Cross-Platform Product Monitoring

Seller teams at Snapdeal needed to know how their products were priced and positioned on competing platforms. Manual checks don’t scale across millions of SKUs. This framework automates cross-platform product matching using image recognition, replacing hours of manual competitive analysis with a continuous monitoring pipeline.

Problem

Product titles and descriptions vary wildly across platforms — the same shoe might be listed as “Men’s Running Shoe Black” on one site and “Sports Sneakers for Men - Noir” on another. Text-based matching produces low recall. Image matching is more robust because product photos are often identical or near-identical across platforms.

Architecture

Image Feature Extraction

Product images are passed through a pre-trained CNN to produce compact feature vectors. These embeddings capture visual appearance — color, shape, pattern, brand logos — independent of how the product is described in text.

Cross-Platform Matching

Feature vectors from Snapdeal products are compared against embeddings from competitor catalogs using approximate nearest neighbor search. Matches above a similarity threshold are flagged as the same product. A verification step filters false positives by cross-checking category and price range.

Gap Detection

Matched products are compared on price, availability, and assortment depth. The system surfaces:

• Pricing gaps: Products priced significantly higher or lower than competitors
• Assortment gaps: Products available on competitor platforms but missing from Snapdeal
• Availability gaps: Products out of stock on Snapdeal but in stock elsewhere

Results

Metric	Value
Cross-platform match coverage	85%
False positive rate	< 5%
Manual analysis replaced	Fully automated

The framework runs on a scheduled pipeline, producing daily reports for seller teams. It replaced a workflow that previously required analysts to manually check competitor listings — a process that could only cover a fraction of the catalog.