Dedicated analytics system created in 2012 for automatic analysis of sales data from a popular auction-sales platform
A dedicated analytics system designed for automatic monitoring of sales and interest in offers on the platform. Ingestia continuously collected data, analyzing millions of active auctions, tracking both sales and the level of user interest (number of auction views).
Key features:
- Continuous collection of data on online auctions.
- Tracking offer views during the auction period.
- Automatic monitoring of auction status changes (sales, closure).
- Analysis of all categories, with a priority on selected segments.
- Dynamic auction queue – optimized data collection.
- Thematic segmentation of data for better performance.
Main project challenges:
- No official API for the classifieds portal in 2012.
- Processing millions of active offers with limited memory resources.
- Ensuring data collection continuity despite changes in the Allegro site structure.
- Optimizing Excel performance for large datasets.
How Ingestia was used in practice:
- Monitoring competitors and analyzing prices in real time.
- Identifying best-selling products in individual segments.
- Analyzing the popularity of new offers and speed of sales.
- Making pricing investment decisions based on hard market data.
- Supporting sales and marketing campaign planning.
What made Ingestia unique?
- Continuous data collection without API access.
- Tracking not only sales, but also auction interest (offer views).
- Dynamic auction queue – optimized for millions of active records.
- Scalable approach: segmentation by topic and dedicated dashboards.
- Practical approach: system built for daily commercial decision-making.
Scalability and technical challenges:
- The data volume required division into thematic segments.
- Each segment had its own Excel application connected to the database via ODBC.
- Report files reached sizes of several gigabytes.
- Excel memory limitations determined the mode of work and data organization.
- Performance optimization by splitting data and minimizing redundant operations.
