Computers never have a problem understanding structured datasets, but for unstructured data like free text, it is always a pain. With AI and machine learning technologies, we are now able to tap into the richness of information stored in the unstructured datasets.
For instance, service centers of big companies usually receive thousands of support tickets per day. Even though there’s a lot of valuable information captured in the ticket content that could potentially be used to know better about some customers or improve the product/service itself, it’s impossible for a human being to go through all these tickets just to generate insights with statistics.
That’s where the power of Natural Language Processing (NLP) text clustering comes in!
Envisioning the high potential, Service Ticket Intelligence (STI) – part of the SAP AI Business Services portfolio, recently had a GA release (2002a) introducing a new machine learning-based feature called Ticket Clustering. This new capability is aiming to generate insights from the content of voluminous tickets received by service centers, and generate actionable business value out of these insights.
In this blog post, I’d like to show you how we could get insights from unstructured text, by following a real business use case.
Use Case Example
Company A has a big customer X that generates over 2300 tickets received from January 2019 to February 2020. The account team of customer X would like to get an analysis of the tickets to better understand the customer feedback and challenges encountered, so they can plan what and how to pitch to them with new offerings.
These tickets get uploaded to Service Ticket Intelligence using the Ticket Clustering feature and 100+ clusters are generated with a list of keywords attached for each cluster to show their meaning. The team arrange the clusters in sequence with the descending number of tickets, and realize that the top 5 clusters account for over 54.07% of total hours for resolving all tickets, and over 53.10% of overall tickets counts.
With that, the account team focus on these 5 clusters and start to take a closer look based on the keywords provided and sample tickets that fall under each.
Here are the two main insights pointed out after the analysis:
- Insight 1: Cluster 5 (contains 73 tickets with 249 hours spent to handle the tickets) is about access (role) request and password tools.
- Insight 2: There are two main descriptions identified among the top 5 clusters, mostly referring to manual monitoring, which can be technical, MOCC (Managed Operations Control Center technical alert handling), and APO.
Two follow-up actions have been identified by the team in charge:
- Action 1: The service center manager proposes to create a self-service/automated access (role) request and password tools, which will potentially have the following benefits:
– Up to ~73 total tickets per year can be reduced
– Up to ~249.41 total hours for ticket handling can be saved
- Action 2: The account team understands customer’s challenges on manual monitoring, and comes up with a sales play around it to better tackle customer needs.
To sum up, the clusters formed out of these 2300 tickets by the Ticket Clustering feature of Service Ticket Intelligence have been used to generate concrete action plans and achieve tangible business benefits.
There are many more use cases where the Ticket Clustering feature can be leveraged in enterprise applications to create business value, and ultimately improve the customer experience.
Stay tuned with Service Ticket Intelligence and SAP AI Business Service by following these useful links below, and let us know if you have any comments in the section under the blog post!