How to Do a Root Cause Analysis in Customer Service

Sharad Khandelwal
SentiSum CEO & Customer Service Expert
Understand your customer’s problems and get actionable insights
See pricing

Imagine you read a CSAT survey result that angrily calls out an agent for being unhelpful and slow.

You look closer at the corresponding conversation and you see the problem was the customer wants to update their password, but the agent failed to help.

You think "Great! This is easily solvable with some training."

You quickly kick off a new agent training around the topic. But negative CSAT scores keep coming. And, not just for the original agent, but all of them.

This issue is now driving 10% of your customer contact volume. You think "What the hell going on here?".

It's prime time for a root cause analysis.

You use our guide below to dig deeper and you find its not the agents fault at all. It's a bug in the product.

You let the product team know and they fix it in a matter of hours—saving costs, agent time, and improving your customer's experience.

Now, let's look a little closer at how to do that root cause analysis.

In this article:
  • Benefits of root cause analysis in customer service
  • How to Do a Root Cause Analysis of Your Support Tickets
  • Examples of a customer service root cause analysis

Benefits of root cause analysis in customer service

Getting to the root of reasons for contact in customer service is mutually beneficial for your team, your company and your customer.

We work with customer service teams of all sizes and company types, and their use cases for support conversation root cause analyses vary, too. 

Generally, they fall into three categories:

1. Improving the agent experience

By identifying the true root cause of support conversations and CSAT feedback, you’re able to better train agents and better equip them to quickly and diligently solve the issue.

For example, if you know that 5,000 tickets per month are about a specific, easily solvable issue, you can build a knowledge base article that your agents can send to the customer. A win for ticket resolution time.

CSAT feedback is also a valuable resource for agents. Root cause analysis of CSAT feedback, for example a topic analysis, reveals the true driver of the score and the associated sentiment. 

With our customers, we help them identify the behaviours and words that agents use which create negative sentiments, this tech-led QA technique helps each agent know where to develop their skill set.

2. Minimizing ticket volume & bringing down costs

Getting to the root of ‘reasons for contact’ is the first step towards tackling those reasons.

For example, in a Support Insights podcast episode with Lauren Cunningham, Loom’s customer support lead, we learned how Loom brought their contact rate down from 4% to 1%.

Over the pandemic, Loom’s customer base exploded from 700,000 to 10 million! At a contact rate of 4%, that meant an enormous surge in ticket volumes.

Starting with a root cause analysis of their customer conversations, the Loom team identified FAQS, things users commonly struggled with, and particularly time-consuming issues. That analysis fed into their ticket minimizing strategy, like their new knowledge-base which receives over 1 million in traffic each month—significantly reducing the need for customer’s to reach out to an agent.

Bringing down costs in this way is top of mind for many support teams. Gousto shared in this interview, that their real-time support analytics:

“Significantly reduces the lead time to unearthing customer insights and sharing these across the business, which allows us to positively improve [metrics like retention and reducing customer contacts].”

In a separate interview, Gousto’s Director of Support shared how root cause analyses in customer service also helps them identify the most frequent ‘reasons for giving a refund’ to their customers. By prioritising fixes to these issues, the company significantly reduces the financial impact of refunds.

3. Improving customer experience (CX)

Many of our customers invest in AI-support conversation analysis because they want to improve their CX.

Support conversations are a valuable source of customer feedback, to quote the UK government website:

“We believe user researchers should be as close to our user support tickets as our support teams and developers. They are a constant source of feedback and feature requests that, unprompted by us, reveal the issues that are most important to our users.”

This sentiment is echoed by Loom, Vinted, Gousto and many other customer service teams who now work diligently to feed real-time customer insights to other departments in their company.

How to Do a Root Cause Analysis of Your Support Tickets

Your root cause analysis can be as easy or complicated as you make it. I recommend you kick off with a simple project to ease everyone in and prove the value.

Here’s the process we guide our customers on:

1. Identify your goals

Starting with an idea of what you’re looking for saves a ton of time. For example, some goals you might be interested in are:

  • Identify frequently asked questions (FAQs): “Users always ask X.”
  • Find the most common issues: “What’s causing our top 3 drivers of customer contact?”
  • Find time-consuming issues: “Which tickets are taking the longest?”
  • Reduce costs: “What topics are driving refunds?”
  • Create growth: “Which areas of the buying journey are presenting a barrier to purchase? What topics are the cause of cancelled subscriptions?”
  • Improve CSAT: “What’s the most common topic in our lowest CSAT scores?”
  • Improve a particular product area: “How are customers reacting to our new feature release?”
  • Find agents who need training: “Why are X, Y, and Z agents receiving low CSAT scores?”

