Configuration

COLORS
CUSTOM CURSOR
Skip to main content

espialtech

PriceLens

No Comments
PriceLens
Market Intelligence

Competitor signals are fragmented, noisy and location-dependent.
PriceLens turns them into a brief that tells you what to do next — and shows its working.

Project Details

Storefront HTML drifts constantly, pricing varies by location and delivery postcode, reviews are noisy, and stock availability is reported inconsistently across merchants. Any system that reads this data has to be resilient before it can be intelligent. PriceLens uses fetcher orchestration with Playwright and Puppeteer modes, proxy rotation, retries and deterministic request profiling to survive that drift.

Structured extraction normalises product, price, availability, delivery and review data into typed records. A monitoring graph models location → category → competitor edges with schedule-aware crawling. On top of that sits the AI layer: review sentiment and demand signals via confidence-calibrated extraction, price elasticity hints grounded in stock trajectory and review velocity, and generated strategy briefs that produce concrete "what to do next" recommendations with the assumptions and evidence references attached.

Project Research

The novel piece is confidence scoring at the scrape-to-insight boundary. When a storefront changes its markup, a naive pipeline does not fail loudly — it silently starts extracting the wrong number, and the insight built on top of it is confidently wrong.

PriceLens gates every insight on the reliability of the data underneath it, and refuses to generate a brief on data it does not trust.

Project Results

PriceLens automates 85% of the market monitoring workload, renders insights in under two seconds, cuts manual effort by 60% and shortens review cycles by 40%. Analysts stopped assembling the data and started arguing about what it means, which is the point.