Travel Website - CompanyName
(Note: Name is Confidential)==========================================
Bhuvnesh Singh and Varun Gupta - INSEAD MBA 16J
bhuvnesh.singh@insead.edu varun.gupta@insead.edu
Business motivation
Our client CompanyName Hotels & Resorts is a hotel chain based in Country, specializing in 4 and 5 star hotels. It is owned by Group CompanyName, which also owns CompanyName Golf, Villas & Condos, CompanyNameservice its incoming tourism division and CompanyName The Club.
The client has its own website and receives traffic from several countries. In this case, we have looked into the traffic data of the client and have tried to look for bottled down factors that drive the booking and hence the revenue for the client. Further, we have looked at the possible clusters in the traffic data.
Data Description
First, let’s take a look at the first 1000 rows of data
Let’s review the contents of the data
- Type of User: Returning or New Visitor
- Source: Organic vs Paid Promotions
- Users: Number of User Visits
- Sessions: Number of User Sessions
- No of Pages Visited: Number of total pages visited
- No of Transactions: Number of monetary transactions
- Revenue: Revenue generated
- Dummy Variable: We created dummy variables categorical data such as ‘user type’, ‘source’ and ‘device type’
This is the correlation matrix of the various factors:
Here we use PCA to identify primary components:
Let’s see how the top factors look like:
Based on the PCA, we can potentially reduce the number of attributes from 17 to 8, and make the decision making process relatively easier. The 8 new buckets could be defined as:
User Engagement (Sessions, Number of Pages Visited, Revenue, Users, Number of Sessions)
Kind of User (New, Returning)
Device (Mobile, Desktop, Tablet)
Organic Source (Direct, Referrals)
DFA Promotions (DFA CPM)
Google Promotions (Google CPC)
Email (Email, Newsletter)
However, from a business sense, bucketing these attribute does not make much sense hence we do not recommend dimension reduction in this case.
Now, we look into the revenue and transactions data through device segments, and try to find out simple patterns. let’s look at how Revenue/Transaction and Transactions varry for the type of device: Desktop, Mobile and Tablet
It is clear that out of all the device types , tablet is one of the most profitable for the client. Not only the average revenue per ransaction is highest for the tablet, but among mobile and tablet tablet outperforms mobile by garnering similar transaction numbers for less than 20% of the page views. We recommends that the client further studies the source segmentation of tablet users and spends more marketing budget on acquiring tablet users. On the other hand form product perspective the client should look into the reasons behind very high number of mobile page views with marginal returns - perhaps there is a lot of scope for improvement by making the mobile experience simpler and easy to use.