In this episode of the Consumer Insights Podcast, Thor is joined by Victoria Gnatoka, Customer Experience Product Manager at Expedia.
By next year, 70 to 80% of all the data we receive will be unstructured. Is your organization prepared?
In this episode, we are joined by Victoria Gnatoka, Customer Experience Product Manager at Expedia. Drawing on over 10 years of experience, she digs into how adopting an omnichannel approach can help obtain unstructured data that can be turned into quality, actionable insights through AI, data engineering and other technologies.
We also discuss:
You can access all episodes of the Consumer Insights Podcast on Apple, Spotify, or Spreaker. Below, you'll find a lightly edited transcript of this episode.
Thor:
Hello everyone and welcome to the Consumer Insights podcast. Today, I'm excited to have a remarkable insights leader joining me for what I know will be an illuminating conversation. I'm thrilled to introduce today's guest, Victoria Gnodryka, customer experience manager at Expedia. Victoria is a customer and strategy focused professionals with more than 10 years of experience assisting senior leaders and strategic planning by putting the... Sorry, by putting the customer first. She's most passionate about managing complexity and thinking about future AI enabled solutions in her daily research work. She has been working on unstructured data enablement and government, sorry, and governance for the last 10 years and loves the area of what solutions it offers to brands and companies. Thank you so much for joining me, Victoria. Now, okay.
Victoria Gnatoka:
Thank you so much for having me. It's a pleasure.
Thor:
So, Victoria, we are super excited to get to know you, but to get things started, could you take a couple of minutes to tell us about yourself, your role, and how you got to where you are today? How did it all begin?
Victoria Gnatoka:
Of course. So I manage customer experience program at Expedia group today, where really what I do is I look after our omni channel customer feedback and customer data ecosystem, which serves as a foundation for our insights. And I do it at a global level across all brands, all points of sale, which makes it really challenging, but also really interesting. I have done similar programs and managed similar programs in the past for other companies such as Nordstrom and United Airlines. And really where my passion is, is taking that data and feedback that your customers give about an experience that they have and transforming into an insight that can drive business decisions. Another important part of my background and also my role today is to manage unstructured data. And I always say that's where really it all began. Because unstructured data represents all of the data that we receive today from social media, emails, surveys, reviews, et cetera, which basically is a language. And to me, language always meant emotions, understanding, empathy, all of those feelings that today brands want together from their experience with customers to understand what really makes them like their brands, love their products, etc. However, when I graduated with my linguistics degree many years ago, I didn't know that things like text analytics, natural language processing, natural language understanding would be solutions that businesses would want implemented scale. At the time,
I just knew I loved languages. I loved the emotions and the understanding that it can produce. And I graduated with my degree. I was fluent in five languages. But I really started to pursue my roles more like on the market research side.
And it was only about five years into my career that I realized there is actually an opportunity to put language in use to drive these business decisions. And I started to manage and get into customer experience programs where, again, the interest in the unstructured data started to surface. And that's where it really came together for me, my passion for language as a linguist, but also my passion for driving business decisions.
Thor:
Wow, language, structured data, unstructured data, all of it helps us uncover insights. What are insights? How would you define it? And how has that definition changed over the course of your career?
Victoria Gnatoka:
I love this question because I think about it all the time. Well, first of all, of course, you know, as linguists and me, I want to say what an insight really means. Well, what it means is our ability and power to look into or within something, right, within a situation. So really to dive deeper.
And to me, in my experience, insight means knowledge. And the reason that I say this is because I think insight shouldn't be trapped in a PowerPoint or in a dashboard, it cannot be simply a fact or a statement. And the reason that I define insight as a knowledge is because I think it's something that you as an analyst or a manager of insights, you put time together to really understand your customer.
