Creating a 360-degree Customer View Using Predictive Analytics and AI

Welcome to another Prime TSR podcast. I have a special guest, Geremy Reiner, who is an Associate Principal at Prime TSR. Welcome, Geremy.

Thanks. Great to be here, Rob.

And today, we are talking about predictive analytics and AI, and you have a lot of experience with your professional background in this. I'm really excited to have this conversation with you, and I wanted to start with a question to give our audience; sort of help them visualize predictive analytics, and a lot of it has to do with personalized customer experiences. But explain that to me. What does a personalized customer experience look like in real life?

Absolutely. I'll give a couple of examples to start with just because there are many technologies that we can dig into, and they don't really speak to what that personal experience looks like for a customer or client, or even our internal employees. One of the best examples and many people actually use these services. If we look at somebody like Netflix, when they started, a large part of their customer base was scrolling through hundreds, if not thousands of potential videos to watch, and movies, and TV shows. And Netflix needed a way to focus a lot of their customer base on what was important to them. They made a choice, a conscious choice, to provide targeted specific information and allowed their customers to find what they were looking for without going through multiple steps. They also wanted to make sure that that user base had an easy way of integrating their system into their daily lives. One of the things they looked at was, well; they have all of this information when somebody is interested in, let's say, Saturday morning cartoons. They need to be able to tag everything that is a cartoon and then present that to a segment of their audience. For the average layperson, that's a really simple concept. All we have to do is make sure we have a piece of meta information on a video that says, "This is a cartoon". But as you start getting more and more data and you want to get more and more personalized, it's not about the grouping of people, but actually tracking that data for one specific person, so that when they log in, and it's a Saturday at 10:00 AM, they want to see that at the top of the list. But if it's a Friday night, you may be looking for a romantic comedy or an action-adventure movie. And Netflix wanted to get to the point where they could take all of this information about demographics, and personal information, and search history, and even watching history, whether or not somebody finished an entire movie or only just watched the first half an hour of it, to identify what their likes and dislikes are, and really create that perfect profile, so that everything formed together to create that perfect experience so that when you log into Netflix, you get a list of videos that might be the most important to you, and they took it to a point where it wasn't really about hard and fast rules. When Geremy logs in, he sees these videos. They wanted to make it so that it was more intelligent, all based on algorithms. And that's really what we're talking about when we talk about predictive analytics. When we talk about the personalized customer experience and how we build these things within our organization, it's all about taking that data. The user experience that we present to them today, analyzing how they interact with our systems and how they interact with our organization and our people, taking all of that data and forming a complete picture, and then feeding it back and having an endless loop between our customers, our employees, and our systems to create that picture and that perfect experience so that our users get what they need, get what they want. We can also provide them with additional services based on our experience with their data.

That's incredible, and I think you've just explained why I'm addicted to Netflix. It's a good segue into the next question, because they've essentially predicted what I want to watch with extreme confidence. It's a pretty good analogy as far as precise customer experiences. But you said something interesting about predicting. Predicting is a big part of, obviously, personalized customer experiences because you're predicting what the customer wants, so explain how predictive analytics works and what are some use cases of predictive analytics?

Absolutely. So, I mean, really, when it comes down to it, predictive analytics is an entire family of services. We're talking about at the base level, taking data and looking at that data, creating relationships within it, and then using some fairly standard, and I say standard by the IT term, and that's information to the analogy, of creating some basic patterns around how we're going to take that data and then use it to create what would really be just a response, and some of the methods that we take to do that are varied. Still, we really look at information retrieval, translation, and pattern recognition through to sentiment analysis. If I kind of touch base on each of these, and when we look at information retrieval, we're really talking about pulling data. So, when users log in to our system for the first time, they log in and register with their base information. First name, last name, maybe a little bit of information about where they're located, a phone number, and then we can take that, and we can do some really interesting things with information retrieval. We can actually do a search and say, "You know what? They're on Facebook. Would you like to add a Facebook connection into your profile information? Or LinkedIn? Or any of the other social media platforms like Twitter, et cetera", to create a more robust identity for those users, and that translates into pattern recognition and audience segmentation that I was talking about. When we really talk about sentiment and audience, we're trying to take all of these information pieces and put them into a bigger picture to understand better what our user is really trying to consume from our systems. Some prime examples of this are some of the biggest organizations in the world. Amazon is probably at the top of our list right now for somebody who does integrations that pull data from many sources and then provide it back to you. So, when we log in to Amazon now, it's a lot different than we did when it was back in 2004, 2005. As a note, Amazon started its Amazon Prime service in 2005, and the idea there was to actually bring in all of their user bases and have them continually use the service. That was the whole idea of customer experience for them. Amazon wanted to tie the user experience back to the user's perception. So, they didn't want their users to go out and say, "You know what? Our competitor has something for 3 cents cheaper. So, we're just going to go with them". We wanted to give them an experience that continued to have them come back to the system. The easiest way to do that was to tie it in with everything they were trying to do. So, we now have an Amazon that provides you with almost anything you would want to buy online, and the way they did this is they looked at it as much like any other organization would do. They took that initial identity, and then they bring in, "What is their feeling on a brand or a product?" Sentiment analysis. "Do they like/dislike something?" and those are the first borders that can be created when we start looking at predictive analytics. You can easily create likes and dislikes and tie that back to the larger story of what they need and what they want from your organization.

