In Data We Trust

Data sits at the heart of CSPs’ trusted relationships with their customers, but the secret is using it effectively. Reed Peterson of DataStax discusses what CSPs can do to build trust, overcome common hurdles, and really maximize the value of their data.

Below is a transcript of this podcast. Some parts have been edited for clarity.  

Michael Hainsworth: Mobile operators are sitting on a treasure trove of data about their customers, from location to surfing habits to dropped calls and everything in between. But cultural, technical, and regulatory silos within the communications service provider can prevent them from leveraging this information to the benefit of the enterprise or retail user — and ultimately the CSP itself. Reed Peterson knows this. As the SVP of telecom strategy at DataStax in Atlanta, he uses Apache Cassandra as a system to help large customers interpret an insane amount of data to do good for both its client and consumer. This at a time when trust in telecommunications companies is at a five-year low. He believes telcos should be among our most trusted brands — and he attributes that to a lack of understanding on the part of the user — and the CSP.

Reed Peterson: For years, MNOs have had a significant amount of user data, right? Including locations or behaviors, social networks usage and spend, and they have taken great care to protect and secure that, not to sell or divulge that information. For example, they know how much user spends on voice or text or internet connectivity and other services. Now, there are regulatory and other reasons for that, but the fact remains that for years, telecoms companies have held user data privacy and security as paramount, which should engender significant consumer trust, I think.

MH: It hasn't. Why hasn't it? The concern seems to be that the CSP doesn't have the best interests of its end customer at heart. That it's more focused on itself. How do we square this circle?

RP: Yeah, I think that's a great question. I think a lot of times we have a perspective that something that costs us money isn't necessarily something that helps and benefits us. You look at number of the OTT players who basically sell the data, for some reason we have a lot more trust of them. Now, that's changing with some recent challenges and issues that have come up, but for the most part, I think we've kind of viewed CSPs as just utilities and maybe not nearly as much trust as we should have in them.

MH: Telecom companies believe trust is rooted in network reliability, but the customers looking for evidence, the CSP will do the right for them and the planet. As we sort of see that shift as well. Is changing this perspective a matter of changing culture, and if so, how do we do it?

RP: Absolutely. I think transparency is paramount here. I also think creating more of a community, the customer wants to feel involved. They want to feel like they're part of something bigger than themselves. They want that company that they work and associate with, as a CSP, to represent their values and their perspective. I think telcos have an opportunity to become a clear steward of data privacy and protection. Consumers made it clear, they want to understand and feel good about how their data's being used, and CSPs that are true stewards of that consumer data protection and privacy will be clear winners. I think really it does come down to transparency, building a community, standing for things that your customers want you to stand for. I think those are all critically important in building and engendering that trust.

MH: You mentioned the remarkable amount of data that any given telecommunications company acquires over the course of a day, let alone a year. Those call data records, that are so prized as something to secure, tell me about how we leverage that though, so ensuring that a CSP is doing the right thing when it comes to that customer interaction, leveraging that data for good.

RP: Great question. Here's the thing, an operator that serves eight million customers generates about 30 million call data records a day. That's about 11 billion records annually. If you take a much larger mobile network operator, a lot of them in the US, are around a 100 million subscribers, that's 375 million call data records per day, 370 billion per year. That is a significant amount of data. What do you do with that data? How do you treat that data? Well, this is where kind of digital transformation and using machine learning and artificial intelligent data models, to use that data, to develop these kind of models around how you understand customer data and customer engagement.

With all this data, you know what the customer's doing, you know where they're at, you know what they're, all the different aspects of their engagement with you. What do you do with that? Well, you need to develop these machine learning models. Those machine learning models are really to figure out how to ingest that data, process that data, train that data, use the data models to ultimately give more predictive capabilities to serve the customer better. It really has to come down to the customer. Everything is about how you engage that customer at the point of interaction at the edge, and 5G is going to push more of that. I really think it does come down to engaging that customer, using that data to better understand, better interpret what they need and respond to their needs or their challenges.

MH: I suppose when you throw machine learning algorithms and artificial intelligence at that insane amount of data that's generated on a daily basis, you create algorithms that aren't just designed to help the customer when they have a problem. You could use that as predictive analysis to indicate where things seem to be going wrong and address them before they hit the fan, so to speak.

RP: That's absolutely right. I mean, machine learning, it's basically the process of feeding data into any number of algorithms, like decision trees or logistic regression, linear regression, ultimately to make better decisions. Companies are using machine learning to forecast the future, right? Prices and ratings, looking at the weather, they use it to detect aberrations: fraud, intrusions, disease. You can use it for classification, right? Facial recognition or categorization, spam detection. We use it a lot, with a lot of our customers on recommendation engines. Netflix, or a content provider, how do you provide a hyper personalized experience in what they're interested in based on that data? I mean, machine learning really is not rocket science. There's no magic. It's just billions and billions of rows of data and two buckets of math, a little bit of linear algebra and some multivariate calculus.

MH: The problem though, is that data is siloed. If you want to build trust, but still comply with data privacy regulations, like GDPR, how do you protect the customer's network activity, something that's critically seen as a serious privacy issue, how do we rethink that practical role that data plays in building trust in a customer first CSP?

