The challenge
Conventional e-commerce personalisation reacts after the visitor has already disengaged, and demographic data does not explain purchase motivation. The retailer was allocating discount and coupon spend broadly rather than to the visitors whose decision was actually still in play, giving margin away to buyers who would have converted anyway.
At a glance
DataClickstream, pre-checkout segment only
MethodBehavioral time-series analysis, anomaly detection, intent prediction
PrivacyNo personal data
What we did
SWICON built a behavioral model focused deliberately on the pre-checkout stage of the customer journey. Behavioral patterns change drastically once the payment button is pressed, so a model trained across the whole journey dilutes exactly the signal that can still be acted on. The model works on clickstream time series, predicting purchase intent and the remaining time and clicks to the decision point, which makes a timed intervention possible instead of a generic pop-up. No personal data is required.
The result
Conversion rate increased by 8–12%. Coupon cost fell by 50–60%, because discounts were allocated only to visitors whose decision was still open instead of being distributed blanket-style. One intervention, both a revenue uplift and a cost reduction.
