Vanessa Gates 00:13

Welcome back to masters of support. I am Vanessa Gates and today with me I have by fellow Playvoxer, Tadas Labudis. He is our Senior Director of Product Management here at Playvox. Welcome Tadas, thank you so much for joining me today.

Tadas Labudis 00:28

Hey, Vanessa, thank you for inviting me. It’s great to be at the beginning of this great new podcast.

Vanessa Gates 00:35

Yes. So Tadas, I had the amazing pleasure to meet you in person just about a month ago in Colombia, over some coffee and some wonderful bonding when Playvox had their summit. However, I know some of our listeners may not know who you are. So if you can just a quick moment, tell us a little bit about Tadas and kind of what you do here. Maybe even tell us a little bit about what you do outside of work.

Tadas Labudis 01:06

Yeah, sure. So I’m Tadas Labudis. As you can gauge from my name, I’m from Lithuania. This is a common Lithuanian name. And I work [as] the Senior Director of Product Management Playvox, and only been here since the beginning of 2022. And the way I ended up at Playvox is through the acquisition of my startup, Prodsight. So you know, Prodsight has been a big part of my life for the last four years, four to five years before. And outside of work. I have a nine month old baby daughter named Sophia, I spent a lot of time at her these days. And outside of that when I have time, I love geeking out on coffee, making coffee, different methods, espresso and things like that, and playing drums, when I get a chance.

Vanessa Gates 01:56

I did not know you had a nine month thoughts or something I learned here today. I’m sure you’re getting tons of sleep these days? That’s awesome. Congratulations on the new baby. So you mentioned Prodsight. So can you tell us a little bit more about what Prodsight is?

Tadas Labudis 02:18

Yeah, sure. So, you know, actually don’t come from support background, like a lot of Playvoxers do. And obviously, we do serve the sport community that I became fascinated by the world of support through my work with Prodsight over the last 4 to 5 years. So originally, I started working as a product manager, and I saw a lot of value in support tickets, the actual data, the inquiries that our customers would send to us in my previous role. So I found myself spending a lot of time reading and analyzing those tickets to kind of inform my decisions as a PM, kind of like using that as a voice of the customer almost. So, you know, before I was a PM, I had some startup experiences. And I felt that there might be, you know, a product idea in formalizing, making something out of this ticket analysis process. So and I kind of got this itch to start something again. So I left my job and started researching and prototyping this product. And the purpose of that product was really to help companies understand their support tickets, what’s in that free text, pile of data that they’re getting every day. So they did the idea seemed to really resonate with these early customers, we got a few of them paying for the solution early on as a prototype. And kind of like snowballed from there. So you know, we raised some investment from venture capital firms and angel investors, and then hired people and then kind of like started building more and more complex features, and signing more clients. So you know, what I learned about the space is, you know, the problems we’re solving really, were replacing a manual tagging process. So a lot of support teams learn about what’s inside these tickets with the contact reasons through tagging. And that means applying tags in a drop down, or kind of typing in the tag and applying it to a ticket as they’re working on those tickets. And then, you know, product teams weren’t really looking at that, as a data source. It just felt like this massive, big, overwhelming information source. So it seems rather, you know, maybe like speak to some customers directly or do like a survey and kind of use that as a signal. But I think that was like a massive missed opportunity. So, you know, the really the purpose of Prodsight was to, you know, take valuable data and make it logistically like, easy to use and hopefully have these companies make better product, support and business decisions. So this is kind of the background. I’m happy to talk more about Prodsight if you’d like.

Vanessa Gates 05:13

Let’s take a second and talk about why people are tagging support interactions.

Tadas Labudis 05:20

Yeah, so the way I see it, and a lot of our customers, you know, they come to see, when customers contact support, it’s rarely a good thing, right. So, if you don’t have a problem, you don’t write into support. So most often than not, the customer might experience some kind of issue or friction and whatever they’re trying to do. And then you know, the last thing to do is write the support ticket, and then wait for it for some help. So businesses need to understand what’s causing customers this pain and through experience I’ve seen that support tickets, they typically fall into these three categories. So you either have a like a product or a service issue. So the thing that you’re selling or offering doesn’t work as expected, or is broken. And another categorie is, it might work as expected, but customers don’t know how to use it so they need information about either prior to buying it, or after we bought it, they have trouble. And the last thing is things normally around payments. There’s always issues with the need for things, or subscribing to things and getting double charges and all kinds of issues. There’s other categories, and this varies from company to company. But these are kind of like the broad buckets. And then what tagging does is helps you quantify and break down these broad categories into more actionable, smaller, more specific buckets that give insights to the business.