Yeah, and that's really interesting, and I think those are some good examples, and I'm wondering, even for companies that are not Amazon or not as big as Amazon, and they're exploring these things because one thing I've noticed is that these are all possible now. I mean, there used to be technology barriers, like the number of servers or number of engineers, et cetera. With the cloud and things like that, many of these tools are here, and a lot of it companies are exploring and just haven't taken some of these first steps. Could you walk me through, if you were a company, what are the technical considerations and challenges when you're creating these personal experiences?

As a note, customer experience is something that is not just for large companies. We see this with every organization that we can integrate with. Kind of the first entry into creating the first platform is going to be all about data. Are you getting the information that you need to start deriving real results out of your user base? So, doctors' offices, or even as simple as our pharmacies that we go to that are fairly small, can take advantage of this. Every mom and pop grocery store in the last 30 years has been doing some sort of predictive analytics. They just haven't been automating it, and that's really what we're talking about when we talk about predictive analytics. It is creating some sort of automation within the technologies that we use to derive these things so that it's not just people doing it. It's actually our technology providing us with additional information. And when I say that it made mention of mom and pop grocery stores doing predictive analytics for 50 years even, and that's because every grocery store out there needs to know what to stock on their shelves for the customer base coming through their doors. We do this instinctively as people.

Interesting.

We analyze everything around us and make guesses about what somebody is going to expect from us. The big piece that we're trying to do now, and it's really an amazing transformation in the way that companies run, is that we're trying to leverage the data and information that is out in the world to create better predictions and provide that back to the people, so it's not just gut instinct, it's data-driven decision-making. And the easiest way to do this is to start collecting the data that we already have. Identity information, buyers' profiles, and even basic information about where our users go to on our websites or how they interact with our brick and mortar stores, and then taking all of that information and creating what everybody calls algorithms, which is really just a formula to say, "Based on these inputs, we're going to present this as a decision and say to our customer service representatives, 'This is what we should be doing next'". So, you can take all of that data and ingest it, and if we're looking at the cloud, and you take that, and you put it into a big, basically unstructured data source. That then, we can take some other technologies and do predictive analytics, some basic pattern recognition, and data mining to create some sort of value-based and scoring system to say, "This is what this user will probably need or want in the future". This isn't a large task. We do it every day. I mean, honestly, go to our offices every day and know that our boss expects certain tasks from us. If they haven't told us what the tasks are for the day, we make intuitive leaps in our own daily task list to say, "Our boss is probably going to expect these things". Now, if we can take all of the data, of everything that they've asked for, for the last year from us and assume a few basic principles. It would be pretty easy to create a formula to say, "You know what, on February 2nd, this year, they're going to expect a report on something". And that's really what we're trying to do every time we look at our customer interactions as well. Every year, I buy filters for my furnace. So, it's not really outside the norm for me to go onto Home Depot's website every year around the same 30 day period and order the same filters. Now, if we can take that data and actually preempt it and send an email and say, "Hey, last year about this time, you bought a filter. Would you like to buy it again?". It's an easy way of re-integrating with the customer base that may not interact with us on a normal basis, and that's really the first step to getting into predictive analytics for any organization. You take the data you know, and the timelines you know. You can create some really interesting interactions for your customers based on that data.

Yeah. That's really interesting, and I like your analogies, and I've actually never thought of it that way about mom and pop shops have been doing it. We can predict what our bosses or managers will want, even if they don't tell us. Predictive leaps. We don't know for sure if they're going to want these reports, but we can do it with some level of confidence that this Tuesday at the end of the month, I think he or she would expect it". So, I liked how you put that because it really puts a human touch on us, something we've already been doing. Now, we're just systematizing it.

Exactly. This isn't really a new concept. It's a new approach to the same concept that is realistically the human condition. We always try to predict what the future is going to be or what's going to be there in the future, I should say. The key, though, is really that now we have the tools, and it's always a great term to use, but new tools are being generated all the time by these cloud service providers. AWS, Azure, all have toolsets for managing what we have called in the marketplace, at least big data, and big data doesn't need to be extremely large sets of data. It can be as simple as three lines of text. It's all about the approach of using data to make better traces.

That sounds great. Obviously, everyone should be doing this. Like, "Oh, you know what? I should be getting off the phone right now and just building predictive analytics for my children". But if someone comes up with that one, I'll gladly pay for that one because that's right now, that's still mysterious. What are some of the challenges that companies run into when they want to create…? Why isn't this a bigger thing? I wouldn't say it's in the hype level, but it's real. But what are the challenges that companies have when doing this, and how do we resolve those?