You put all the data, you did an analysis and you develop this knowledge, which is an insight. And then you bring it to the table to your stakeholders. And your goal with that knowledge is for them to listen to it, to empathize with it, and then to take it forward in the organization. And then they can take that knowledge and share with others. They can act on it. They can be educated upon it. And I think really that's to me what insight represents. What I have observed in my 10 plus years experience in the insight space is that when I started, really insight was a recommendation a lot of organization and teams that I worked on, everything they wanted and thought that insight is, is just give me a recommendation. But it was really not that effective. Then it became a fact. Then I had a time in my career when insight equals news, bring me something new. I want something new. And today what I've started to observe in the last several months, was the release of generative AI, in, 2000, in 2021 is that today, unfortunately, Insight starts to lean towards being that summary or summarization of data, which unfortunately, I don't think that's what it is. I really think Insight is a deep knowledge that look deep into something to understand the problem and then bring that knowledge forward.
Thor:
Fantastic. Absolutely love that. Absolutely love that definition. Now, Victoria, I know you're a strong believer in taking an omnichannel data approach to your insight strategy. For anyone who might be less familiar, could you share a bit more about what you mean by omnichannel data approach?
Victoria Gnatoka:
Of course. So Omnichannel data approach, so let's unpack it a little bit.
When we talk about Omnichannel, these are the different channels that you can collect your data from. So of course you can collect data from session replace, from all of your customer databases, loyalty programs, but you also have the data coming through surveys, reviews, emails, chat bots, conversations with agents, social media.
Perhaps videos that your customers post, reviews that they share on your site. All of these are different channels that companies today give customers an ability to share their feedback with them to live, and that's where we always collect the data. And what Omnichannel approach means is really instead of working in silos and instead of having just one team focusing on either driving insights from, let's say, a survey channel, of data or maybe chat bot channel of data or email, but really look at it all together. And the biggest benefit that I've seen of doing that is not the vast majority of data that you're going to receive, but first of all, the quality and then the ability to see the blind spots. Because a lot of times what happens to the customer, they of course don't see you as all these, you know, 10, 20 plus channels. They see you as one brand, one company. And one promise that you make to them that you're going to deliver that experience. So having that omnichannel approach, you can look at it as a customer would look at your, which is one approach, one omnichannel experience. And then when you put all this data together, you start seeing similarities where perhaps you maybe collect too much data, where you may maybe collect too little. And of course, it's a big unlock for insights because where sometimes you might see, I've been in this situation many times where, you know, you want to address a business problem and you lack the data and you either start producing the insight that's not really actionable or you just simply say, we don't have enough data, which is not really true. You do have the data, but it might be sitting in all this other channels. And the first time when I tried this omnichannel approach, it was really eye opening to how much deeper you can go when creating the insights.
Thor:
I love that. And why do you think that taking an omnichannel data approach is so important? Were there any particular experiences that made you realize their value?
Victoria Gnatoka:
Yes, there were so many. One of my favorite that comes immediately to mind, I worked on a big accessibility project for a big brand. And we really were trying to understand what drives the accessibility, the digital accessibility experience with our sites and apps. And the traditional approach was to look at this one particular data set. And we really didn't see much. So an immediate assumption was that perhaps customers who have those accessibility needs simply either are not equipped with the right tools to leave that feedback or share that experience, or perhaps it's in our end where we did not do the right job to collect that data. However, what turned out is when we, having that omnichannel approach, when we dived into our other data sets, such as the experience with the chat bots or conversations with an agent, it was in those transcripts where actually the experience was happening. Because naturally, as a customer, when you could not have enough accessibility to access something, perhaps you couldn't see the image, you couldn't read the font, maybe the keyboard didn't pop up, et cetera, you naturally would want to reach out to someone who could help that someone with either a chatbot or an agent. And all of a sudden we were presented with this richness of data that allows us to produce insights and deliver actions on how to improve experience. For example, if you needed a wheelchair assistance or maybe you wanted to bring your service dog to your hotel room, et cetera, et cetera. And all of a sudden we had this rich data. And again, we wouldn't be able to do that if we were simply focused on this one single data set. This is just one of the many examples where when I think back, I really think that having all data come together, you can see how it's flowing through and again, where your customer chooses to give you that feedback about their experience.
