CX
Mar 26, 2024
In a recent conversation, Hugo Munday, Director of Customer Service at ThriftBooks, sat down with us to share his winning strategies for delivering a consistent, delightful service experience for thousands of worldwide book lovers who order their next literary adventure through their online store. ThriftBooks has an award-winning customer service team recognized by Newsweek for superb customer service two years in a row.
As the world's largest online independent used book seller, we here at ThriftBooks pride ourselves in doing everything we can to help our customers find what they’re looking for, and get them the stories they seek on their doorstep. Our promise is to always provide quality books, both used and new, delivered directly to our cherished customers.
Listening to how well we deliver on that promise hasn’t always been easy.
Analyzing the voice of our customers has always been a priority, but we struggled with two issues: scalability and freshness. Getting a statistically relevant snapshot of customer feedback was done manually. By the time we were able to analyze the conversations, the freshness of what happened had already gone stale, rendering a lot of those learnings obsolete.
As you can imagine, trying to surface information in the moment to help our agents and other departments make improvements was a tall task. It would often take days to find answers behind an unexpected surge in shipments taking longer to arrive or questions about website features.
We’ve learned and evolved a lot in recent years, particularly with the help of automation and technologies like generative conversation intelligence.
Here are the lessons I’ve learned and how my team has been able to up-level how we tap into customer conversations to offer more value across the board.
Hear more from Hugo firsthand in this customer story video.
Widening Our Snapshot (of what Customers Are Saying)
What is your current vision for customer service? Is it to be able to have deep visibility into customer needs? To improve agent performance? To report back with confidence to your leadership team?
If you’re anything like me, it’s all of the above.
The only way to check the boxes on all of those priorities is by having a real snapshot of 100% of your customer emails, chats, and calls.
The initial barriers we had in getting this degree of visibility boiled down to our approach to technology. We’ve used QA platforms in the past, but many were limited in the depth of data they surfaced, which I’ll expand on a little more further down in this blog.
When I began to learn more about the possibilities of generative conversation intelligence for customer service, I discovered there was an incredible amount of information that could be surfaced, without extra effort, to better monitor customer service performance and even more ephemeral metrics like tone and empathy.
We started by capturing three key things:
Customer sentiment - when are customers expressing frustration, or delight, and how can we reduce friction and replicate moments of positive feedback?
Agent sentiment - how is the agent delivering empathetic responses across channels, and how can we help coach them in key areas?
Customer intent - what are the top reasons behind customer inquiries, and how can we play an active role in tightening any loose screws in our operations
Because they’re completely automated, insights powered by gen AI enable us to be more specific about enforcing clear guidelines, such as agent adherence to brand tenants, that were harder in what I would call analog QA platforms in the past. By going beyond just average QA scores, we can capture a broad spectrum of our customer behaviors, their comments, and their overall experience with our brand.
Asking the Right Questions
One of the most important lessons I’ve learned has been figuring out how to ask the right questions. I will go into an example of how we shifted our line of questioning in one key area: our quality assurance strategy.
I’ll start by sharing how we ran our customer conversation reviews before. If we were trying to solve last week's problem, we were looking at a sample set of at least 200 conversations, reading those with a handful of team members, and making a judgment call on how to build the rubric.
The real pain for us was how time-consuming the whole thing was—and it resulted in over-stressed subject matter experts and supervisors who couldn’t keep up with the volume of work. We knew something needed to be done.
So, we turned to our quality provider to help us figure out how we could give our workforce better feedback without burning out our entire QA team. We spent weeks trying to re-think our scorecards and figure out how to ask different questions to streamline the scoring process. We came to the conclusion that even if we had fewer questions, there were many core competencies that were hard to grade objectively.
As it turns out, we were asking ourselves how to create a more efficient rubric, when we really needed to reimagine what the grading process looked like as a whole. When we understood that shift, it helped us zoom out and think bigger about how digital transformation could help us in our journey.