RP: Great question. Oh, well, that data needs to be protected, right? It needs to be secured from all aspects of the network. The second thing is you need that data, wherever the customers are, to be able to engage and interact. As you look at the different forms and functions of digital transformation, you're looking at 5G enabling a significant amount of data. As all that data goes out to the edge of the network where you separate compute and storage, you need to be able to access that data and replicate that data intelligently from core to edge. Having said that, there are certain things that can't be replicated or shouldn't be replicated, and that's why you need intelligent replication. From our perspective, that's really Cassandra's strength, right? It really does come down to a distributed database with multi-region clustering capability and intelligent replication, so that you're replicating the right data for the right environment and application rather than kind of blindly duplicating data everywhere. Like I said, this is an incredibly important for security, regulatory issues, GDPR, et cetera.

MH: I wonder though, and I know that companies in the social media space are getting a lot of flack lately for the role that they play in every aspect of our lives now, but as people we're sort of accustomed to giving up a certain degree of our data privacy in exchange for something. Do these social media sites and others that monetize data, give CSPs some wiggle room to use customer data in a more flexible way?

RP: I absolutely think they should. Years ago, at the GSMA, we made a big push on same service, same rules, right? [Reed Peterson was SVP of strategy at GSMA between 2007 and 2020] What we were trying to do is look at the rules and regulatory regime of the OTT players and what the regime that was set into place for telcos, and it's very, very different. Our perspective was, if you're offering similar type services, there should be similar rules, not more rules, but do it the same. As you look at it, the telcos have so much data that they have not used, and for good reason. Again, going back to the very first question, that should engender trust, having said that, I think they can use a little bit more of that data to benefit the customer.

Now, if you're selling someone's data to make a buck, that really doesn't engender customer trust. That's not the transparency that we were talking about earlier, but if you're using that customer data to give them the power and to put the power of their data in their hands, I think that's where it gets a little bit more interesting. Shift the data power to the consumers so that they get to control what happens with their data, how it's used and ultimately, so that it benefits them to what they need. Do they want recommendations? Great. Let's give them the best recommendations possible. Do they not want to be bothered? Great. Let's not bother them. Whatever we can use that data to better serve those customers I think is the best way to do it. I do think that social media companies have allowed the perspective of that to open the door a little bit more for telcos.

MH: Well, then, let's talk about how we go about that. You touched on intelligent replication of data. How does a CSP share data securely across its departments that ensures the customer, whether it be enterprise or retail, or what have you, get what they need when they need it while still meeting those regulatory burdens?

RP: Great question. I mean, the easy answer is Cassandra and data stacks. The broader answer is that you need to bring all of your data properties into one place. You need to at least try to do that and, so, you have data housed all over the organization and some of these behemoth organizations, it's a significant number of databases and data lakes and Hadoop clusters, and so many different things that are happening across that organization. You need to have a unified data strategy. Part of that comes with getting all the right people in the room to brainstorm about what you actually need that data for. There is some data that will need to be on-premise. There is some data that will need to be in the cloud, everybody's moving to the cloud, but telcos, by their inherent structure, they're infrastructure providers and they provide this communication service based on infrastructure.

So, because it's infrastructure, there will always be an on-premise need for data. Having said that, the customers are out in the world, out at the edge. A lot of the other data needs to be able to be in the cloud. We look at it as is a hybrid cloud perspective, on-prem and in the cloud. Then, we view it also as a multi-cloud perspective, where no responsible executive team is going to put all of their workload in one place. It just wouldn't be smart. You're going to see this multi-cloud, hybrid cloud world, where that data does need to be replicated. Companies need to come together and look at their comprehensive strategy. Right now you have individual groups within those companies that are developing the strategy. That's fine to some extent, but that has to roll up into a broader organization-wide data strategy to be able to use all aspects of that customer data from each different department to benefit that customer at the edge of the network.

MH: Okay. Then we've broken down the silos, we've established, which data can be shared amongst departments securely while still complying with GDPR regulations, etc. When solving a customer problem, we're advised to provide personalized solutions. The largest mobile operators in the west have as many as 470 million customers. How do we accomplish that effectively and efficiently at scale?

RP: Another good question, Michael. This one's interesting because you're right. Most of us, we have CSPs for one reason, right? That's connectivity, but we all have slightly different needs for that connectivity. You run your podcast and you need significant bandwidth at certain times, right? As you're going through all of this. I may be traveling a lot. I may not need nearly as much bandwidth, but I do need flexibility and coverage. Everybody needs something slightly different. Whether your kids are all homeschooled, on Zoom calls, or you watch a ton of Netflix, or you can't have that buffering, whatever it is you have certain needs and the data will tell the story. Whatever the data tells you about what the customer needs, that's where you go. As a customer, if I'm having an issue and I jump online to contact my CSP about my issue and that CSP, they get a customer service rep and they get on the phone, they say, "Hey, what's your problem?" I start to explain my problem and only then do they start to look into my problem.