These will help accelerate your direction of travel and, importantly, tell you who to get involved in the planning process.

To make your analysis worth it, the results have to make an impact. The number rule of getting other teams to listen and act on your findings is to bring them in early and make them feel like their needs are being met (Ed Deason, Pret A Manger’s Ex-Director of Experience gave us this great advice in our guide to Customer Journey Mapping.)

The UK Government’s Product User Research team note in their planning stage that:

“To make sure our analysis focused on what was most useful, we initially ran workshops with each product teams to agree the objectives of this research.
In these sessions we printed off a random sample of tickets and gave each team member 5. Everyone was asked to write on Post-its what things they’d want to capture about the ticket, for example the type of request or the date it came in.
We then refined everyone’s Post-its and, by the end of the session, came up with a basic framework for analysis.”

2. Categorise support tickets

Anyone who works in the customer service industry will know the sheer volumes of conversations that happen every month. They’re not only high volume and frequency, they’re also made up of complex, qualitative free-text.

So, how do you approach the problem of reading them all and finding patterns?

Your first point of call is categorise all your conversations or CSAT feedback using topic tags. This will allow you to separate relevant tickets and track volume patterns.

You have two options here:

  • Manual categorization:

If you have only a handful of tickets and a good amount of time to invest, a manual analysis is your best bet. 

To identify the root cause of your support conversations, you first need to categorize them into topics using tagging.

Use our guide here to build a tagging taxonomy that suits your goals (we also tell you what a tagging taxonomy is, if you aren’t yet familiar with the term).

With your taxonomy in place, you must either go through your support request backlog and apply the tags or ask your agents to tag new conversations as they come in—which would only show topics in the latest tickets.

  • Automated categorization:

If you have thousands of monthly calls, chats, and emails, you’ll need an automated free-text analysis tool to categorise them.

The video on our homepage is an example of SentiSum’s machine learning-based analytics platform, which automatically understands, tags topics, and reports the hidden sentiments in customer support conversations.

3. Drill down with a ‘five whys’ analysis

Sakichi Toyoda, the founder of Toyota, popularised the “five whys” root cause analysis technique.

It’s as simple as asking why five times, which typically gets you to the core of the problem.

5 Whys

Source: Kanbanize

In a customer service root cause analysis, the five whys technique might look something like this.

👉 Our top reason for contact is ‘I can’t login’

👉 Why? Because they can’t find the login page (you’ll have to drill down into conversations to find this level of detail—SentiSum does this automatically for you).

👉 Why? Because they’re first visiting the homepage, which doesn’t have a clear login button.

👉Why? The website team didn’t think about it.

We got to our root cause pretty quickly here, but you get the idea. The five whys helps you understand what the real problem is and, if you’re lucky, points you to who can fix it.

4. Report Your Analysis

Periodically, or at the end of the project, you’ll want to create visual reports that show the volume of each topic mentioned and give guidance on the ‘why’ behind it.

Depending on the goals you set initially, your analysis should point to quick fixes or bigger projects that need to be initiated.

Top tip: Illustrate your quantitative data (e.g. # of monthly tickets on each topic) with qualitative data (e.g. an example quote from a customer about the issue). This really helps spur action amongst other departments. (This is how Heidi, Head of Support at Trivago, influenced other departments).

Examples of a customer service root cause analysis

There aren’t many detailed examples out there on the web, but my favourite examples are these ones:

Here, the UK government’s user research team used a customer service root cause analysis to identify quick fixes (gaps in product pages or documentation) and also used support tickets to do a usability test on certain features. Luckily, they only had a small number of tickets to manually analyse.

Zendesk’s documentation team member does a basic categorization to find common issues in support conversations. This is then used as a foundation to start a knowledge base.

Leverage SentiSum for your customer service root cause analysis

SentiSum customers use our platform to automatically surface important topics and customer sentiments from customer service conversations. With our accurate, granular tags you can unlock time-saving automations and experience-boosting insights in real-time. 

Book a product tour with us here to see how we can help you.

Support Insights Community
Join a community of 2200+ customer-focused professionals and receive bi-weekly articles, podcasts, webinars, and more!
Thank you! Your submission has been received!
Oops! Something went wrong while submitting the form.

Customer Sentiment Analysis FAQs

How is sentiment analysis useful? 5 examples of customer sentiment analysis

Here are 5 ways sentiment analysis is useful in customer service:

Prioritize customer issues:
Sentiment analysis can help businesses quickly identify and prioritize customer issues based on the emotional tone of their messages. This can enable customer service agents to respond promptly to unhappy customers and resolve issues before they escalate.

