Many of HPE's customers do not benefit from the content offered on the HPE website due to the large amount of information available. For this reason, to give a useful solution to users, Machine Learning tools in the form of recommenders were applied. When a user lands on the HPE website, they will find a wealth of useful tools and information related to its products. In other words, a platform with an infinite number of resources to find the desired content.

However, this strength is also an obstacle: locating a specific asset among such a volume of information was a difficult task for the user. 

HPE was finding that this amount of data was overwhelming visitors, leading to frustration and abandonment. This problem leads to waste and disuse of the knowledge that HPE offers. 

Generating engagement on the platform

This is the goal of many online sites: to generate engagement. In the case of HPE, this purpose becomes even more important as a result of the following findings:

​​According to internal reports from the Customer Relationship Department, visitors are ready for a business contact (nurture) when they have checked up to 5 assets. 

However, most visitors did not go through such a high number of assets and they left the session early. What was the result? Low user engagement as a result of the overwhelming amount of information.

For this reason, generating engagement on the platform was essential to achieve the business objectives. The challenge: find a way to attract users by encouraging them to navigate from one document to another.

Betting on a recommendation system

Navigability and UX are two key requirements to be taken into account in web design and HPE's was not going to be less. The User Experience was improved by redesigning the asset directory on a new platform. It now displayed useful information powered by Machine Learning tools from Google Cloud Platform.

This Machine Learning solution included two recommendation engines:

• The recommendation system is based on the content the user is currently viewing. 
• And the second recommender is trained based on the user's behavioral data. That is, based on the content that the user has visited before.

This recommendation system aims to increase the number of documents viewed per person, pushing the user to consume a greater number of documents in as few visits as possible.

Average of 6.71 documents per user of the recommender

That's right, the recommender made it possible for the user to consume a greater number of documents in the fewest number of visits, thus increasing the number of documents viewed per user.

After implementing the recommendation engine, we found that, after its implementation, these are the results:

• An average of 6.71 documents was viewed per recommender user, compared to 1.76 viewed by other users.
• 46.74% lower bounce rate for referrer users.  
• 07:16 minutes more session time per referrer user. 
•  More than 500 million registrations in total.
• An average of 10 active users daily 
• 5 new users every day.
• More than 180 GB of data in total.

This data shows that the implementation of the recommender is successfully pushing users down the conversion funnel from nurture to lead.