METHODOLOGY
Evidence over hype.
Every published card starts with a real request for an app, website, or tool. We show a short original excerpt, a source link, and a verified request date; the proposed build is our interpretation.
How a signal becomes an opportunity
- Source responsibly. We use permitted public APIs and public pages, honor access rules and rate limits, keep a minimal excerpt, and retain the original source in our private review tools. Signed-in visitors see a short attributed request excerpt, its platform, and a link to the original public post. We do not separately publish source usernames or complete raw posts.
- Normalize and deduplicate. Repeated complaints are grouped by workflow and buyer. A unique source cannot be published twice.
- Score the evidence. A reviewer applies the published rubric below and records a plain-English rationale. The rationale matters more than the number.
- Review before publication. The source, original date, excerpt, audience, existing alternatives, proposed first version, and first test are checked before publication. Automated intake remains in a private review queue. Versioned source-checked editorial collections are reviewed separately from that intake.
The 100-point evidence rubric
| Factor | Points | What earns the score |
|---|---|---|
| Intent and specificity | 0–25 | An explicit problem, search, workaround, feature request, or buying signal tied to a concrete job. |
| Buyer clarity | 0–15 | A reachable role or segment is identifiable from the source and workflow. |
| Urgency and frequency | 0–15 | The pain is costly, risky, time-sensitive, or repeats often enough to justify action. |
| Repetition | 0–15 | Independent reactions, comments, or corroborating sources show the signal is not isolated. |
| Reachability | 0–10 | Prospective users can be found through a focused community, profession, or installed product. |
| Existing alternatives | 0–10 | Users already spend money, time, or effort on a workaround and can describe its gap. |
| Recency and source quality | 0–10 | The source is current, inspectable, attributable, and specific enough to verify. |
A verified request can still have a low opportunity score: existing software may solve it, or buying intent may be unknown. Scores from 70–84 indicate stronger evidence for customer discovery. Scores of 85 or more indicate unusually strong observable evidence—not guaranteed demand. Replies and upvotes are not counted as independent requests.
What “First asked” means
This is the earliest original publication date we can verify in the reviewed sources linked to that request. When reviewed requests share a problem cluster, we use the earliest dated source in that cluster and retain its link. Older requests may exist elsewhere. Collection dates, import dates, and reply dates never stand in for the original request date. Unknown dates remain unknown.
Source coverage
Collectors support Reddit, X, Bluesky, Hacker News, and GitHub through their permitted APIs or public feeds. Access and availability vary by platform. The library also includes individually checked public forum requests. Source filters report published requests, not a promise that every platform has current results. Existing alternatives, author updates, and historical workarounds remain part of the context.
What a score means
A score reflects the strength and usefulness of observable evidence. It is not a market-size estimate, valuation, revenue forecast, investment recommendation, or guarantee that a product will succeed. A single post can be useful evidence, but never proves a market.
How AI may be used
AI may help normalize, classify, or summarize research. It may not invent a quote, source fact, buyer, demand signal, market number, or willingness to pay. Source review remains required before publication; automatically collected posts are never published directly.
Corrections and removals
Every source subject or visitor can submit a correction or removal request. Include the opportunity title or ID so we can find the underlying research. We record the request for human review and can archive a card without deleting the original request history.
Methodology last reviewed: November 2, 2026.