Thor:
There are so many things you're saying that I want to unpack, but if we switch gears for a bit and go back to one of the concepts you used in the very beginning. You talked about structured data and unstructured data. And if we zoom in a bit more, if I know you've developed quite a bit of expertise working with unstructured data, like open -ended feedback from surveys, reviews, and conversations, why do you think insights leaders should be incorporating unstructured data into their insight strategies?
Victoria Gnatoka:
Great question. Well, a lot of the research and studies out there show today that only 20 % of insights are surfaced from structured data. And that happens pretty with less effort and pretty easy. And then 80 % of insights are surfaced from unstructured data, which obviously requires a lot of times. There is also research recently done by Qualtrics, a customer experience management company, that shows that by 2025, 70 to 80 % of all of the data that we receive will be unstructured. Because the amount of how customers interact, so again, think about social media, videos that become transcripts, conversations with agents etc, that all drives that unstructured data. And we're having more and more of it. So really it is sort of, I call it a gold mine that I feel Insights team have been sitting on for a while and it's really no longer the time to avoid it.
Because if you avoid it, you just keep missing on opportunities of where you can drive the insights from. And also this is just the reality of the interactions and how, what are the different mediums that customers get experience with today.
So I really think it's very important every organization has it, that it takes time and effort to put a structure around it.
Thor:
I think that's such good advice and for insights leaders looking to incorporate more unstructured data into their insight strategies. How do you suggest they going about doing this? What's step one? What's step two?
Victoria Gnatoka:
Well, step one, I would say is commitment and allocate dedicated resources. When I started to work on unstructured data, probably about 10 years ago now, a lot of times the ratio on the teams I've been on would be literally one to five. One person focused on unstructured data and then the rest of the people on the team focused on that structured analysis. And it cannot be like if you don't have that unstructured program in place and managing that unstructured data, it is a big effort where, especially if you're in the beginning of this journey, where it simply cannot be someone's part -time job or you cannot spread evenly across an already existing insights team, you really need to commit and dedicate resources to that. Especially in the beginning when you want to mine that data, understand what's actionable, what's unactionable, and how to tie it back to your structured data. So it's really about dedication and resources. And step two, I would say it's figuring out what are the business problems that today perhaps your structured data does not address, that unstructured data will help you address. What I have seen in the past, a lot of times unstructured data will be sort of a nice to have. And that's really not true. An example that comes to mind when I worked for this large retailer, we were analyzing reviews and surveys to understand that experience with the product. And we also thought that that feedback and unstructured data that we receive from those sources would also tell us a story of what customers want more of, right? So what are they looking for?
And what we totally ignored at the time is that search bar that's on every brands and company site because we thought it's being used kind of as a filter where customers would go and input and say, I'm looking for XYZ brand or size, which turned out to be not true because they were doing this by using already existing filters on site. What they were doing by using that search bar was actually to search for those trends they saw somewhere else, like on social media. So they would input a lot of keywords and phrases trying to find those products on our site that we even didn't have. So when we pulled that data and those queries into our ecosystem, that would help us answer that question of what else can we sell on our sites? What brands should we consider? What may be new products customers want in a particular line? And all of it was really answered through that search bar on a site, which we thought the intended use was completely different. And that's where I think the that was to me an example of you have to look for those opportunities as a step to how unstructured data, what business problems it will help you solve. Because again, a lot of times like in my example was the search bar in relation to product reviews is completely two different things of where you think the data is to where it actually lives.
Thor:
I absolutely love that. And just trying to unpack some of the things you said there. I mean, you mentioned that it's often somebody's part -time job. Why do you think that is? Why do you think insights teams don't incorporate unstructured data into their strategies more often?
Victoria Gnatoka:
Yeah, there's so many reasons and I must say that in my, again, almost 10 years of experience in working with unstructured data, the reasons that were there about 10 years ago are still here today. So they are, first of all, the complexity of analysis. To analyze unstructured data, a lot of times I worked in an organization's team that simply there are no tools or processes in place because unlike structured data that can be put into you know, Excel or any Power BI tool, it's not as simple. And that is really the complexity.
