EDMONTON, AB — May 14, 2026 — Napkin Deals today detailed the second generation of its deal-aggregation engine, the system the team calls “the well.” Listings now move through eight automated stages — discovery, scraping, AI extraction, industry classification, deduplication, currency normalization, embedding, and publication — with no human in the loop.
An index that cleans itself
Marketplace data rots quickly: listings sell, expire, and reappear under new names. The v2 engine fights back nightly. A phantom-listing sweeper retires deals that have quietly disappeared from their source. Per-source health checks automatically disable feeds that degrade. Cross-listing detection collapses the same business advertised on multiple sites into a single canonical record, and re-listing matching preserves price history when a deal returns at a new number.
Built for a global market
Because the lower middle market is global, the pipeline normalizes asking prices and financials across 30 currencies using daily exchange rates, and classifies every listing into a standard industry taxonomy. More than 600,000 listing scans have been attempted to date, and the corpus compounds daily.
The engineering principle behind the release is simple: aggregation is table stakes, but a self-healing index is a moat. Every stage that removes a human touchpoint drops the marginal cost of covering one more business — and that cost now rounds to pennies.