ProductPulse

At Hammer, we often get questions about what is happening in a market. Which brands use a particular ingredient? How do their product ranges compare across countries? What do similar products cost on the shelf? The information is often available, but spread across many product pages. Reading all of these manually quickly becomes a project in itself.

I work on ProductPulse, a platform that brings this product data together and makes it useful for market research. It lets us explore products by category, brand, country and ingredient, and go from a broad question to the individual products behind the answer.

My work includes the infrastructure behind this collection: managing a large-scale database, building scrapers and developing the data pipelines that bring their output together. At this scale, keeping the data collection and processing reliable is a substantial part of the project.

Another part is making the data comparable. Retailers describe and categorise products differently, and ingredient lists come in different languages. A search that works for one country does not necessarily work for another. I build the processing and analysis behind the platform, using NLP and language models to help organise the data and answer more specific questions about products.

For example, an ingredient supplier might want to know which brands use a certain ingredient in a particular product category. ProductPulse helps narrow down the relevant products and compare the observed assortments. We can then inspect the product labels that support the result, rather than having to take a summary at face value.

What I like about this project is the connection between the technical work and the original question. Collecting more data is useful, but the interesting part is making it something a researcher can actually work with. The platform brings that collection, processing and analysis into the same workflow.