Finding The Buyer Among Visitors

The Case:

Being able to accurately predict which visitors will turn into buyers is a crucial part of any web business. Knowing who your buyers are can not only help you predict future sales and improve your business planning but it can also help you better serve the needs of your customers.

The Challenge:

Customer segmentation is an old problem which boils down to answering two key questions: Who are my buyers and what do they value the most in my products/services?


Getting answers to these questions is hard enough when you have data on your customers such as purchasing history, age, gender, etc. However, most small web stores don’t have access to such data which makes it hard for them to use traditional models.


Often times the only data that these stores can get is the browsing history of their customers.

The Solution:

To solve this problem we’ve extracted over 30 features from the visitor’s browsing history such as frequency of visits, time between visits, Page rank of the referrer site and more.


We then fed these features into a custom machine learning algorithm to produce a model capable of identifying the visitors with the highest probability to purchase within the next week as well as outlining the reasons for their purchase

The Outcome:

The final model was tested in a marketing campaign where it managed to identify the most promising future buyers with an accuracy of over 90%.