Personalize customer interactions: By detecting the emotional tone of a customer's message, sentiment analysis can help businesses tailor their responses to the customer's needs. For example, if a customer is expressing frustration, a customer service agent can respond with empathy and offer a solution to address the issue.

Improve customer experience: By providing personalized and efficient customer service, sentiment analysis can help improve the overall customer experience. Customers who receive prompt and effective solutions to their issues are more likely to remain loyal to a business and recommend it to others.

Analyze customer feedback: Sentiment analysis can be used to analyze large volumes of customer feedback to identify trends and patterns. This can help businesses identify areas for improvement and make data-driven decisions to improve their products and services.

Monitor brand reputation: Sentiment analysis can be used to monitor online mentions of a brand or product to detect negative sentiment and address issues before they become a larger problem. This can help businesses protect their brand reputation and maintain customer loyalty.

What is real time sentiment analysis in customer service?

Real-time sentiment analysis in customer service refers to the process of analyzing the emotional tone of customer messages or conversations as they are happening, in real-time. This enables businesses to quickly identify and respond to customer issues, prioritize certain conversations, and personalize interactions based on the customer's emotional state.Here are some examples and analogies to help understand real-time sentiment analysis in customer service:

Real-time monitoring: Real-time sentiment analysis involves monitoring customer messages or conversations as they are happening, in real-time. This is similar to a security guard monitoring a building in real-time for any signs of danger or security threats. Just as the security guard can quickly respond to any threats they detect, businesses can quickly respond to customer issues as they are identified.

Prompt customer service: Real-time sentiment analysis allows businesses to quickly identify and respond to customer issues before they become larger problems. For example, if a customer is expressing frustration about a product issue, real-time sentiment analysis can alert customer service agents to prioritize that customer's message for a quick response. This can help the business resolve the issue before it leads to a negative online review or loss of customers.

Personalized interactions: Real-time sentiment analysis can help businesses personalize their interactions with customers based on their emotional state. For example, if a customer is expressing happiness about a recent purchase, a customer service agent can respond with enthusiasm and congratulations. Conversely, if a customer is expressing frustration or anger, a customer service agent can respond with empathy and an apology. This personalized approach can help businesses build stronger relationships with their customers.

Improved customer experience: Real-time sentiment analysis can help improve the overall customer experience by providing prompt and effective customer service. Customers who receive quick and effective solutions to their issues are more likely to remain loyal to a business and recommend it to others.

Continuous monitoring: Real-time sentiment analysis can be used to continuously monitor customer messages or conversations, providing businesses with a wealth of data that can be used to improve their products and services. For example, if customers are expressing negative sentiment about a particular product feature, a business can use that information to make improvements and better meet the needs of its customers.

Overall, real-time sentiment analysis is a valuable tool in customer service that can help businesses quickly respond to customer issues, personalize interactions, and improve the overall customer experience.

What type of information do companies analyze when conducting sentiment analysis?

Here are the two overarching areas of customer information you can include in your sentiment analysis:

Text data: Sentiment analysis of text data is like analyzing a written letter to detect the writer's emotional tone. By detecting the emotional tone of customer feedback, customer service chats, reviews, or social media posts, companies can gain valuable insights into how their customers feel about their products or services.

Voice data: Sentiment analysis of voice data is like interpreting a person's tone of voice during a conversation to detect their emotional state. By analyzing phone calls or video chats with customers, companies can detect the emotional cues in a customer's tone of voice, such as frustration or anger, and provide a more personalized response.

What are the main goals of sentiment analysis?

The main goals of sentiment analysis are to gain insights into customer emotions and opinions, and to use these insights to improve customer satisfaction and loyalty. Here are some examples of the main goals of sentiment analysis:

Understand customer feedback: One of the main goals of sentiment analysis is to understand customer feedback and opinions about a product, service, or brand. By analyzing the emotional tone of customer feedback, companies can gain insights into what customers like and dislike about their products or services, and make improvements accordingly.

Improve customer experience: Another goal of sentiment analysis is to improve the overall customer experience. By understanding customer emotions and opinions, companies can address any issues or pain points and provide a better customer experience. For example, if sentiment analysis reveals that customers are frequently complaining about long wait times, the company can take steps to reduce the wait times and improve the customer experience.

Enhance customer engagement: Sentiment analysis can also be used to enhance customer engagement by identifying opportunities for positive interactions with customers. For example, if sentiment analysis reveals that customers are expressing positive emotions towards a new product or service, the company can engage with those customers to learn more about what they like and how they can improve the product or service even further.

