Hand-selected AI models. Your existing data. Prompts for a set of problem themes. Fixes on your existing tech stack.

Arrow pointing right

The answers are there, and now you can actually find them.

Step 1: You share signals.

You share your Google or Adobe Analytics clickstream feed with us. Everybody has this, it usually does not contain PII, and it takes 5 minutes to export.

Step 2: We validate signals.

Sometimes, your site tagging does not accurately capture user behavior. We “checksum” your signal data to make sure that it reflects actual user behavior. For some companies, this step alone is the most valuable part of our engagement.

Step 3: Cleanup and Perfunctory Analysis

Clickstream files are huge. This is what makes them such important troves of behavior data. That said, not everything is valuable and the file can be drastically reduced. We take out everything the models don’t need and produce an initial set of insights.

Step 4: Workshopping

Unlike most AI-focused companies who believe site merchandising should be fully automated, we know that buyers and merchandisers understand products and customers better than AI. We sit down and discuss our initial findings and explore ideas for optimization together.

Step 5: Insight Iterations

Now the good stuff! We take hand-selected models trained to find a discrete (and growing) set of problem/opportunity types. We produce daily insights and, if you like, manually add them to your existing product discovery, content management, merchandising, or other appropriate platform.