You have to kind of unpack and have the tools in place, whether internal or you find like a third party partner that's going to partner with you on that, that you can actually unpack and transform those words into actions.
So that's definitely number one. The other is perceived value and ROI of unstructured data. Unfortunately, because it is unstructured and it takes time and effort and it feels that it's going to take longer time to produce an insight. A lot of time, honestly, structured data just wins over because it's very easy to produce an insight or drive action from structured data. It's fast and it can be quantified versus with unstructured data. It's not that easy to quantify. You really have to perform several approaches and analysis to come up with a method that works for your organization so that perceived value and ROI of unstructured data unfortunately is there. Another reason is lack of integration with the organization's existing systems. If you already have existing reporting or existing structures in place, a lot of times they simply do not support. So you cannot put together structured and unstructured data and let's say have a Tableau dashboard. It's simply not going to work. A lot of these tools are simply not designed to ingest text and present it in a structured way that then your teams can access and look at the trends. So what happened in my experience, most of the time you would need to partner with data scientists and engineering teams to make sure that these different tools and system are integrated to support unstructured data. And of course that requires time and effort. And then the final consideration of a reason why I've seen companies, it seems might be reluctant to work with unstructured data is privacy and security. So it is obviously data that is coming through the language. And I always am fascinated by it because if customers are trusting you so much that they're leaving you private information in the data, trusting you not only going to action on it but you're also gonna actually protect their data. To me, that's always mind blowing. But of course, it's your job as a brand and company to protect their data. Usually there is a lot of privacy and security information masking solutions that you have to implement when you work with unstructured data to protect the customer. But that is also, I've observed, can be one of the reasons of why companies might be reluctant. So to summarize, its complexity of analysis, lack of integration with existing systems, perceived low value and ROI, and security and privacy.
Thor:
Now, in preparation for this episode, you also share that you think there's a tendency for insights professionals to try to be broad, getting caught up in focusing on the quantity of insights rather than the quality of them. If we take this into the context of what we've been discussing so far in this episode, how do you think insights leaders can ensure quality when incorporating more types of data sources?
Victoria Gnatoka:
Great question. Yeah, I think that what I've observed again, and like insights like to be trapped in a dashboard or a pound point. And often it feels that the broader we go, the more of the dashboards or PowerPoints we have, it's going to drive more actions, which is not true. I'm really the supporter of having less insights, but more actionable insights. And that's where I think the having this incorporating more data sources actually helps. And the reason for that is that naturally it might feel bad by incorporating more data sources and bringing more data. You actually going to make it more difficult for the insights team, but that's really not true. First of all, to incorporate more data sources and bring more resources, you of course need to pre-process the data.
You need to clean up the data and imagine if every time somebody on your insights team had to do it manually, every time they wanted to broaden the scope or the data sample that they have that is simply not feasible. And that of course drives the quality of your insight that you might be missing something.
You might have those blind spots. But if you already have integrated most of your data sources in one place, you don't need to worry anymore about this pre-processing. So for example, today in the feedback ecosystem that I manage in my daily job, I already have sources coming from 10 different places and daily when I log in, I don't need to worry about if they come in regularly, if there is a data lag, if the data has been cleaned up. It's all already been automated. All I need to worry about is how to drive the insights from that. Another thing that's really powerful when you bring additional data sources into your ecosystem, you're going to realize that not all data is good data, and also not all data is actionable.
I've seen in my experiences that about 30 to 40%, especially of unstructured data, tends to be unactionable, which is very normal because, again, it's a language, it's a text, it's a transcript of a video or social media post. And naturally, as customers use that language to share their experience, there will be filler words, there will be generic phrases, there will be generic complaints. And that's where you put systems in place to filter it out. So you actually, by adding more data sources, you actually come out as a result with more data, but more quality data, because in that process, you kind of filter out that unaction ability. And that helps you to narrow down that insight.
