This site separates captured facts, interpretations and untested hypotheses. Current research uses AI-assisted source analysis; no named human sign-off or hands-on product testing is recorded for the published analyses.
Counts can be checked against the published evidence. A developer’s capability claim establishes what was advertised, not whether it works. A research label does not turn either into validated demand.
The five labels
Scroll horizontally to see all columns
| Label | Means | Example |
|---|---|---|
| Fact | A captured field or reproducible count. A listing claim is a fact about what was advertised. | “353 ratings in the 23 October 2026 snapshot.” |
| Inference | Our reading of what the facts imply. Arguable. | “The product is designed to quote before leaving the site.” |
| Hypothesis | A decision we would make if the inference holds. Untested. | “Teams would pay for a weekly niche-change alert.” |
| Review | Source analysis with a stated sample, process and date; the word does not certify a human sign-off. | “AI-assisted analysis of sixteen full descriptions; no product testing.” |
| Correction | A change made after publication, described rather than hidden. | “Reclassified as adjacent; see note.” |
The labels are strict about one thing: a hypothesis never gets promoted to a fact because it is repeated across pages. If a paragraph needs three inferences stacked on each other to work, it does not belong on the site yet.
App Store facts and Google keyword reports
The App Store snapshot records public listing fields captured on 23 October 2026. A Google keyword report is separate provider research with its own seed phrases, market, language and capture date. Saving a keyword report does not update the Apple facts or the published source-analysis date.
Search volume, keyword difficulty and CPC are provider estimates. They are not app downloads, revenue, unique customers or proof that a SaaS direction is viable. When a public conclusion uses a report, the relevant values, scope, date and limitations belong in that page's readable evidence.
Official sample reports can be read in the catalog without starting a new provider query. Free accounts do not generate new reports. Private research is planned for Pro and requires an explicit submission and verified paid access when it opens. Open the research catalog · Plans and availability · Public samples and private research.
How AI-assisted research is disclosed
AI assistance is used to classify captured descriptions, compare advertised workflows and draft explanations. Programmatically reading records to recompute tags does not count as individually analyzing every product. Each analysis states its sample and limitations.
Current pages do not record a named human reviewer’s sign-off. If a human review is added, the page will identify the reviewer, date and scope. Installing products, interviewing buyers and validating demand are separate steps and must be recorded when performed.
The publication gate
A niche or direction page goes live only when all of the following are true. These are editorial rules, not a scoring model.
-
The pool is reproducible
The field, matching rule, categories, window and rule version are published with a candidate manifest. The interactive catalog may use a different search mode; any mismatch is stated rather than hidden behind a similar-looking link. The invoicing manifest records its summary-based candidate set.
-
The stated sample was analyzed
The page identifies which full descriptions were analyzed and why they were selected. A complete threshold-selected analysis covers every record above that threshold; a partial progress note reports what remains. Misclassifications stay visible.
-
The claim is one sentence and falsifiable
One paragraph about what the group shares, plus the cheapest observation that would break it.
-
The limits are published
What the snapshot cannot answer appears on the same page as the claim, not in a disclaimer elsewhere.
-
Someone can be told they are wrong
Every page carries a correction route. A correction that changes a claim is described on the page with a date.
Claims we do not make
- No revenue or download estimates. The data does not contain them, and inferred figures are usually wrong in a way that flatters the idea.
- No “validated” ideas. Ratings are not demand. A direction stays a hypothesis until someone talks to buyers.
- No market-size projections. Counting apps measures supply. Demand requires a different dataset that this project does not have.
- No AI-visibility guarantees. We do not promise that any page here will be cited by a search engine or an AI assistant.
- No paid placements, affiliate links or sponsored niches. Inclusion is based on the stated research scope and evidence, not a payment from a developer.
Capture, analysis and correction dates
The snapshot date identifies when the source data was captured. The analysis date identifies the recorded source-analysis work; it does not imply a human sign-off. A text-correction date identifies a wording or scope change and does not refresh the data. Rebuilding the site changes none of these dates automatically.
28 October 2026 correction: removed the earlier guarantee that every review had been read and signed by a person. Current records support AI-assisted source analysis, not that guarantee. Future data refreshes must preserve the original dates or explicitly record which facts were rechecked.
30 October 2026 clarification: distinguished frozen App Store facts from separate Google keyword reports and clarified the Micro SaaS research purpose. This wording change did not recollect app data, rerun source analysis or validate demand.
Public evidence stays public. Published conclusions, their supporting facts, comparisons, sources and corrections do not require registration or payment. Optional personal, bulk or monitoring workflows can have separate limits without hiding the basis of public research.
Where this method is weak
Three honest weaknesses. First, the attention threshold (50 ratings) is arbitrary and was chosen to keep the reading list short; a different threshold would change which products look like the market. Second, everything here reads public marketing copy, so the gap we identify may be a gap in the copywriting rather than in the product. Third, the catalog excludes incumbents by construction, which systematically makes markets look more open than they are. The reviews repeat these weaknesses on the pages where they matter.