Jev 给 50 万条电商会话数据打行为标签

Andrew Blumson @Andrew_Blumson原推发布于 2026.10.04

在 X 上打开原推

@Replit x Jev x Atta = video attached. I don't just build games 😉 The demo build is to show how a few different pieces can work together on a large dataset. I created a fictitious e-commerce company with 500,000 customer sessions. The data includes things like country, age, device, traffic source, products viewed, cart activity, checkout behaviour, purchases, returns, delivery cost and customer comments. The first stage uses normal filtering for the obvious stuff. For example: Show me UK customers aged 18–24 who came from TikTok, looked at running shoes, but didn’t buy. That part doesn’t need AI. Then Jev comes in. Jev looks at the customer behaviour and makes fast decisions around things that aren’t directly stored in the data. Things like: - Purchase intent - Likely reason for abandonment - Price sensitivity - Whether they should be retargeted - Return risk So instead of just having 500,000 rows of raw data, the dataset becomes much more useful. Then that enriched data is analysed and visualised using Replit’s new Atta technology. From there you can ask questions like: “Why are UK TikTok users abandoning running shoes?” “Compare them with German customers.” “Show me high-intent customers who didn’t buy.” “Which traffic source has the biggest opportunity?” And the data can be turned into charts, heatmaps and visual analysis. So the full flow is basically: 500,000 raw customer sessions → normal data filtering → Jev makes AI decisions → the data is enriched → Atta visualises and helps analyse it → Replit can then turn the insight into an application or action What I like about this is that it shows AI doing something beyond just chat or content generation. It’s using AI to make decisions across a large dataset, combining that with normal data processing, then making the results easy for a human to understand. Demo video attached. Note: the Charts are from a different download as I didn't want to record my screen while I downloaded and uploaded the CSV, so the Charts are from a previous run that I did earlier. #Replit #Atta #Jev #Data #Analysis