Methodology
Every niche on NovelRadar gets a single headline number — the composite — that's the weighted geometric mean of four independent axes. We use a geometric mean (not arithmetic) because we want to reward niches that win on multiple dimensions, not ones that spike on one and crater on another.
Default weights are 1/1/1/1. Paid users can adjust weights live and re-rank all niches in <100ms.
Reader appetite minus title supply. High = hungry market, thin shelf.
90-day momentum across Royal Road, Reddit, and NovelUpdates.
KU penetration, paperback price depth, audiobook arbitrage.
Series friendliness and cross-channel readability.
For Amazon / KDP data, we display only what's licensed via the Amazon Product Advertising API (PA-API) — covers, titles, BSR, KU status. We scrape public category pages for scoring-only internal signals (e.g., top-100 BSR density per niche keyword), but those signals never appear on the public page. The signals.is_displayable column on every row enforces this at the database layer.
Royal Road and NovelUpdates data is republished as public facts (chapter counts, follower counts) with link-out to the source. Reddit thread titles are quoted under fair use with permalinks.