Yeah. I mean, the hardest challenge is really the data. So, we talk about the amount of data in the world at this point, surpassing our ability to actually filter through it all. There is a lot of noise in the data as well. Some of the biggest obstacles to enter into any sort of predictive analytics system will be figuring out what data is important. Identifying that master data model, and identifying what's really going to be key to this transformation that you want to make into your organization, and really, it's going to be the normal technology learning curves. These are new tools or newer tools; I should say because they have been around for a while. But the entry-level is really the learning curve about how to look at your data and how to use unstructured data when for so long, we've focused on normalizing every piece of data that comes into your organization. It's all about the relationships in the data and strict one-to-one or one too many relationships to that data point. A large part of these projects ends up becoming; how much data are we actually going to look at? What's the important data set within that? How do we take that raw data and turn it into something that is good business domain data? Something that's going to drive value either internally or externally, and then finding a way to use that to actually create some of these analytic routines.

And walk me through who is responsible for this? It started with very big data. It was a very big technical word when it began; I guess we would say back in the day. And now, it's not just a technology thing. It looks like everyone within the company, including marketing, are doing it. So, who owns this? How do you structure this process to get the data in the right place? Who owns this process?

So, when we're talking about data, it's typically marketing and IT. But the real conversation here is really about the customer experience, and one of the key tenants to customer experience and how we integrate with our internal and external customers is that it's not just one team or one person's responsibility in an organization. When you start making transformations, like putting in predictive analytics and focusing on that larger customer experience for how they're going to consume your services, or even just provide data to your services, it really becomes the entire organization's responsibility. And in a real way, everyone becomes a part of the chain. From your frontline support and customer service representatives, all the way through to the executives and everyone in between. Every voice becomes another piece of the dataset that feeds the machine, which is going to give you better and more refined analytics. Part of the reason it's called big data at this point is because it's a big job. It's all about taking every single piece of data that you can, and then refining and scaling that data, and then creating the tools around it to create real interactions with your user base. Not just reports and dashboards, but taking all of that information throughout the organization and enabling each and every person to feed that back to your consumers or to other systems to make better use of the information itself. And it sounds really circular because we're really talking about everyone having to do this. Who's the owner then? And the owner really becomes someone new. And sometimes, it's just the marketing organization that takes on the responsibility for managing the larger vision. Sometimes it's the IT organization to ensure that the data stays consistent. But realistically, it still becomes everybody's responsibility, and everybody has feedback into the process.

That's really interesting. Is there anything else that you wanted to discuss or a question that you wanted me to ask?

When we start talking about predictive analytics, and when we start talking about the customer experience, we tend to try to take the view of what's best for our company, our organization, and whenever we start down this path, it's really more about creating more personalized experiences for our users. And our users aren't just our customers. They're our internal employees who enter their time into a time management system or even integrating new tools on our email systems. Customer experience, and really, it's just the personalized experience, something that is both external and internal. So, whenever we start down this path, keep an eye on making more relevant, personalized experiences, not just for your customers, but for your employees, because the more you empower your employees, the better that external customer experience is going to be.

That's amazing. Specifically, on the data parts that you mentioned, it is everyone's job, and I think that's the biggest shift I think we've all seen, is that it used to be just a technology person's job to store the data somewhere. That was the bare minimum. Now everyone's asking, "Well, what are we doing with that? And oh, by the way, you store the data that way? Why would you do that?". Because nobody really ever asked, you know what I mean? Nobody really asked what was happening with this data. I think that's super interesting how it evolved.

Absolutely, kind of a history of data. The old filing cabinet and the storage room, and now we're at the point where we have databases, and file shares, and blog storage, and cloud repository’s, photos everywhere, and everything's in a search engine. But the downfall is that it's not nearly as easy to find the important information.

What happened to the good old SharePoint days? We just put it in, we checked it in, checked it out, and that was the end of our storage.

Even before that, the internet. Everybody had one. Now, the internet idea is even beyond that. It's a concept that's morphed into this major keystone for everything that we do.

Yeah. That's great. Well, thank you so much, Geremy. I really appreciate your time. I've learned quite a bit about predictive analytics, and now, when I watch Netflix, I'm going to be conscious of what they're doing and how they're targeting me.

The "Recommended For You" section has always been interesting to look at when you start thinking about predictive analytics.

Yeah. Yeah. If they were good, if they took their chaos theory to the next level, what they could do is they would provide someone else's recommendations for you and just see how much people would flip out. But I think with just internal, not for the customers.

Actually, really great. The people like you, you actually have a tag like that. Netflix.

Yeah.

It's an amazing thing. They started doing that. That's audience segmentation, actually.

Yeah. That should be an ad for them. Do you know what I mean? Almost like in Burger King, pretending like they had run out of Whoppers. Here's the Burger King freak out. Burger King pretends they ran out of Whoppers and people freaked out. And so, Netflix can do the same thing if they provide someone else with a stranger's recommendations just to see how bad it is, and that fits the tag line just for you. Well, that's all for today. I appreciate it.

I appreciate it. Robbie. Thanks.

Okay. Thank you.