The second challenge of CX Garage commenced on 7th October 2019 and focused on extracting valuable information from unstructured text using content comprehension. This is a domain in the area of machine learning comprehension and is required to innovate business solutions in the CX product portfolio to address customer needs. The challenge was sponsored by the SAP CX Innovation Office and led by Jaspreet Kaur in Bengaluru, India. The team located, in the Startup Studio space of the Bangalore campus, had participants coming in from different lines of business. Our 4 participants were Pratik (C4C), Chetan (C4C – Service), Gagan K (S/4 HANA), and Abhivyakt (S/4 HANA).
Challenge – “Content Comprehension in CX”
Objectives:
- Research the market of content comprehension.
- Identify data resources of unstructured content within the SAP C/4HANA space and create a compelling user story supporting the mentioned use case.
- Implement a PoC following the same using open source framework tools and identified data resource and integrate this into the SAP CX Portfolio.
- Document and share learnings and showcase it to the respective product management teams for productization.
The aim of the challenge was to utilize the unstructured data in CX suite and develop insightful tools for different personas, using machine comprehension. CX suite has lots of unstructured data in each of its cloud solutions. Implementing machine comprehension models to analyze such textual data and to provide valuable business insights, which would enhance the impact of the existing product, was where this challenge was focused. Google’s BERT, being the game changer in Natural Language Processing (NLP), was also recommended to be used while designing the solution.
With the scope being so broad and the challenge timeboxed for 16 weeks, the team had to look across all the five cloud offerings in SAP C/4HANA, converse with different Product Managers, and brainstorm multiple use cases where machine comprehension could be introduced. But the challenge (along with the finalization of the use case) was to find the right data needed for its implementation.
With respect to the mentioned objective, the CX Garage team identified two areas where content comprehension could be applied. First, the team started with analyzing service tickets and the potential to increase efficiency by providing an excerpt to the service agent. The team then focused on Commerce Cloud, researching how existing user reviews can be utilized to improve product descriptions. The following two use cases were finalized:
- SummUp – SAP Service Cloud’s Ticket Summarization
- ABSA – Aspect-based Sentiment Analysis for SAP Commerce Cloud Reviews
The key deliverables included:
- User story
- Working prototype – for the above-mentioned use cases
- Documentation of learnings and implementation
Service Cloud’s Ticket Summarization
In SAP Service Cloud, incidents can be created from different channels (phone, email, social media, etc.). However, most of the incidents are created from email channels and will have lengthy and multiple conversations.
Usually there will be multiple conversations with the customer to identify the real problem and to get various other details.
As the length of the communication thread increases, service agents need to spend more time on reading all previous memos to understand the issue each time the ticket is assigned. Oftentimes, the interactions contain duplicate and redundant information. Going through each of the conversations becomes frustratingly difficult for the service agent.
Solution – “SummUp”
To address these challenges, the Innovation team came up with a solution named “SummUp”. SummUp will provide a full summary for each ticket interaction.
Use Case:

Click here to watch the demo of the solution integrated
into SAP Service Cloud.
The Service Ticket Summarization use case has received a very good feedback from the Product Managers and other colleagues from SAP Service Cloud. They are also planning to add this to the backlogs of Release 2005.
ABSA – Aspect-based Sentiment Analysis for SAP Commerce Cloud Reviews
E-commerce companies are always on the lookout to increase sales. Product reviews and feedbacks are of great significance, as their analysis can provide insights on market trends as well as customer demands. Understanding the customer mindset can also lead to better marketing campaigns. SAP Commerce Cloud caters to a wide range of products, and it is very difficult to manually analyze every product review. Also, in this competitive market, it is not sufficient to understand the sentiment for an entire product. Drilldown to aspect-level sentiment analysis can help to understand consumer acceptance of different features as well as shortcomings of a product. Such insights are invaluable and can help to both prevent loss of sales and to design optimized marketing campaigns.
Use Case 1:
Mark wants to buy a laptop which is lightweight, as he is a frequent traveler. But on the product description, there are no details about the weight and therefore the product gets ignored.

After adding this missing information to the description, new customers would be able to find more details about the “Weight” aspect, thereby preventing such loss of sales.
Use Case 2:
Sales expert Bruce wants to boost the sales of a particular laptop. Using ABSA, he can easily uncover that device storage has been the least-preferred aspect of the laptop, and this can be fixed by including additional storage.

Therefore, Bruce creates a new marketing campaign with a combo offer for the laptop along with external storage to increase the sales of the product.
Click here to watch the demo of ABSA.
The ABSA solution and use cases have been widely appreciated by various colleagues from SAP Commerce Cloud and SAP Service Cloud. We are currently in talks with Product Managers from the respective areas to look for a way forward for productization of the PoCs.
If you are interested to learn more about CX Garage or the Challenge 2 outcomes, please contact Jaspreet Kaur (jaspreet.kaur@sap.com).
Comments are closed.