And another power of having this omnichannel approach that I think to insights to drive that quality versus quantity is you can create segments of channels where you pull the data from and that helps you drive that quality. So for example, if you're working on a particular problem, instead of looking at it all, you might say, well, I want just signals from transcripts, emails and social media. And to address this program for this particular brand and for your different brands, right? It might work differently. So for example, at Expedia group, we have hotels.com, Vrbo, Expedia brands and for each of those brands, obviously different data sources will work differently just by the nature of experience on those brands. So naturally you create those kind of data channel six segments and that again allows you to create that quality of insights rather than being too broad and just focusing on the quantity.
Thor:
And how do you see the timeliness fitting into this idea of quality?
Victoria Gnatoka:
Well, first of all, I think real time data and real time insights is not there. We're not there yet. We always try, we always talk about it, but the reality is we're simply not there yet. With this concepts that I've described, I think the timeliness, you really can actually shorten the time from the time the data was produced to the time the insights was produced. And the reason I say that is because traditionally, again, if you have all these data sources sort of sitting in different silos across organization, I've observed it can take anywhere from two weeks to two months to get from the time the data was generated to the time your inset was actually produced. Because it has to go to all the systems, again, all this processing. Sometimes, again, if it's lived in a different part of an organization, you might not have access to it. So by the time you have an access.
So I think what you actually do, you again, you will not have it real time, but you will have it as real time as possible. And another really big unlock was having this omnichannel data approach when we think about timeliness of insights in this scope is that actually you will realize when you have it all together, there are data sources that actually are real time. And if you action on them quickly, you can have a very quick turnaround. So think about if your customer right now, as you and I speak, your customer has an issue with your brand or product, and they potentially are having now a conversation with that chat bot or with an agent. Seconds later, you already have transcript as a data input on your hands that you can look into and pair it with other transcripts, look at the trends, and produce an insight. So you're really having that knowledge of how all this works, you can speed it up. Again, it's never going to be real time, but you can get as close to real time as possible and close the loop with the customer faster.
Thor:
Victoria, I think you've shared some really insightful learnings with us today. And if you had to summarize, what's the one big takeaway you want listeners to get from this episode?
Victoria Gnatoka:
My main key takeaway is do not ignore the unstructured data and break the silos to focus on this omnichannel approach. Again, by 2025, 80 % of your data will be and probably already is unstructured. You cannot simply collect it or have it out there and not act on it.
Because to provide that good customer experience and great experience with your products and brands that's really the way to go about it. Otherwise, you probably should consider other ways of what type of data you want to have if you don't action on it.
So I would say embrace that. Of course, embrace the AI that's out there today to help you set up all these processes and automate them and make your insights team more empowered how to create that knowledge with an organization based on the data you have. I don't think we have lack of data, but I also don't think that we have too much data. I think if you create this omnichannel approach and again kind of identify your blind spots and remove the unnecessary, you actually end up with this good amount of data that I'm sure your insights team will appreciate.
Thor:
Wow, this has been such an amazing conversation, Victoria. Your perspective on insights is truly noteworthy and I think we can all learn from it. Now, before we end today's episode, I'd love to return to some of the moments of our conversation that really stuck with me. When I asked you about the definition of an insight, you told me that in your experience, insights mean knowledge. As an insights leader, your job is to make sure those insights are not trapped in a PowerPoint. It's much more than a simple recommendation. An insight is a deep knowledge about a specific topic. When we spoke about unstructured data, you told us by 2025, 70 to 75 % of all the data we will receive will be unstructured. It's a gold mine we've been sitting on for quite a while. Let's not make sure not to avoid it. If you're at the beginning of this journey, remember that it requires a commitment, dedication. It cannot be somebody's part -time job. Two. Ensure you have the resources to pursue it. And three, be very clear about the exact business problem you're looking to solve through it. Lastly, do not ignore the unstructured data and do everything you can to break the silos to focus on the omnichannel approach. Now, I know that I've learned a lot from talking to you today, and I'm sure our audience has as well. Thank you so much for joining me.
Victoria Gnatoka:
Thank you for having me. It was a pleasure.