Vanessa Gates 06:54

I think my when I call in to customer support, I feel like it’s always for payment. That is usually my customer pain always. Or if not, it’s some type of random question that falls probably in think you said, a product services or issues and I feel like that’s the other one that I’m usually always calling for. You talk about like quantifying it, you talk about kind of, you know, essentially high level, why are they calling in and in in figuring it out. But is there a successful way to kind of tag these as, as we’re putting them in different buckets, per se?

Tadas Labudis 07:30

So I think there’s some best practices around the process. So a lot of companies early on, maybe when they’re a startup stage, they start very informally. And you know, at first, you might not have any tags whatsoever, and then you’re like, Okay, actually, we don’t need to understand where like these 1000s of tickets are coming from. So we can like start managing those problems. So, you know, at scale or like more sophisticated teams, they would typically have what’s called a tagging taxonomy. So it’s really a fancy word for a structure for your tags. So then you’re thinking about the taxonomy, the best mental construct would be to think about the report you’d like to see, once the tickets have been tagged. So let’s say you’re looking last month, and you’ve received, say, 100,000 tickets, what were the about? So how many different categories or topics would we have? And then what kind of structure would there be? Would it be like a flat long list of things from like, most popular to least popular? Or would there be like categories, like we talked about, like billing and product issues. So there’s many different ways to organize this data, it taxonomy, if it’s managed, and if I’m like, thoughtfully created, will help. You know, if it’s reading these reports, making them useful, but also will help deciding what tags should be applied and when. So this is kind of like a planning phase. And then the next step is how that taxonomy is communicated to the agents that actually has to tag, and how that’s implemented inside of the support CRM. So so, you know, having taxonomies, one step, if your agents are not using it consistently, say if one agent is tagging ticket in one way, and other agents is tagging in a different way, maybe the third is even like forgetting to apply tag. And so there’s no tag in a conversation that leads to like, basically, that, you know, bad data in bad data out. So your report is not going to be as useful or insightful, or the stakeholders looking at the report are not going to trust it. So if you only tag 20% of tickets and left the others alone, can we trust that, you know, category number one is the most important issue, like what did we analyze 100%, like would those numbers or ranking change? So I think these are two things to keep in mind, like keeping the structure and then being consistent with it.

Vanessa Gates 10:10

Awesome, do you see some type of rhyme and reason to when you should be updating these tags?

Tadas Labudis 10:19

Yeah, so obviously, once you’ve implemented this taxonomy, what happens is new issues start appearing, right. So maybe you’ve released a new product that could come with like five or 10, different new issues that you’ve never seen before. So you need to have a process that kind of like Crito, like a reflection points like, right, we have these tags in this taxonomy. Like, should we add new tags, should we remove tags that are no longer relevant? Because the problem is, this is like, when you go to a restaurant, and you see a massive menu, maybe there’s like five pages of like dishes, like, the longer the menu, the harder is to pick, like what you’re going to eat. So it’s the same with this taxonomy. If you open the drop down, and it’s like an endless list of things, some of the things that you’ve never seen before, don’t understand, you’re not going to be able to efficiently pick the right tag. So either, you know, what we see is like, the result of that is, teams start picking tags that are like the easiest, or like the ones that are like the most general. But what that does, is kind of creates these like black hole buckets, where it’s like, issue: Other or like billing problem: Generic. And that, again, is like, yeah, having a long taxonomy is not helping, because you’re not tagging at that granular level, to coming back to that, like, very high level overview that people already know, like, we already know, when we have billing issues.

Vanessa Gates 11:47

So you’re mentioning here, you know, they’re going into these other buckets. And I guess, essentially, that’s just gonna end up opening up a whole other can of worms and probably caused unnecessary headaches of trying to identify what those calls are. But on average, how much time do you think these agents are spending tagging?

Tadas Labudis 12:12

Yeah, so in the companies that, you know, are serious about tagging, and that’s part of the boot process, you know, you can see anywhere between like 30 seconds to 60 seconds in choosing a tag for tickets. So that might be like, maybe at the beginning of the conversation, looking through the list and applying the relevant one, or maybe like track the conversation, adding additional tags, as it evolves or other types of topics come up. So if you think that has an average of 30 seconds per ticket, and then you have a situation where a company receives, say, 100,000 tickets per month, you know, if you do like some quick math, that’s about like 50,000 minutes or 833 hours. I’ve calculated this before, I’m not that good at Math on the Spot.

But that is a lot.

Tadas Labudis 13:03

That’s a lot of time. And that’s, that’s just the tagging, you know, if you’re running any kind of like monthly report, that overhead is gonna come on top of it.

Vanessa Gates 13:12

I know we like to be efficient with our time, we want to maximize as much as we can get done. So how can AI essentially help with tagging?

