Imagine you are a customer success manager. You have customer comments in hundreds of NPS responses, customer interactions & your own meeting notes sitting in archives. You want to get signs of action out of this untouched territory of raw information but it's almost impractical to go through and extract the actions by yourself.

Usually customers or product users provide signals of their satisfaction, discomfort or sentiment at various stages of the interaction. It could be in comments of support tickets that they are raising, long email chains that might have happened over time, part of nps survey responses or even in casual catch up. Along with these sources of customer signals there could be various dimensions which might be impacting the actual quality of thier experience. For example your support team, feature robustness, billing, renewal process, product intuitiveness and many more all have an impact the the resulting experience of the customer.

Now, you can imagine how complex this whole network could be in terms of even extracting or tracking the most basic insights. Such as who is at risk and needs the most attention out of all of my customers. We’re far from getting some automated guided actions from these sources with today's approaches. Which is why today customer success done right has a prerequisite need of data wrangling and business intelligence efforts. First you must gather the data then you must clean it, then you can begin to use it.

But also keep in mind, the majority of these inputs today are not used at scale as they are text based interactions. (Meeting Notes, NPS responses, Email Chains). The are used subjectively on individual levels but are hard to extract value at macro levels to distill account level and cohort level insights. This is where Natural Language Processing comes in.

"Customers or product users provide signals of their satisfaction, discomfort or their sentiment via various channels of engagement."

Natural Language Processing is among the hottest topics in the field of modern business intelligence. Companies are putting tons of money into research in this field. Customer Success as an industry is just one of the many. Everyone is trying to understand Natural Language Processing and its applications to make their business insightful. Every business out there is integrating it into their business over the next 5 years as most SaaS tools have some R&D that has been brewing over the last 2 -3 years that is begining to be brought to life in their platforms. To understand the power of natural language processing and its impact on our lives such as the customer success manager above, we need to take a look at its applications. These are few applications of natural language processing:

Applications of NLP

• Targeted Advertising
• Voice Assistants
• Search Autocorrect and Autocomplete
• Hiring and Recruitment
• Social Media Monitoring
• Chatbots
• Survey Analysis
• & More

Forrester reports that 74% of companies say they want to, only 29% are actually successful in actioning their analytics

Though there are tools which will help CSMs to build scorecards and send notifications; a textual actionable assistant is vastly missing in the customer success domain. Most progressive companies today say they want to be data-driven. But as per findings of Forrester reports that 74% of companies say they want to, only 29% are actually successful in actioning their analytics. So, the missing link here seems to be actionable insights for companies that want to drive business outcomes from their data.

"A textual assistant that provides actionable insights is vastly missing in the customer success domain."

One key for customer success leaders looking to leverage NLP is understanding how it works when applied to customer success problems. Customer Success data has alot of challenges that must be uniquely addressed. Although tools like Medallia exist to perform text analytics across a spectrum of general use cases. The need for insights to be specific and actionable for CSM's makes the general application of NLP lacking. Many times inferring the actual tone or theme of conversations becomes tricky. Especially if such topics are subjective in nature. For example, ease of use, product suggestions and negative survey sentaments all usually consist of words and phrases that have diverse potential connotations based on their use in context. Other issues are based on grammer itself. Grammatical properties of words plays a vital role in making them relevant in any conversations. All of these are common challenges in applied NLP but these are even more challenging when used to solve Customer Success problems.

To cover such scenarios we have supervised models which evolve over the period of time but have been pre-trained on the core principles of customer success. We have had text assistants and chatbots since the 2000's but the challenge is still the same. Builiding compelling customer success experiences has a higher bar to exceed. So, here comes DataPlant’s NLP engine to help CSMs and Execs take actions based on these hidden indicators of customer health.

How NLP in Customer Success Works

Customer Success by nature is multidimensional. In Gainsight, they call the concept "Getting a 360 view of your customer". Meaning customer success is rooted in having visibility of your customers experience across all the departments in your organization. For that reason NLP must collect text from all the available sources (e.g. nps, email, support comments etc.) and parse them. Data preporation alone is a complex task but only once that is done can you begin to pass cleaned text for training. During training your algorithum would need to utilize statistical as well as graphical methods to get insights. This is why we have used our unique experience as a team to build something powerful yet purposely geared for customer success.

DataPlant’s centralized Knowledge Base is trained specifically for the customer success domain

We have solved all of the challenges that usually make NLP unactionable for CSM's before in practice for different companies. Now we have begun to productize those learnings. By building a centralized knowledge-base which is trained specifically for the high actionablity benchmark in customer success; we had some universal truths along the way that we're am excited to share with you over the coming weeks. Our centralized knowledge base approach understands the relevant terms and the processes of the customer success domain and accordingly it maps or highlights important insights with context. This approach itself has novel takeways in linguistics, analytics and customer success. Tons of stuff we will dive into together.

To learn more about how we are building for the future of customer success and see how we have broken down the challenges of applied NLP in customer success. Check back for Pt. 2 of this series. We will have a ton of gems that any organization can apply. In the meantime, please reach out and let us know what you think of NLP. Is it the future of customer success? Will AI take all our jobs? lol. Share and like below. Until next time, readers.