Start with the leading insight
A dossier brings complaints, competitor gaps and commercial signals into one research view. Start with the top opportunity card, then read the quote behind it. Identify the task the person was trying to finish, the obstacle they encountered and the workaround they used. A specific workflow matters more than an angry adjective. The card is a starting point for investigation, not a recommendation to build.
Open the source link before interpreting a complaint. App Store links identify the app listing rather than a permanent link to that individual review. Reddit links point to the original discussion where available. Check the date, surrounding context and relevant product version. A complaint about an older release may describe a problem that has already been fixed.
Read the sample before the score
Evidence counts describe stored excerpts, not unique customers or the size of a market. One person can post more than once, and people who leave negative reviews are not a representative sample of all users. Collection covers limited recent pages and configured sources. A niche with more excerpts may simply have better source coverage.
Frustration scores summarize the severity of the collected complaints. They are derived analysis, not customer satisfaction surveys. A high score from three reviews deserves a different level of attention than the same score supported by dozens of distinct workflows. Use the confidence label as a reminder to inspect sample size. It does not provide a statistical confidence interval.
Understand opportunity scores
Each cluster's Opportunity Score combines frustration, evidence volume, trend movement and commercial signal. The default weights are 40%, 25%, 20% and 15%, respectively. The score tooltip explains the inputs. Missing components are omitted and the remaining weights are re-normalized, so read the coverage line alongside the number. A score based on fewer inputs carries less context.
The workspace ranking uses a separate composite with editable weights. Its purpose is to help compare tracked niches under explicit assumptions. Neither score estimates the probability of success. Scores can change when evidence, source coverage or weights change. Compare the underlying workflows as well as the rank, and keep a written reason for choosing a problem.
Interpret movement carefully
Trend badges compare complaint volume in successful source snapshots. Worsening means the sampled volume increased beyond the configured threshold; improving means fewer complaints appeared in the sample. Neither establishes a population-wide trend. New means a baseline or newly observed evidence, and a single snapshot cannot establish sustained movement.
Check source health beside the dossier header and open the sync details when needed. A blocked source is a collection limitation, not evidence that complaints disappeared. A competitor with zero low-rated reviews may be healthy, or relevant complaints may sit outside the fetched pages. Compare collection windows before comparing products.
Treat commercial signals as proxies
Search volume measures interest in a keyword, and CPC reflects advertiser bidding for that keyword. Neither directly measures software revenue or willingness to switch. Read the exact keyword, provider and fetched-at date. Manual entries are researcher-supplied; fixtures are fictional examples. Missing keyword data should remain missing rather than being interpreted as zero demand.
Revenue scenarios multiply visible assumptions about affected users, conversion and price. Rating counts are an adoption proxy, not a customer census. Change the assumptions to see how sensitive the result is. A large modeled number tells you to investigate those assumptions; it does not turn the scenario into a forecast.
Turn the job into a conversation
Read the Jobs-To-Be-Done statement as a draft: when someone faces a situation, they want an action that produces an outcome. Test it in interviews. Ask about the last time the problem happened, the steps they took, what they tried and the time or money they already spent. Avoid pitching your solution or asking whether they like an idea.
Use the interview script to guide the conversation, then follow concrete details rather than reading every question mechanically. When recruiting from original discussions, contact people respectfully and follow platform rules. The recruitment kit provides a template for manual outreach; it does not send messages for you.
Decide what evidence would change your mind
Before building, define a small validation experiment and a stopping rule. Look for repeated problems among a specific buyer group, existing effort to solve them and a credible route to reach those buyers. Record contradictory evidence too. A complaint can be serious without supporting a viable business.
Explore the demo dossier to practice these steps with labeled sample data. When you need your own tracked sources and plan-specific research access, review pricing. The strongest outcome of a dossier is a clearer next experiment, including the decision to stop researching an idea.