Then they've got a problem, right? If, by the time I get on that call, they've already seen all the data that's come up to that customer service representative that says, okay, so Michael had an issue. The issue is that he joined a new service and that new service hasn't been working, I can see that he's had a number of dropped calls over the last two, three days. I can immediately get on and say, “Hey, Michael, I see that you've had an issue. I apologize for that. I see that the service isn't working as well for you as we would both like, let me offer you a $40 discount credit to your account right now. Let me take a look at that. It also looks like, I see over the last couple years, you've had this much data usage. I think you should bump that up. I see your needs. I see who you are as a person from that data. I'm giving you specific actionable items that I can make you happy and engender trust.” Ultimately, it's about transparency. The data will tell the story.

MH: I can imagine as well that if you've built your data system effectively and you have an efficient and effective machine learning algorithm, you may very well know the customer's problem before they experience it. The best systems would even call the customer. The customer wouldn't have to call them, and recognize where the problem lies and then offer solutions. How do you do that though at scale, how does a CSP scale up its expertise in machine learning when this is a brand new avenue for this industry?

RP: Yeah, I mean, and that would be such a phenomenal response, right? I'm having issues and the CSP calls me and says, "Hey, this is the issue you're having. We've already solved it. We've sent a truck to your facility, your Internet's down, we've already got it. We're covered. You'll be up in 10 minutes." Before I even have to do anything, but, again, it all does come down to the data. How do you do that? How do you do that with so many people? You collect as much data as you can, and then you have to figure out how to train the models around that data. Again, it's ingesting all the data, collecting data from all aspects of your business, right? That can be IoT sensor data. That can be processes. That can be people. Data can be simple as temperature humidity, or it can be a full video feed, whatever it is you need to then process that data.

Then you need to train the data. Algorithms train the machines on actions that need to be taken based on the data, or train the customer service bots on what needs to happen. Then you develop the data models, the prediction models for patterns and recommendations and predictive analysis and 'what if' scenarios. Ultimately, that is going to provide a better customer service experience, but it really does come down to data. Now, we think that the best way to do that is through partners, right? Develop partnerships with data companies that can help you do that. It's a really hard journey alone, especially if you're just looking at massive amounts of data and don't know what to do with it, which is a lot of the position that I think CSPs have been in for a while. That's my view, collect as much data as you can, find partners that can help you analyze, collect, and intelligently replicate that data. Then, develop the models around it to better serve your customers, but if you keep your customers at the core, they'll tell you the story and the data will tell you the story.

MH: I can imagine as well, because any decision that gets made in any corporation has to be based on return on investment, the idea of partnerships where you may be weak or lax in one particular area allows you to monetize this remarkable amount of data that you've got, not only to ensure that you've got better customer service, but that it also creates a new revenue stream for you.

RP: Yeah, absolutely. I mean, and 5G is going to change everything there, right? I mean, as you look at the G's, you've gone through 1G, 2G, text messaging, 3G, mobile and wireless connectivity, 4G, more kind of cloud IP, mobile broadband, 5G is kind of this unlimited data capacity world. Where its significant speed, significant throughput, very, very low latency. That's going to change so much of what you can do out there at the edge of the network. The 5G applications of the future are going to be different today. The same way that the first generation of video game consoles were different from PC games on 17 floppy discs. This ultra-low latency and ultra-wide bandwidth doesn't mean just faster response and higher resolution images. It means access to cloud computing resources and transparency to the user, applications that follow the user as they move, so vehicles, drones, so many different use cases for this.

MH: Well, you point this out, that's basically industry 4.0, there are so many more moving parts in 5G, but that gives us the ability to create this fourth industrial revolution. What are some of the best use cases for predictive analytics when it comes to the factory floor?

RP: When I used to live in the UK, I used to have my groceries delivered. I always thought that the person who was delivering my groceries would basically go and go to the store, pick up my goods and then bring them in the bag to my house. Well, really what happens is you've got this factory that's just right outside of London and a swarm of bots running on one of the world's densest private 5G mobile networks, and it's revolutionizing warehouse pickup and fulfillment. You have all of these bots that are programmed and using machine learning and, in some cases, artificial intelligence, to maneuver around this factory, collect the items and do it with amazing speed using cloud data emerging tech and really this also comes down to sustainability, right? It allows you to function with more effectiveness and more efficiency. I think it's a great thing.

MH: In a 5G world, CSPs are being urged to partner to fill in gaps and institutional knowledge, as we've been discussing, in areas like machine learning algorithms and IoT. How do we securely break down data silos, not just within a CSP, but break down those data silos so that partners can use that data?

RP: You need to be able to have that data always available and easily replicated so that people can see it wherever they are. You need to be able to give access to the partners to the right data in the right places. I do think that, as you work with these partners, you'll need to break down some of these silos. That does come with, my recommendation is always get in a room, get the right people in the room, brainstorm, have the conversation, talk about how you can use data more effectively, more efficiently, ultimately to better serve your customers, but also create significant operational efficiencies across the organization. Most of the partners will know that and they'll understand that. As you work with them, try to figure out ways to benefit the customer at the end of the day and the data will tell the story.

<< Go to previous episode