Sitecore Cortex: How Smart is it Really?

Ghodrat AshourniaJanuary 2020

Seamless And Personalized Digital Experiences

Sitecore’s flagship digital experience platform is used by many companies to create seamless, personalized digital experiences. At its center is the Sitecore Experience Platform. Sitecore introduced Cortex in the most recent version.

Three Use Cases for Sitecore Cortex

There are three different use cases for Cortex:

  • Use as a Sitecore-internal prediction engine

  • Integration with external services

  • Integration with machine learning

Cortex can be used as an internal prediction engine for the Sitecore Experience Platform. Take the example of personalization suggestions. Sitecore selects the user experience that achieves the best overall result, in order to optimize the experience for different user segments.

Easy integration enables Cortex to use external providers for data processing and analysis. One example is content tagging, which is available out-of-the-box with Sitecore 9.

Sitecore typically uses Cortex to communicate with Open Calais. This means that every product can be automatically tagged in Sitecore at the touch of a button. The most powerful approach is to integrate Cortex with a machine learning server. This gives companies full control over how their data is analyzed.

Machine Learning Is More Than Just Technology

But technology is not the be-all and end-all. To be able to exploit the opportunities machine learning offers, you need an interdisciplinary team with a range of skills:

  • Digital marketer: The digital marketer is in charge of marketing goals and target group segments and understands every aspect of customer behavior. They can identify patterns and develop use cases.

  • Data scientist: The data scientist is an expert in machine learning, is able to analyze data and patterns and select the right algorithm for the machine learning service.

  • Sitecore expert: The Sitecore expert is in charge of implementing the integration steps between the machine learning service and Sitecore.

Step 1: Setting a goal and collecting data

Once you’ve got the right experts on board, the first step is setting a goal. Imagine the following straightforward, fictitious use case. Let us assume that a tourism company wants to show users the destination that is best suited to them on its homepage as a way of increasing its revenue. Other use cases might be automatic customer segmentation for e-mail marketing or campaigns, churn prevention and product recommendations. The digital marketer, data scientist and Sitecore expert work together to establish what kind of data is needed and what user behavior should be recorded.

In our use case, they defined the following dataset:

  • Age

  • Gender

  • Marital status

  • Country

  • Past trips

  • History of visiting the website

This data is required to train machine learning. If data is incomplete or available on another system, such as a CRM system, Sitecore experts can import it into the Sitecore Experience Data Base (xDB).

The digital marketer can use the different data collection tools provided by Sitecore to configure Sitecore to collect the user behavior data needed for machine learning. In our case, the team set different goals based on the climate of the destinations and defined different events for requests for brochures or guides in order to establish what customers were interested in. In addition, results were assigned to the checkout pages to identify users who had actually generated a monetary value for the company.

Step 2: Exporting to the machine learning server and model training

If the appropriate data required is available, Sitecore experts can take the necessary steps to enable Cortex to export the data to the machine learning server and retrieve the predictions. During this step, the data scientist should develop the training algorithm and work together with the Sitecore expert to define the structure of the data model for the machine learning server.

Once training has come to an end and the predictive results from machine learning are at hand, the result is usually stored on a facet1 of a contact2. A new additional facet is created so that personalization can be applied based on this facet at a later stage.

Step 3: Personalizing content

We now have the user’s information and a prediction regarding the destination they will choose next. We can use this to personalize the teasers on the homepage, based on the current user’s predictive results.

Conclusion

Providing personalized content is an absolute must in the age of information overload. Machine learning is a broad field and opens up many avenues. According to estimates, Amazon’s recommendation engine generates more than a third of customer purchases using artificial intelligence to identify, analyze and provide the most suitable product recommendations.

Another example is the enormous investment IBM has made in artificial intelligence with its Watson platform for various areas such as travel or hospitality for hotels. However, the process of machine learning is complex. It requires experts with different skills to work together and high-quality infrastructure for data storage and data analysis.

Even beyond machine learning, Sitecore 9 has several attractive functions such as personalization suggestions or automatic content tagging. Clients looking for a rapid and reasonably priced solution can use the external services mentioned above to gain initial experience.

Sitecore AI automates personalization

The next AI service provided by Sitecore is already ready to go. At its symposium in Orlando, the company announced the arrival of “Sitecore AI” to address a recurring challenge in Sitecore projects. The experience platform is a powerful tool for tracking activities and providing personalized content. However, the company frequently finds it difficult to implement these functions. 

Sitecore AI is designed to automate personalization by automatically identifying visitor trends, creating customer segments and modifying page elements, offering a personalized experience as a result. Automated personalization has been available since December 2019.

More on intelligent user interfaces

intelligente-user-interfaces-mit-unic

Intelligent UI

This dossier shows how services become intelligent. It points out opinions on the potentials of artificial intelligence, but also how teams have to form in order to create intelligent services for customers. It examines various services with artificial intelligence and it analyses how much intelligence they really contain.

Read more
Nicole Buri1

Nicole Buri

Contact for your Digital Solution with Unic

Book an appointment

Are you keen too discuss your digital tasks with us? We would be happy to exchange ideas with you: Jörg Nölke and Gerrit Taaks (from left to right).

Gerrit Taaks

Contact for your Digital Solution

Book an appointment

Are you keen to talk about your next project? We will be happy exchange ideas with you: Melanie Klühe, Stefanie Berger, Stephan Handschin and Philippe Surber (clockwise).

Melanie Kluhe
Stefanie Berger
Philippe Surber
Stephan Handschin