I have been working on causal inference and experimentation frameworks within the e-commerce tech industry lately, specifically dealing with platform behavioral and transaction telemetry data to determine how customers behaves on those e-commerce platforms.

Lately i have been facing problems where standard textbook statistical theory fails when dealing with messy, real-world data distributions as human behaviour is unpredictable and many unexpected things, which i didnt even thought could happen, can happen in the real world. Such as product delivery delay, causing the customers to be angry and hence put them off from buying the product, each of these variables have to be accounted for.

To model and have these unexpected psychological state among customers accounted for, we use variables such as sudden spikes in customer support chat volume, rapid page refreshes, or immediate session drops.

Also, Marketing headers and automated CRM systems love to push blanket promotional campaigns, but as statisticians,…