Almost all customer feedback systems include surveys with open-ended questions. Indeed, in the popular Net Promoter System (NPS), the survey can consist entirely of just two questions: a likelihood-to-recommend score, followed by an open-ended question explaining the score. As a result, as much as half of the data generated by customer feedback systems is verbatim, “unstructured” text.
One of the problems with unstructured text is that it can be very difficult to identify patterns or trends in the data. Are my customers complaining about customer service? Excited about new cable TV channel lineups? Experiencing dropped mobile coverage?
Until now, the only way to identify patterns or trends in text data has been to simply count the words in the comments and try to infer meaning from which words are used most commonly. These “bag-of-words” techniques infer meaning indirectly, by analyzing lists of words. In order to parse meaning directly, models need to approach the unstructured text differently.
In contrast to traditional models such as bag-of-words, recent advances in Artificial Intelligence (AI) have enabled machines to replicate or exceed human performance in many domains, from recognizing faces, to playing games, and now, to reading and categorizing survey feedback.
Today, Centriam is announcing Centriam Telecom AI, an AI bot trained to read and categorize customer satisfaction and feedback comments in the Telecommunications industry. Centriam Telecom AI is the first commercially available solution built specifically for telecoms, internet, and tv service providers to parse customer comments directly for true meaning.
Centriam Telecom AI quickly reads customer satisfaction survey comments and automatically categorizes each comment for data aggregation and ongoing trending. The Centriam Telecom AI was built for three particular use cases:
- Periodic, relationship-based customer satisfaction or NPS surveys (including transactional NPS)
- Satisfaction or NPS surveys related to recent in-home repairs
- Satisfaction or NPS surveys related to recent in-home installations
For example, the team in charge of the Customer Experience of the install process of home phone, cable, and internet needs to gather information on customers’ satisfaction with different aspects of the install process, from the quality of the installation, to the demeanor of the technician, to the time it took to get the install completed. The organization collects thousands of survey comments about the install process but doesn’t have the time to read and categorize the comments. With Centriam Telecom AI, the team can process the thousands of comments into meaningful categories and sentiments in minutes.
Here are a few examples of what Centriam Telecom AI can provide:
|Comment||Category & (Sentiment)|
|The install was good and the technician was professional||Technician Demeanor (+)
Technician Work Quality (+)
|They should always keep the appointment dates||Wait Time (-)|
|My channels aren’t working and yet I still have to pay||Billing (-)
Install Effectiveness (-)
|The installation was fast||Wait Time (+)|
|I got so many channels, but I’m still not getting ESPN||TV Content (-)|
It’s important to reiterate that the AI reads the comments for their meaning rather than just the words in the comments. It identifies both “They should always keep the appointment dates” and “The installation was fast” as both being about Wait Time, even though neither comment includes the word “wait” or “time.” That’s because, unlike other Natural Language Processing such as bag-of-words, Centriam Telecom AI reads for meaning rather than lists of words.
Once you send your comments through Centriam Telecom AI, you can track the effectiveness of your Customer Experience over time and across dimensions. Are your technicians in Atlanta friendlier than your technicians in Boston? Are there TV channels your customers are starting to ask about that you should consider adding to more bundles? These types of questions can now be answered with Centriam Telecom AI. Connecting these insights with customer data (tenure, number of accounts, monthly revenue, etc.) will enable you to drive action to influence your bottom line.