The five questions research has to answer
Market research is not a document, it is five answers. If a report does not contain all five, it is a literature review with a chart in it, and the decision it was written for is still unmade.
- Who has this problem, and how many of them are there — counted as distinct people, not as a market-size figure copied from a press release.
- What it costs them today: money spent, hours lost, or a workaround they maintain by hand.
- What they use instead right now, and what they say about it when they are not being sold to.
- Whether money already moves: someone paying for a worse answer is the single strongest signal in any research file.
- Which direction the last few months point, separately from the total. A large declining market and a small growing one look identical in a size estimate.
Where the free evidence actually is
The useful material is writing people produced for their own reasons, before they knew anyone was studying them. That rules out anything written to be read by a buyer, and rules in support forums, issue trackers, app-store reviews, question sites and the complaint threads of communities built around a job rather than a product.
For a Vietnamese market the same rule holds with different addresses: app-store reviews in Vietnamese, Facebook groups organised around a trade, and the support threads of the services people already pay for. The language is the barrier that keeps most of this out of English-language research, which is exactly why it is still worth reading.
Why surveys and interviews mislead
A survey asks a person to predict their own future behaviour, which people are reliably bad at, and it asks in a setting where being encouraging is free. The answer you get back is what the respondent believes a reasonable person would say. It is not evidence about a market, it is evidence about politeness.
An unprompted complaint has the opposite property. Nobody writes a four-paragraph post about a broken workflow to be agreeable; they write it because the problem cost them something that day. The cost is already paid before you arrive, which is what makes the record trustworthy.
Interviews are worth doing, but after the evidence, not instead of it — and to ask what someone did last month, never what they would do next month.
Turning evidence into a market analysis
A market analysis is the step where scattered reading becomes a number you can compare against another number. The mechanics are unglamorous: group every post that describes the same underlying problem regardless of wording, count distinct authors rather than posts, then check each group against the supply side — who sells into it, at what price, and what their own users complain about.
The grouping step is where manual research usually fails. One problem is written thirty ways, and a person searching by keyword finds the phrasing they happened to think of and misses the twenty-nine they did not — which makes a real market look like thirty unrelated small ones.
Reading a market trend without a keyword tool
Search-volume tools measure how many people typed a phrase, which is a measure of vocabulary, not of demand. A problem nobody has named yet has no search volume at all, and by the time it does the market is contested. Read the trend off the evidence instead: the number of distinct people raising a problem this quarter against last, and whether new posts are still arriving after the product that was supposed to fix it shipped.
Two directions matter and they are different questions. Momentum is whether the rate is rising. Persistence is whether the problem survives the arrival of a solution — and persistence is the one that predicts whether a market is still there in two years.
How this site runs the same process
Opportunity Market performs these steps continuously rather than once per project: it collects public posts, clusters different phrasings of one problem together, counts distinct authors, and scores each cluster on demand, pain, momentum, competition, monetization potential, market size and evidence quality. Every figure links back to the posts it came from, so the research is checkable rather than assertable.