Oleksandr TverdokhliebLeaflet Analyzer
A retail promotional-intelligence pipeline that turns weekly promo leaflets into structured product, price, and discount data — so pricing and trade teams see competitor moves while they still matter.
My role: Designed and built the extraction pipeline.
Executive summary
Around 70% of trade-promotion spend fails to pay off, and competitive promo data usually arrives too late to react to. The teams who see competitors' prices first are the ones who rebalance fastest.
pyLeaflet Analyzer closes that gap: it ingests retailers' weekly promo leaflets and turns them into clean, queryable data — products, brands, current and previous prices, discount percentages, validity dates, and categories — ready for competitive analytics.
Technical implementation
The system runs as a five-stage, resumable pipeline. Each stage keeps SQL views of what's left to do, so processing is idempotent — it can stop and restart without redoing work.
- CapturePulls leaflets from aggregator sites and retailer microsites.
- ArchiveStores source images and records in SQLite with versioned image storage.
- ExtractClaude vision AI with structured outputs parses PDFs and images into normalized rows.
- NormalizeBrand and category tagging plus deduplication.
- DeliverOutputs to a database, an API, and dashboards.
Who it's for
Distribution companies tracking retailer positioning, CPG brand managers auditing promo compliance, pricing teams feeding competitor data into forecasts, and category managers benchmarking merchandising. It currently ships as a self-hostable Python package with UAE retail-aggregator integration; a managed service, multi-region sources, and price-change alerts are on the roadmap.
Tech stack
Outcome
Promo intelligence that used to take days of manual collation becomes a near-real-time data feed — turning a pile of weekly leaflets into a competitive-pricing dataset teams can actually act on.