“Do you enjoy filling out customer service surveys?”

That’s the question I posed to a group of customer service professionals. The response was a resounding, “No!!!”

Then why do we pester our customers with surveys?

You can eliminate post-transaction surveys with a new sentiment analysis technique called the sentiment arc. It works by analyzing the conversations you’re already having with customers to answer two fundamental questions:

  1. What’s driving customer sentiment (positive or negative)?
  2. Is the customer happier at the end of the contact than they were at the start?
Guest post by Jeff Toister

I teamed up with Balto to chat about the sentiment arc in a live Q&A session. There were so many great questions that we ran out of time before we could answer them all.

Here are the answers to some of the additional questions we received.

How does this differ from what other organizations are doing?

Traditional sentiment analysis evaluates the words a customer is using in a phone, chat, or other conversation to identify whether they are happy, neutral, or upset at the end of an interaction.

The sentiment arc analyzes customer sentiment twice: at the start of the conversation and again at the end. The change in sentiment helps identify specific things agents are doing to improve their customers’ experience.

In one test of 29,000 calls, a traditional sentiment analysis found that 88.3% of calls ended with neutral sentiment. That makes it look like agents aren’t adding much value.

Applying the sentiment arc painted a much different picture:

  • Sentiment improved on 46% of calls
  • 95% of calls that started with low sentiment ended higher
  • Sentiment decreased on less than 1% of calls

The sentiment arc also helped identify specific agent actions that led to improved sentiment, such as asking more questions to understand a customer’s needs.

What companies are starting to make this shift?

Sentiment analysis has been used for quite some time. Contact center software, voice of customer software, and quality assurance platforms all commonly contain sentiment analysis tools.

Sentiment arc is a relatively new concept, so very few companies have started using it. You could be at the forefront of a movement if you implemented this technique.

How do you run a sentiment arc analysis?

There are three steps to running a sentiment analysis on a customer conversation and calculating the sentiment arc.

  1. Evaluate customer sentiment at the start of the conversation
  2. Evaluate sentiment at the end of the conversation
  3. Identify topics or themes that drive any change in sentiment

Smaller contact centers can do this manually without any special software. One option is to have quality assurance analysts assess starting and ending sentiment for each contact monitored.

Many contact centers have sentiment analysis tools included with their contact center software, voice of customer platform, or quality assurance software. These tools can be deployed to calculate the sentiment arc as well.

Does this approach help uncover what made the customer unhappy to begin with?

Yes. Part of any sentiment analysis is identifying the themes that drive customer sentiment. This includes determining specific reasons customers are upset at the start of a contact and evaluating what agent actions work best to help customers feel better.

How can you bring this back to the team to make changes?

A sentiment arc analysis helps you identify specific agent actions that are more or less effective at improving customer sentiment.

For example, a contact center leader used sentiment analysis to discover that effective technical support reps used something called positive positioning. This is a way of providing solutions using more positive language.

“You have to reset your router,” sounds negative and demanding. Positive positioning reframes that same request to become, “Let’s try resetting your router—that often fixes issues like this.”

The support leader found customers are more compliant with technical support suggestions and sentiment was much higher when agents used positive positioning. Average handle time and repeat calls decreased when positive positioning was used, saving the company money. 

The contact center leader trained her entire team on this technique as a result of her analysis.

Is this a real-time assessment or a process that’s run on a schedule?

The answer depends on the tools you are using to analyze the sentiment arc.

Some software is able to analyze a call in real-time. Other tools require some manual inputs after the contact, so it’s more efficient to analyze batches of contacts at one time.

Do you have to calibrate against actual customer surveys?

The sentiment arc is intended to replace your customer surveys. This technique provides a larger data set, more insights, and doesn’t annoy your customers.

However, your executive team might want to compare sentiment arc results to your existing survey before committing to a change.

What do I need in terms of technology, skills, and budget to implement this in my organization?

In many cases, you can get started using the resources you already have.

Check to see if your existing contact center software, survey provider, or quality assurance platform already has sentiment analysis capabilities. You can also evaluate the sentiment arc manually by adding starting and ending sentiment to your existing quality assurance form.

You’ll need the same analytical skills you’re already using to evaluate customer feedback. For example, if you search your survey comments for trends, you can apply the same technique to find trends in customer conversations.

Budget depends on what tools you already have. One way to make the case for more tools is to demonstrate how the sentiment arc can save your company money or increase revenue.

For instance, one sentiment analysis found that customers spent 13% more when agents used clear and specific language to respond to customer inquiries. Highlighting a 13% boost in sales makes a strong case for investing in more robust tools.

Caution: Avoid using non-private tools such as ChatGPT to analyze your customers’ private data.


It’s time to stop pestering customers with endless surveys. Moving to the sentiment arc yields more data, better

Jeff Toister is the bestselling author of The Service Culture Handbook, a step-by-step guide to getting employees obsessed about customer service. Over 3,000,000 people have taken one of his LinkedIn Learning training courses. Global Gurus has ranked his free Customer Service Tip of the Week email as one of the best customer service training programs in the world.