Multichannel e-commerce
E-commerce leaders need instant answers, but waiting on the data team for ad-hoc reports creates bottlenecks. This two-part BI module pairs an executive Tableau dashboard for daily monitoring with a generative AI chat that builds custom charts on the fly.
285M
raw behavioural events in the source dataset, filtered and sampled in Python into one analytical table
2
layers: an executive Tableau view and a plain-English chat for ad-hoc questions
0
direct access to the raw database for the AI, keeping data secure and token costs low
Multichannel e-commerce generates huge amounts of behavioural data across websites, apps and social platforms. Executives need that data to set strategy, but they shouldn’t need a data science degree to ask about top-selling brands or revenue trends.
The public source holds over 285 million raw behavioural events. In Python it was narrowed to one month (November 2019), sampled, cleaned and reshaped into a single analytical table that isolates purchase intent and revenue drivers, small enough for Tableau and the chat app to answer in seconds.
The processed data feeds a Tableau dashboard where leadership tracks revenue trends, channel comparisons and category breakdowns at a glance.
For questions the dashboard doesn’t cover, users ask a Streamlit app in plain English, and the AI writes the Python code that draws the requested chart.
The app queries a controlled dictionary version of the data, never the raw database. This cuts token costs, protects the data and shows the generated code for full transparency.
Leaders get structured reporting and on-demand conversational analytics in one place.
Portfolio project · Public dataset (Kaggle)