Allegro's New AI Suggests Items That Actually Go Together

When you buy a camera online, you might see suggestions for a tripod, a lens, or maybe another camera. That last one doesn't make sense—why would you need two cameras? But many recommendation systems can't tell the difference between products people happen to buy together and products that truly fit together. Polish marketplace Allegro built a new AI called AlleCompanion to solve exactly that problem.

The challenge is that shopping data is noisy. If someone buys a camera and a dog leash in the same order, the system shouldn't assume those belong together. Allegro's solution combines smart data filtering with a neural network that learns which categories of products naturally complement each other. They also added what they call ComCat—a way to merge expert rules, human feedback, and AI reasoning from large language models. Think of it as a teacher guiding the AI to pair a camera with a lens, not with a leash.

This system is already running at massive scale, serving more than 20 million active shoppers every month. The results are measured in real money: Allegro reports significant increases in sales driven by these smarter recommendations, both in regular browsing and in paid product placements. For shoppers, it means less clutter and more useful suggestions—seeing the right accessory before you even realize you need it.

The broader takeaway is that recommendation technology is becoming less about "what everyone bought" and more about "what makes sense for you." As this approach spreads, expect online shopping to feel less like a crowded bazaar and more like a helpful store associate who actually understands what you're doing.