Six consumer data models, fitted and run live in your browser. The data is synthetic with known structure planted in it, so every page can show its receipts: press replay, drag any slider, and watch every number recompute. These pages prove the method, not your results; those take your data. Each page switches between English and 中文 in the top corner.
Learns conversion patterns from member behavior and scores the whole base, so a capped campaign budget reaches the likeliest converters first.
The campaign budget reaches twenty percent of the base. How is that list ranked today?
Learns customer groups straight from behavior, no labels given, and shows the evidence for each split.
Are your customer segments drawn by the business, or learned from the data? And when were they last refreshed?
Reads each customer's churn risk and future value from purchase timing, so retention budget ranks on what comes next.
Whose retention gets the budget first? And how is that list ranked today?
Derives product affinities from purchase histories and ranks a personal next-best list for every customer.
On your mini-program home page, who decides what each customer sees first?
Estimates each channel's sales contribution and its diminishing returns, so the next budget move has numbers behind it.
When was the channel budget split last changed, and on what evidence?
Separates trend, season, and campaign lift in the history, then forecasts with a backtested error band for target setting and staffing.
How was last year's peak-season volume target set? And how far off did it land?