Prevent negative customer experiences: Another goal of sentiment analysis is to prevent negative customer experiences by identifying potential issues and addressing them proactively. For example, if sentiment analysis reveals that customers are frequently complaining about a specific product feature, the company can address the issue before it becomes a bigger problem and affects customer satisfaction.

Monitor brand reputation: Sentiment analysis can also be used to monitor brand reputation by tracking what customers are saying about a brand, product or service on social media, review sites, and other online platforms. This information can be used to prevent a potential PR crisis and maintain a positive brand reputation.

Want to learn more about how SentiSum automates your customer sentiment analysis? Book a meeting with our team here.

How to Do a Root Cause Analysis in Customer Service

Sharad Khandelwal
Sharad Khandelwal
CEO & Co-founder at SentiSum, Expert in AI Analytics

Imagine you read a CSAT survey result that angrily calls out an agent for being unhelpful and slow.

You look closer at the corresponding conversation and you see the problem was the customer wants to update their password, but the agent failed to help.

You think "Great! This is easily solvable with some training."

You quickly kick off a new agent training around the topic. But negative CSAT scores keep coming. And, not just for the original agent, but all of them.

This issue is now driving 10% of your customer contact volume. You think "What the hell going on here?".

It's prime time for a root cause analysis.

You use our guide below to dig deeper and you find its not the agents fault at all. It's a bug in the product.

You let the product team know and they fix it in a matter of hours—saving costs, agent time, and improving your customer's experience.

Now, let's look a little closer at how to do that root cause analysis.

In this article:
  • Benefits of root cause analysis in customer service
  • How to Do a Root Cause Analysis of Your Support Tickets
  • Examples of a customer service root cause analysis

Benefits of root cause analysis in customer service

Getting to the root of reasons for contact in customer service is mutually beneficial for your team, your company and your customer.

We work with customer service teams of all sizes and company types, and their use cases for support conversation root cause analyses vary, too. 

Generally, they fall into three categories:

1. Improving the agent experience

By identifying the true root cause of support conversations and CSAT feedback, you’re able to better train agents and better equip them to quickly and diligently solve the issue.

For example, if you know that 5,000 tickets per month are about a specific, easily solvable issue, you can build a knowledge base article that your agents can send to the customer. A win for ticket resolution time.

CSAT feedback is also a valuable resource for agents. Root cause analysis of CSAT feedback, for example a topic analysis, reveals the true driver of the score and the associated sentiment. 

With our customers, we help them identify the behaviours and words that agents use which create negative sentiments, this tech-led QA technique helps each agent know where to develop their skill set.

2. Minimizing ticket volume & bringing down costs

Getting to the root of ‘reasons for contact’ is the first step towards tackling those reasons.

For example, in a Support Insights podcast episode with Lauren Cunningham, Loom’s customer support lead, we learned how Loom brought their contact rate down from 4% to 1%.

Over the pandemic, Loom’s customer base exploded from 700,000 to 10 million! At a contact rate of 4%, that meant an enormous surge in ticket volumes.

Starting with a root cause analysis of their customer conversations, the Loom team identified FAQS, things users commonly struggled with, and particularly time-consuming issues. That analysis fed into their ticket minimizing strategy, like their new knowledge-base which receives over 1 million in traffic each month—significantly reducing the need for customer’s to reach out to an agent.

Bringing down costs in this way is top of mind for many support teams. Gousto shared in this interview, that their real-time support analytics:

“Significantly reduces the lead time to unearthing customer insights and sharing these across the business, which allows us to positively improve [metrics like retention and reducing customer contacts].”

In a separate interview, Gousto’s Director of Support shared how root cause analyses in customer service also helps them identify the most frequent ‘reasons for giving a refund’ to their customers. By prioritising fixes to these issues, the company significantly reduces the financial impact of refunds.

3. Improving customer experience (CX)

Many of our customers invest in AI-support conversation analysis because they want to improve their CX.

Support conversations are a valuable source of customer feedback, to quote the UK government website:

“We believe user researchers should be as close to our user support tickets as our support teams and developers. They are a constant source of feedback and feature requests that, unprompted by us, reveal the issues that are most important to our users.”

This sentiment is echoed by Loom, Vinted, Gousto and many other customer service teams who now work diligently to feed real-time customer insights to other departments in their company.

How to Do a Root Cause Analysis of Your Support Tickets

Your root cause analysis can be as easy or complicated as you make it. I recommend you kick off with a simple project to ease everyone in and prove the value.