That’s where we discovered the power of automated quality scoring, which removed the subjective nature of scoring criteria and enabled us to automatically review and score every call, email, and chat interaction. We can now deliver personalized AI coaching to each agent to improve key behaviors, and instantly flag compliance violations as they happen. The result has been less overburdened customer service supervisors who can move much faster.
Ditching Manual Deep Dives
Our entire team dreaded the deep dives of our emails, chats and calls. We had a robust set of issue codes, which meant that going into the vast array of common support drivers to analyze and understand what went wrong took a lot of time. We’re talking days at a time.
We deliver physical products, books, to the homes of our customers. This means that hypothetically, if we had items being reported as delivered, but customers didn’t receive them due to porch piracy, we needed to find out what that problem is fast.
But by the time we had completed a reactive manual deep dive, the issue at hand had already subsided, meaning all of our efforts were for nothing.
The only way to pinpoint issues and act on them fast is by using what technology has to offer.
Here’s where generative insights really shined for us. We are able to monitor things like:
Book condition
Order issues
Promotion and reward inquiries
Positive customer feedback
With these learnings, we’ve been able to see the top reasons behind customer service inquiries, broken down by categories, helping us speed up our ability to know what’s going on much sooner.
Analyzing Our Third Party Reviews
In addition to our calls, emails, and chats, we also began to analyze 100% of our third party reviews to find customer insights.
We have approximately 1.8 million reviews that are 4.7 out of 5 stars, and the reviews are chalk-full of qualitative learnings.
If a survey respondent submits a poor review, it flows into our customer service queues and we will respond to try and mitigate that bad experience. Before using generative conversation intelligence, we were manually responding to the bad reviews, but no one was spending much time reading what was being said across all the millions of reviews.
A lightbulb went off when we discovered that we could use generative conversation intelligence to pick up all of our reviews and analyze them in a second layer of analysis. We're now actually able to find much more nuance in what our customers are saying and understand their perspective of the book condition much more thoroughly than just a CSAT survey.
Final Words
Generative conversation intelligence has played a big role in helping us better understand what our customers want and get to the real reasons behind less than stellar service experiences.
By widening our snapshot of customer sentiment, asking the right questions, and ditching manual deep dives, we've been able to overhaul how we understand and respond to customer needs. Through automated analysis of the entire customer sphere, from customer conversations to third-party reviews, we've gained deeper insights into our operations, enabling us to proactively address issues and deliver a consistently excellent experience to book lovers worldwide.
So, don't shy away from AI. It's now more affordable than ever and can assist in pinpointing customer needs and providing timely feedback to that will undoubtedly change your customer retention for the better.
Table of Contents
Widening Our Snapshot (of what Customers Are Saying)
What is your current vision for customer service? Is it to be able to have deep visibility into customer needs? To improve agent performance? To report back with confidence to your leadership team?
If you’re anything like me, it’s all of the above.
The only way to check the boxes on all of those priorities is by having a real snapshot of 100% of your customer emails, chats, and calls.
The initial barriers we had in getting this degree of visibility boiled down to our approach to technology. We’ve used QA platforms in the past, but many were limited in the depth of data they surfaced, which I’ll expand on a little more further down in this blog.
When I began to learn more about the possibilities of generative conversation intelligence for customer service, I discovered there was an incredible amount of information that could be surfaced, without extra effort, to better monitor customer service performance and even more ephemeral metrics like tone and empathy.
We started by capturing three key things:
Customer sentiment - when are customers expressing frustration, or delight, and how can we reduce friction and replicate moments of positive feedback?
Agent sentiment - how is the agent delivering empathetic responses across channels, and how can we help coach them in key areas?
Customer intent - what are the top reasons behind customer inquiries, and how can we play an active role in tightening any loose screws in our operations
Because they’re completely automated, insights powered by gen AI enable us to be more specific about enforcing clear guidelines, such as agent adherence to brand tenants, that were harder in what I would call analog QA platforms in the past. By going beyond just average QA scores, we can capture a broad spectrum of our customer behaviors, their comments, and their overall experience with our brand.
