All Docs
FeaturesAgentOS Scope OutUpdated March 15, 2026

Top-N Auto-Deep-Analysis Selector

Top-N Auto-Deep-Analysis Selector

Available from v0.1.72

Overview

After a skim crawl of a supplier directory (e.g. Kerfuffle.com), the platform automatically identifies the most promising products and queues them for deep analysis. This removes the need to manually sift through every listing — the system surfaces the top candidates based on initial signals, and you confirm before the expensive deep-scrape pipeline runs.

How It Works

1. Paste a Directory URL

Enter a proptech supplier directory URL into the crawler input. The skim crawl runs across all listings, collecting lightweight surface signals for each product.

2. Automatic Scoring & Flagging

Once the skim completes, every discovered product is scored using two surface signals:

SignalDescription
Review countHigher volume of reviews indicates stronger market presence and data richness
Pricing tierSignals revenue potential and market positioning

The top N products by this surface score are automatically flagged as candidates for deep analysis. N defaults to 20 and is configurable.

3. Review the Flagged List

Before any deep-scrape work begins, the platform presents the auto-flagged list for your review. At this step you can:

  • Confirm the list as-is to launch the deep-scrape pipeline
  • Add products that didn't make the cut but interest you
  • Remove products you want to exclude
  • Reorder candidates by priority

This human checkpoint ensures the more resource-intensive deep-analysis stage is only triggered intentionally.

4. Deep-Analysis Pipeline

Once confirmed, the deep-scrape pipeline runs against the finalised list. Each product receives a full scored dossier covering:

  • Replicability — how easy to build with our stack
  • Market demand — review volume and sentiment
  • Revenue potential — pricing tier and market size
  • Competitive gaps — what negative reviews reveal

Results appear on the ranked dashboard, and each product's full dossier is available for drill-down.

Configuration

SettingDefaultDescription
TOP_N_CANDIDATES20Number of products automatically flagged after a skim crawl

Notes

  • Surface scoring (review count + pricing tier) is a lightweight pre-filter. It is not the same as the full composite opportunity score, which is only computed during deep analysis.
  • The confirm step is mandatory — the deep-scrape pipeline cannot be triggered without an explicit user confirmation of the candidate list.