Exploring eCommerce Travel Data

In this notebook, we will analyze an eCommerce dataset. Each row of data contains information about someone that visited the website and may or may not have spent some money on the selection of travel packages that this website hosts. Suppose we are a marketing agency and we are tasked with identifying cohorts of users to incentivize to visit the site again.

Smart Mapping - key drivers of 'Amount Spent'

Lets use VIP's Smart Mapping routine to identify the key drivers of users who spent money on the website.

Immediately we can start to see how users with heavy spending and lower spending are separated. Its also very interesting that 'Sources', which tracks how the user visited the website, is such a strong driver of spending. Lets map 'Sources' to shape and see if that shows us anything interesting!

Its pretty clear that the only 'Sources' value that had substantial sales was the 'App'. Good to know - this makes it easier for us to send notifications to the user about potential deals we can offer.

VIP's AI based Insights

We are specifically interested in finding users that spent a lot of money on the website. From the dropdown menu in VIP, lets click on the RED color to look at the top quartile of spenders.

Lets make an index column to track our known heavy spenders and check some simple statistics!

It looks like our "low income browsers" and "express shoppers" have nearly double the 'Amount Spent' on our products. Lets see if clustering will give us more information about our user base.

Behavioral Clustering

The red and blue clusters are separated on 'Time spent' and 'Products Viewed' but if we look closely we can see that the green cluster is strongly correlated with size, which currently is mapped with 'Household Income'. This is a really useful insight in that these individuals have high enough income that it outweighs their behavior in 'Time Spent' and 'Products Viewed'! We should try to identify our richest customers and send them customized marketing materials.

Highlight Browsers and Express Spenders

Identifying the richest customers

We can use VIP's Anomaly Detection tool to identify users with extremely high household income. We may be able to target these users with a special deal.

Taking action on key insights

Through our analysis, we've identified 3 cohorts of users that are likely to spend money on the website.

Lets make some subsets of the data based on the 3 cohorts we've identified.

We will not run the next cell here, but it is shown to give inspiration on how business decision can be taken from the insights identified in this analysis.