Tadas Labudis 13:24

Yeah, so this is something, you know, we’ve thought long and hard about at Prodsight, because like the product that we created, the whole purpose of that was to automate this tagging process. So automation, has certain advantages that that perhaps is difficult for humans to manage, and has some disadvantages. You know, when you think about this work of tagging, it’s not something that agents enjoy, that’s not their favorite part of their job. Their favorite part of job is solving customer problems. That’s why they got into support. So this is kind of seen as busy work, or kind of like auxilary work, and doesn’t always get like the best kind of investment. So what AI does, if you automate this process, essentially, what you’re doing is you’re training a machine learning model, or an AI model, or some mixture of other techniques and data science to basically like look for looking at text, and then imply what would be the right tag or set of tags for that piece of text. So once that system is set up, or trained or implemented, then it can do this job very consistently. Right? So it’s not going to like skip a ticket just because it doesn’t feel or because it forgets it’s, you know, machines are, like good at doing what they are, like designed to do. So you’re gonna get 100% coverage. That’s the first thing you’re going to analyze more data with this process. And I think the other benefit is, is consistency. So there’s no got, like a human bias in, you know what agent interpreting a ticket in one way, and then another agent interpreting the other way, that machine is going to interpret everything according to the rules or the training it’s received. So that’s the second benefit. And I think lastly, and this is the most important one. Because the other two you can like manage, with humans, like you can get people to be more compliant, and, like follow the taxonomy more. But the thing that’s really hard to do, it’s actually like a story you’d like to share from a customer is, what do you do, if you come across a situation that you’ve never included in your taxonomy? Right? So something has been happening for a while a customer has been talking about is reporting it. And then you’re like, at this point, today, you haven’t been tagging it, because it’s, it feels like it’s a new thing, or like someone in other parts of our organization, asked for a report. So this is like one of the reasons one of our customers, the Bouqs Company, a flower delivery company in the US. They used to receive those requests, and they have the support team dedicated to kind of satisfying those requests. So when it comes from a product or logistics team, they would say, Okay, let’s get you some data that you asked for. So they weren’t getting it was getting to the point where they were getting so many requests. And each of those requests would take like days, that they started turning them down, right. So someone would come and say, “Hey, I know that some of our, like, the case that we sent last week included a gift, but some of the customers didn’t get the gift, how many times was that reported.” So they’re looking at 1000s and 1000s of tickets, there wasn’t a tag for that. So they have to read every single ticket or a sample off those tickets, and then say, okay, like this one was about this issue or that issue, and then kind of backtag retrospectively. So that’s why it takes days, because you’re not only doing it, like whilst you’re serving a customer, you’re now like, going through backfill of 1000s of tickets to retag them. So the good thing about automation is, once you train the model to look for a certain thing, it can like backtag 1000s or millions of tickets in seconds. And you can see the data instantly. So that’s what they were able to do with the Customer AI. And, you know, looked at the investment that they made in the system, and the ROI they got and I think was just over 300% ROI, just from this conversation tagging.

Vanessa Gates 17:42

Amazing. 300 you said?

Tadas Labudis 17:46

Yeah, 318%? Actually, yeah.

Vanessa Gates 17:48

Wow. That’s, that is amazing. That’s amazing. Um, it’s funny, as you’re like, telling this story, I’m just thinking about, like, how most organizations, usually want reports and they want them fast, they want them yesterday. So it’s amazing how like, essentially, AI is almost like your little assistant, you know, constructing these reports and assisting you in whatever fashion you need it, essentially when it comes to this wonderful tagging system.

Tadas Labudis 18:18

Yeah, I think ultimately, you know, it’s not about you know, having technology for the sake of it, or just because like companies are investing in AI, it’s, it’s really about, like, you need data to make good decisions, because without data, you’re gonna make a decision anyways, but it might not be the right one. So it’s kind of like, how can we apply technology to, like a real business problem and give people what they want, and kind of get out of the way from there.

Vanessa Gates 18:47

Wow, this is great. Tadas. This has been amazing. I am learning so much for someone as I may or may not have mentioned in previous podcast episodes, I’m definitely a little bit newer into the industry. And as I’m taking everything in and soaking it up. It’s just amazing. It’s amazing. I don’t know how these contact centers would work without softwares like ours. It’s insane to think that there was once a world of just spreadsheets. Because that is insane. But this has been so great. I am so thankful to have you here today to talk to us about our wonderful Customer AI and kind of what your experiences has been thus far to, to get, you know, to Playox and where you know where you’ve been and what you’re doing now with us. And thank you, Tadas, so much for being here with us today. It’s been wonderful. I am Vanessa Gates. I am the host here of Masters of Support. We thank you for listening and make sure you hit the subscribe button so you can get notified when our next episode is out. And I hope you have a wonderful day and always be a good human. We’ll see you soon. Bye bye.