1. Define the problem, not the product
Write the problem in the words the affected person would use, with no mention of a solution. "Small shops cannot reconcile bank transfers against orders without typing them in twice" is a problem. "An AI reconciliation tool" is a product, and analysing a product tells you only whether you like your own idea.
The test is whether someone who has the problem would recognise themselves in the sentence. If the sentence contains a technology, rewrite it.
2. Size the demand from evidence, not from a report
The industry figure you can cite in one line — "a $47 billion market growing at 12%" — is almost never the market you are entering. It is a category label that includes incumbents, adjacent products and revenue you could never reach. It is useful for a slide and useless for a decision.
Count what you can observe instead: how many distinct people described this problem in public in the last ninety days, in how many separate places. It will be a small number and it will be real, and a small real number is a far better basis for a decision than a large borrowed one.
3. Measure pain, not interest
Pain is what the problem costs today. Look for three things in the evidence: a workaround someone built, an hour count, or a sum of money. Someone who wrote a script to work around the problem has paid for it in labour and told you the price. Someone who says a category sounds interesting has told you nothing.
4. Map the competition honestly
List every product a person with this problem could buy today, including the spreadsheet and the intern. Then find, for each, the specific reason the people in your evidence are not using it: price, missing feature, wrong language, wrong segment, or they tried it and left. That list of reasons is the actual specification of your product, and it is worth more than any feature brainstorm.
"No competitors" is almost always a research failure. A market with genuinely no one in it is more often a market with no money in it.
5. Test willingness to pay before you build
Find evidence of money already moving: a subscription someone resents paying, a competitor's price that gets quoted in complaints, a person asking what a solution would cost. If money appears nowhere in ninety days of evidence, treat that as a finding rather than an omission — you may have found a real problem in a market that has never paid for anything.
6. Check the trend over the last quarter
Split the evidence in half by date and compare the two halves. A market whose second half is thinner than its first is a market you will enter late. This one comparison catches most of the opportunities that look attractive on a total and are actually finished.
7. Write it as a scorecard, not an essay
End with one line per dimension — demand, pain, willingness to pay, competition, momentum — each with a number and a link to the evidence behind it. The point of the number is not precision; it is that a colleague can disagree with a specific figure instead of with your general feeling, and that you can compare two opportunities side by side six weeks later.
That scorecard is exactly what an opportunity page on this site is: seven measured sub-scores, each one clickable down to the posts it was computed from.