Three files with near-identical names and different edit dates | Naming rules are probably missing | For, but it is only one setup |
| Public help request | A writer asks how others mark a “final” file | Others share the confusion | For the language, not the size |
| Consented conversation about the last occurrence | The person re-sent a file because nobody knew which copy was approved | The cost is rework and delay | For |
| Consented conversation, second person | Uses a single shared document and has no problem | The problem depends on how files are shared | Against, and it narrows the target |
The last row is the most valuable one. It does not kill the idea. It tells the seller that the problem lives in teams that pass files by message, not in teams with one shared document. That is a narrower, more honest starting point. Once the problem is clear, how to match an offer to evidence you already have helps you check whether you can actually support a service around it.
Three ways early research goes wrong
Researching only friendly sources. Friends, followers, and past colleagues tend to agree with you. Include at least one source that has no reason to be kind.
Mistaking a complaint for a purchase. People complain about many things they would never pay to fix. A real problem usually leaves a workaround, a repeated cost, or a half-built fix. A casual gripe leaves only words.
Letting research become the project. More notes feel productive. If new sources have stopped changing your picture, the next step is a small test, not another round of interviews.
When to stop and what to do next
Stop researching and decide when new sources keep repeating what you already know. At that point you have three honest options: narrow the idea to the situation where the evidence is strongest, drop it because your drop condition was met, or move to a small scoped offer. For the last option, how to reduce an offer to a small first version shows how to keep that first version small enough to deliver and review.
Pause immediately if the problem turns out to be mainly legal, medical, financial, or security-critical and you are not qualified to work on it, or if your sources require private data you cannot protect. Research that finds a good reason to say no has still done its job.
The problem research log
Copy this into a document or spreadsheet. Complete the top block once, then add one row per source.
| Field | What to record |
|---|---|
| Claim being tested | One sentence with a visible failure condition |
| Situation studied | The specific workflow or setting |
| Drop condition | What you would need to see to abandon or reshape the idea |
| Source type | Own experience, artifact, public discussion, or consented conversation |
| Permission status | Cleared, public and paraphrased, or not usable |
| Observation | What happened or was said, with no identifying details |
| Interpretation | What you think it means, kept separate |
| For, against, or narrows | How it affects the claim |
| Pattern count | How many separate sources show the same friction |
| Decision | Continue, narrow, test small, or drop |
The log passes when every row has a source you were allowed to use, every observation is separate from your interpretation, and at least one entry argues against your idea. A log full of agreement has not been tested, so go find the entry that disagrees.