Here’s the process we guide our customers on:

1. Identify your goals

Starting with an idea of what you’re looking for saves a ton of time. For example, some goals you might be interested in are:

  • Identify frequently asked questions (FAQs): “Users always ask X.”
  • Find the most common issues: “What’s causing our top 3 drivers of customer contact?”
  • Find time-consuming issues: “Which tickets are taking the longest?”
  • Reduce costs: “What topics are driving refunds?”
  • Create growth: “Which areas of the buying journey are presenting a barrier to purchase? What topics are the cause of cancelled subscriptions?”
  • Improve CSAT: “What’s the most common topic in our lowest CSAT scores?”
  • Improve a particular product area: “How are customers reacting to our new feature release?”
  • Find agents who need training: “Why are X, Y, and Z agents receiving low CSAT scores?”

These will help accelerate your direction of travel and, importantly, tell you who to get involved in the planning process.

To make your analysis worth it, the results have to make an impact. The number rule of getting other teams to listen and act on your findings is to bring them in early and make them feel like their needs are being met (Ed Deason, Pret A Manger’s Ex-Director of Experience gave us this great advice in our guide to Customer Journey Mapping.)

The UK Government’s Product User Research team note in their planning stage that:

“To make sure our analysis focused on what was most useful, we initially ran workshops with each product teams to agree the objectives of this research.
In these sessions we printed off a random sample of tickets and gave each team member 5. Everyone was asked to write on Post-its what things they’d want to capture about the ticket, for example the type of request or the date it came in.
We then refined everyone’s Post-its and, by the end of the session, came up with a basic framework for analysis.”

2. Categorise support tickets

Anyone who works in the customer service industry will know the sheer volumes of conversations that happen every month. They’re not only high volume and frequency, they’re also made up of complex, qualitative free-text.

So, how do you approach the problem of reading them all and finding patterns?

Your first point of call is categorise all your conversations or CSAT feedback using topic tags. This will allow you to separate relevant tickets and track volume patterns.

You have two options here:

  • Manual categorization:

If you have only a handful of tickets and a good amount of time to invest, a manual analysis is your best bet. 

To identify the root cause of your support conversations, you first need to categorize them into topics using tagging.

Use our guide here to build a tagging taxonomy that suits your goals (we also tell you what a tagging taxonomy is, if you aren’t yet familiar with the term).

With your taxonomy in place, you must either go through your support request backlog and apply the tags or ask your agents to tag new conversations as they come in—which would only show topics in the latest tickets.

  • Automated categorization:

If you have thousands of monthly calls, chats, and emails, you’ll need an automated free-text analysis tool to categorise them.

The video on our homepage is an example of SentiSum’s machine learning-based analytics platform, which automatically understands, tags topics, and reports the hidden sentiments in customer support conversations.

3. Drill down with a ‘five whys’ analysis

Sakichi Toyoda, the founder of Toyota, popularised the “five whys” root cause analysis technique.

It’s as simple as asking why five times, which typically gets you to the core of the problem.

5 Whys

Source: Kanbanize

In a customer service root cause analysis, the five whys technique might look something like this.

👉 Our top reason for contact is ‘I can’t login’

👉 Why? Because they can’t find the login page (you’ll have to drill down into conversations to find this level of detail—SentiSum does this automatically for you).

👉 Why? Because they’re first visiting the homepage, which doesn’t have a clear login button.

👉Why? The website team didn’t think about it.

We got to our root cause pretty quickly here, but you get the idea. The five whys helps you understand what the real problem is and, if you’re lucky, points you to who can fix it.

4. Report Your Analysis

Periodically, or at the end of the project, you’ll want to create visual reports that show the volume of each topic mentioned and give guidance on the ‘why’ behind it.

Depending on the goals you set initially, your analysis should point to quick fixes or bigger projects that need to be initiated.

Top tip: Illustrate your quantitative data (e.g. # of monthly tickets on each topic) with qualitative data (e.g. an example quote from a customer about the issue). This really helps spur action amongst other departments. (This is how Heidi, Head of Support at Trivago, influenced other departments).

Examples of a customer service root cause analysis

There aren’t many detailed examples out there on the web, but my favourite examples are these ones:

Here, the UK government’s user research team used a customer service root cause analysis to identify quick fixes (gaps in product pages or documentation) and also used support tickets to do a usability test on certain features. Luckily, they only had a small number of tickets to manually analyse.

Zendesk’s documentation team member does a basic categorization to find common issues in support conversations. This is then used as a foundation to start a knowledge base.

Leverage SentiSum for your customer service root cause analysis

SentiSum customers use our platform to automatically surface important topics and customer sentiments from customer service conversations. With our accurate, granular tags you can unlock time-saving automations and experience-boosting insights in real-time. 

Book a product tour with us here to see how we can help you.

Join a community of 2139+ customer-focused professionals and receive bi-weekly articles, podcasts, webinars, and more!