The questions worth asking before you start.

Kerfuffle is the primary supported source, with purpose-built extraction rules for its listing structure. That said, the source directory management settings let you add other directory URLs beyond Kerfuffle — G2 PropTech and GetApp PropTech are named expansion targets. Each directory gets its own crawl frequency (one-off, weekly, or monthly), and the scraper plugin architecture lets site-specific extraction rules be defined for new high-value domains. If a site isn't yet in the plugin registry, the crawler falls back to generic heuristics.

No. Every composite score is built from four transparent, separately reported dimensions: replicability (how buildable is this with our stack, 1–10), market demand (review volume, recency, and sentiment ratio, 1–10), revenue potential (pricing tier crossed against PropTech category market size, 1–10), and competitive gap (frequency and severity of recurring negative review themes, 1–10). Each dimension has a written LLM rationale alongside its number, and you control the weighting of each dimension from the settings page. Change the weights and the entire product table re-ranks in the background — no AI re-run required.

Access is restricted to a hardcoded allowlist of approved email addresses. Any sign-in attempt from an address not on that list is rejected with an explicit 'access restricted' message. There is no public sign-up flow. The root URL routes unauthenticated visitors straight to the sign-in page — there is no public-facing marketing funnel to stumble through. It is a deliberately closed internal tool.

A full dossier includes: all four scored dimensions with written rationales, raw review samples, pricing signals normalised to a standard tier (free through enterprise), a feature list extracted from scraping, an AI-drafted mission statement for a competing SaaS, a prioritised 10-item suggested feature list shaped by market gaps, and a competitor weakness analysis drawn from negative review patterns. Each of those three text artefacts has a copy-to-clipboard button that outputs formatted Markdown. You can also export the entire dossier as a Markdown file or PDF from the same page. Dossier version history is retained every time a product is re-scored, so you can track how the opportunity has shifted over time.

Several signals surface movement automatically. Monitored directories run a weekly re-crawl and flag any newly discovered products with a 'new since last crawl' badge on the dashboard. When a product's composite score moves more than 1.5 points between runs, a directional indicator (▲ or ▼) appears on its dashboard row. If a shortlisted product drops significantly, you can opt into an in-app notification. A weekly email digest also lands in your inbox summarising the top new products discovered, the top scorers, and any products with notable score swings — configurable per user in notification settings.

Yes. When starting a crawl you choose the depth: 'Skim only' discovers listings and pulls surface data in seconds with minimal AI cost, while 'Full analysis' runs the complete scrape, score, and dossier pipeline. The typical workflow is to skim first to gauge directory size, then review the auto-flagged top candidates before confirming the expensive deep-analysis run. AI token usage and estimated cost are tracked per crawl run and broken down by pipeline stage on the run detail page. Cumulative monthly spend is logged, and you can set a cost threshold — the default is $5 per run — that triggers an in-app alert if a single run exceeds it.