This Trick Will Help You Uncover Trends From Your Support Data

By now, most companies have set up some sort of customer self-service tool. Whether it is a chatbot, a help site, or simply a search engine, most businesses have seen the value in letting consumers navigate their site, and find answers and products on their own. 

It is clear these tools can satisfy the need for immediate support, as well as reduce the workload of your customer service agents, improve CSAT, and simplify your operations.

However, is there something more self-service tools can do for you?

A new source of customer data

Digital support channels, as the ones we just mentioned, have a substantial advantage when compared to phone or in-person assistance: they leave a trace.

They allow us to log all activities and generate a large amount of data from which we can draw conclusions and identify patterns and trends.

Sure, you could also do that with in-person assistance, but you’d need an employee to write down the reason for contacting the support, the behavior of the customer, what happened after the request was closed, how happy the customer was with the assistance received… 

Doing this manually would be extremely time-consuming. 

Yet, most digital tools let you do that naturally, automatically and with practically no effort from your side. 

However, simply gathering data has no benefits whatsoever. The question remains:

  • How do we read that data?
  • How do we understand it?
  • What can it tell us and what’s its benefit? 
  • How can we drive change from it?

What does the support data tell us?

That is a tricky question. It really and truly depends on the support tool you’re using, the data that it is gathering for you and your own business challenges. 

Self-service support tools tend to come with metrics and KPI dashboards to understand what’s going on. These may include:

  • Number of calls/chatbot sessions/searches
  • % of satisfaction regarding the answers you provide
  • Number of unanswered questions
  • And a long etcetera.

Nevertheless, most tools tend to show only and understand a small part of the data they gather, therefore not taking advantage of its full potential. 

Read also: We analyzed 4 million chatbot conversations. Here’s what we found out

The long-tail of your support requests

Let’s take an AI chatbot as an example.

Do we know the intention behind most requests? 

Are there any trends among these queries?

What are customers most interested in?

Most chatbot platforms provide data dashboards showing the top 10, 50 or 100 requests asked by your customers, but what about the rest of them? 

The top 10 or 50 requests received by any support tool tend to be short—they’ll usually have no more than 2 words. An example could be a ‘cancel account’ request. 

However, these common requests only make up 20% of the total queries. The remaining 80% will be less common, and also longer (4 words or more). These longer requests are what’s known as long-tail queries’. 

Identifying trends among the top 20% is usually easier. Queries have one or two words, so it’s easy to see what they’re all about. However, what happens when users input long queries to express their intention? 

Let’s re-use the example of ‘cancel account’. What if the user says ‘I no longer want to use this service’ instead?

The intention might be exactly the same, yet, the request doesn’t include the same words, and it’s going to be difficult for your team to identify it as the same type of request on a large scale.

Most self-service tools and Conversational AI platforms don’t have a semantic engine powerful enough to understand natural language and real meaning. Therefore, they are unable to identify these requests as having the same intent. 

This creates a couple of issues:

  • The first issue is that most of the long-tail requests won’t be understood, and therefore, they’ll remain unanswered. If a user types ‘I am disappointed because the technician didn’t come and I no longer want to keep using your services’
  • The second issue is that, if the system doesn’t understand the meaning behind all requests, it can hardly group queries with the same intention and identify trends.

At Inbenta, we use a unique kind of AI that truly understands queries by looking at its meaning. Plus, we created the perfect environment for you and your team to understand customer requests at scale.

Read also: The ultimate guide to Conversational AI

Automatically identifying support trends with semantic clustering

As we’ve seen, most Conversational AI and self-service support software will only show you the top 100 requests. 

What if we said there is a way to analyze longer requests, together with the shorter ones, and group them by meaning?

Inbenta’s Semantic Clustering groups semantically equivalent search queries — words, phrases and sentences — into clusters based on meaning. 

A visualization tool allows you to see the different subjects/topics from the grouped queries.

Below you can see an example from our a small sample of data, at Inbenta:

  1. In this first screen we see the main topics from all queries (long or short). Inbenta analyzes not the top 10, 50 or 100, but the top 20,000 queries.
  1. We can dig deeper and deeper if needed. For instance, one of the main subjects is ‘chatbot’. If we want to see what kind of queries our users make regarding chatbots, we can just click on the ‘chatbot’ circle and we see further information:
  1. If we look closer to the circle at the center, we can see relevant metrics, like the number of times it was asked, how many times the chatbot found an answer to the query, etc.

How does that benefit my support strategy?

Analyzing customer interactions can bring a lot of benefits to your business. However, the most important ones we’d like to highlight are the following:

  • Identify trends. By easily analyzing the content with semantic clusters you can identify trends and requests from your customers, no matter how they express themselves. 
  • Identify content gaps and unanswered topics. You can filter the clusters to only show topics with no answer, or simply analyze the answer rate for a cluster in order to identify where you’re not giving any answers and what content you should focus on next to maximize self-service.
  • Understand what matters most to customers at scale. Seeing how many times users searched a specific topic will give you a better view of hot subjects for your customers.

Get started with conversational and chatbot analysis

Try Inbenta’s platform for free for 14 days and get access to our chatbot module, where you can navigate its semantic clustering.

Check out our similar articles