Asking the Right Questions
One of the most important lessons I’ve learned has been figuring out how to ask the right questions. I will go into an example of how we shifted our line of questioning in one key area: our quality assurance strategy.
I’ll start by sharing how we ran our customer conversation reviews before. If we were trying to solve last week's problem, we were looking at a sample set of at least 200 conversations, reading those with a handful of team members, and making a judgment call on how to build the rubric.
The real pain for us was how time-consuming the whole thing was—and it resulted in over-stressed subject matter experts and supervisors who couldn’t keep up with the volume of work. We knew something needed to be done.
So, we turned to our quality provider to help us figure out how we could give our workforce better feedback without burning out our entire QA team. We spent weeks trying to re-think our scorecards and figure out how to ask different questions to streamline the scoring process. We came to the conclusion that even if we had fewer questions, there were many core competencies that were hard to grade objectively.
As it turns out, we were asking ourselves how to create a more efficient rubric, when we really needed to reimagine what the grading process looked like as a whole. When we understood that shift, it helped us zoom out and think bigger about how digital transformation could help us in our journey.
That’s where we discovered the power of automated quality scoring, which removed the subjective nature of scoring criteria and enabled us to automatically review and score every call, email, and chat interaction. We can now deliver personalized AI coaching to each agent to improve key behaviors, and instantly flag compliance violations as they happen. The result has been less overburdened customer service supervisors who can move much faster.
Ditching Manual Deep Dives
Our entire team dreaded the deep dives of our emails, chats and calls. We had a robust set of issue codes, which meant that going into the vast array of common support drivers to analyze and understand what went wrong took a lot of time. We’re talking days at a time.
We deliver physical products, books, to the homes of our customers. This means that hypothetically, if we had items being reported as delivered, but customers didn’t receive them due to porch piracy, we needed to find out what that problem is fast.
But by the time we had completed a reactive manual deep dive, the issue at hand had already subsided, meaning all of our efforts were for nothing.
The only way to pinpoint issues and act on them fast is by using what technology has to offer.
Here’s where generative insights really shined for us. We are able to monitor things like:
Book condition
Order issues
Promotion and reward inquiries
Positive customer feedback
With these learnings, we’ve been able to see the top reasons behind customer service inquiries, broken down by categories, helping us speed up our ability to know what’s going on much sooner.
Analyzing Our Third Party Reviews
In addition to our calls, emails, and chats, we also began to analyze 100% of our third party reviews to find customer insights.
We have approximately 1.8 million reviews that are 4.7 out of 5 stars, and the reviews are chalk-full of qualitative learnings.
If a survey respondent submits a poor review, it flows into our customer service queues and we will respond to try and mitigate that bad experience. Before using generative conversation intelligence, we were manually responding to the bad reviews, but no one was spending much time reading what was being said across all the millions of reviews.
A lightbulb went off when we discovered that we could use generative conversation intelligence to pick up all of our reviews and analyze them in a second layer of analysis. We're now actually able to find much more nuance in what our customers are saying and understand their perspective of the book condition much more thoroughly than just a CSAT survey.
Final Words
Generative conversation intelligence has played a big role in helping us better understand what our customers want and get to the real reasons behind less than stellar service experiences.
By widening our snapshot of customer sentiment, asking the right questions, and ditching manual deep dives, we've been able to overhaul how we understand and respond to customer needs. Through automated analysis of the entire customer sphere, from customer conversations to third-party reviews, we've gained deeper insights into our operations, enabling us to proactively address issues and deliver a consistently excellent experience to book lovers worldwide.
So, don't shy away from AI. It's now more affordable than ever and can assist in pinpointing customer needs and providing timely feedback to that will undoubtedly change your customer retention for the better.
