Try MBA against a fake-store dataset.
Pick a product
We'll show the top complementary products MBA would recommend, with the same scoring (support, confidence, lift) the real engine surfaces in the admin.
Top recommendations
Customers who bought the Backcountry 2-Person Tent also bought:
~60 seconds per job
On a store up to 100k orders, the full mining pipeline completes in under a minute. This demo serves the result from a pre-computed cache, the real product reads from your live order history.
Profit-aware ranking
The same engine produces revenue-weighted rules with the HUI algorithm (Pro tier). Cross-sell suggestions weighted by margin, not just click-through rate.
Same data, MCP-accessible
The recommendations you see here are exactly what an AI agent (Claude Desktop, Cursor, Cline) gets from the MBA MCP server. Basket AI: four named agents over eight tools, real auth, SSRF-guarded.
The math, briefly
Support
What fraction of orders contain BOTH products. High support means the pairing is common in your real history.
Confidence
Given a customer bought product A, what fraction also bought B. 80% confidence = 4 in 5 customers who bought A also bought B.
Lift
How much more likely the pair is than chance. 1× = no signal (random co-purchase rate). 3-5× = strong signal, this pairing is genuinely connected, not a coincidence.
The recommendations on this page are ranked by an opportunity score (lift × confidence), which weights both how genuine the pairing is AND how predictable it is. The real engine adds a fourth column for profit lift when you opt into cost-aware scoring (Pro tier).
Ready to turn your order data into revenue?
Install on your platform in under 10 minutes. Or book a consulting call and we'll do the launch for you.