April 9, 2020
Five trends that matter in Big Data for travel –  John Carney

Five trends that matter in Big Data for travel – John Carney

So it’s great to be here today and really looking forward to meeting you all in person over the next couple of days. And it’s been about, I think, around 15 months since our last conference in Dublin. And it was around that time as well when my journey at open door started. And so I thought maybe a good place to start today really is, you know, to look at some of the trends that I’m seeing in the data space and the data science base that I think are particularly relevant to travel retailers so let’s get started. So really this is the first trend I’m seeing this is you know, our first observation data is now really the primary driver of success in e commerce today so curated product discovery personalization 24 Seven recommendations are really now a normal user experience on so many e commerce platforms that we use today from Amazon in the US and Europe, so Alibaba in China, and of course, flip cards. Now in India, and this was not always the case in the 1990s, ecommerce was much more transactional. So it was really all about utility. So, you know, I’m sure some of you like me remember eBay when it came out first, it works but you know, it wasn’t certainly wasn’t pretty. And before the internet, consumer retail was really all about brand and demographics. So data really has come of age and is now taking center stage really in every leading e commerce platform today. This is our second observation. So personalization works. Okay, first off, it actually does work in travel retailing, you know, my observation is really Is that personalization has always been viewed as something that is optional. You know, a nice to have this is not have Amazon, Google, Facebook, Netflix or booking Sears it’s absolutely mandatory and it’s now an expectation for all of your customers as well. And this is reflected not only in the enormous profitability of these companies but also their customer satisfaction scores. So in the US, the latest average customer satisfaction index score is 77, Amazon, Google, Netflix booking all details comfortably. What’s interesting here though, is that Facebook does not but I think this can be at least partly explained at least by my next observation. So data privacy concerns go mainstream. Okay, so data privacy regulation, such as GDP or used to be an arcane area something of interest only Today’s a geeks like myself, but now following you know, the very public failures of Facebook and other internet giant’s to protect the personal data of their users. The significance of data privacy and security is now really well understood by most people. And what we are seeing and what GDP or. And similar new data privacy regulations globally are enabling is actually a fundamental shift. And that is a shift where the ownership of data is now moving from corporations and private enterprises that managed the personal data to the individual consumers and citizens from which it is derived. And this actually, I think, is a very positive thing. Two important trends we’re seeing in the area of analytics over the past 12 months is the investment in what we call heart and mind analytics and also geo location analytics. so harsh really refers to customer analytics or insights that capture a customer’s immediate intent or impulse to buy and this is normally measured by tracking surgical. behavior on ecommerce sites. And there’s a wealth of technologies out there that can help you do this mind is a bit more complex. And this really refers to customer analytics or insights that capture, you know, a customer’s transactional history, their demographic profile, and other historical purchasing patterns that measure customers natural, slowly changing propensity to purchase a product. And as you can probably guess, combining both of these insights can be very powerful from a retailing standpoint, and really, I think represents the next major step in this area. geo location. geolocation analytics is also growing in importance in travel for really obvious reasons and travel. Obviously, if we can track where customers are physically, then we can make better interest recommendations and offers and you’ve also all probably noticed how Google is already using geolocation information to graze effects to improve search. results. And the fifth and final trend I have here is arguably the most important for travel retailers. So we all know that customer acquisition costs have increased steadily over the past few years. So advertising with Google is expensive. Advertising with Facebook is expensive retargeting and remarketing is expensive. And even traditional media advertising continues to be expensive. And you don’t even know if it’s making a significant difference to your bottom line. So in this environment, analytics like customer lifetime modeling are growing in importance. So this is because these sort of analytics can determine Hamish to invest in a retargeting and remarketing or even an email campaign for a specific person. If customer lifetime value is high, then invest because you will earn that investment back in the future from that customer. If it is low. Well then don’t invest good money after bad. So these are our top five transfer data. But is there a product for travel retailers that responds to these trends and delivers a solution as the box but of course there is and it’s called t data. So we promised you this at our last conference, we hired a world class data engineering and data science team. And we’ve built us t data is on general released or customers since the end of June of this year. And we are already working with some of you here in this room to implement the full solution. So really exciting times. So it was important when we design t data that the platform enabled business agility and gave your technology teams the flexibility that they need to deliver against their own roadmap and priorities. With this in mind, t data can be implemented in phases starting with T data acquire which acquired Your customer data from multiple sources. Once this is delivered t data warehouse can be implemented, which integrates all of your customer data into a scalable and robust Big Data Warehouse in the cloud. Once your warehouse is in place, you can visualize, manipulate and interpret the integrated customer data and deliver this data to downstream systems via the T data API’s. And finally, if you want to unleash the power of this data for personalized travel retailing, you can implement t data predict which deploys customer insights part by machine learning and 2019 our investments in T data will continue to data will will grow to become a product family solving the difficult data problems every travel retailer encounters as they implement a customer centric strategy. So this includes a new tool for identity resolution which we’ll talk about in much more detail later and you blueprints or industry data model for travel, retailing to accelerate on premise data warehouse bills, where that’s the most and YouTube for online prospecting that captures and tracks the short term buying intent or impulse of customers. So this is the heart and the heart and mind analytics I mentioned earlier. And finally, a new tool for tracking and managing geo location analytics for customers all delivered following the privacy by design principles of GDP or, and similar regulations. So the future is bright for t data and its product family. But, you know, let’s take a step back here. What do we have in T data today that’s exciting and important to you as travel retailers. What’s you know, our secret sauce? Well, let me explain this with a scenario. So remember her two or three slides back she is about to purchase a flight hotel or maybe even a package you’re traveling retailing websites. So wouldn’t it be great if we knew three things? Number one issue a loyal and frequent customer or a marginal customer that could turn where we can answer that with teenagers clustering algorithm? Number two, if we give her a discount now will we earn that investment back in the future? Well, we can answer that with T data is live scores. And number three, which type of hotel she most likely to purchase? city center beach countryside, Five Star Force star, three star, the list goes on, as you know, well, we can answer that with T data’s propensity models. So these questions and a wide range of other questions that are really pertinent to travel retailing can be answered by T data is machine learning algorithms and models. But how does all of this work? I mean, how does all of this work mathematically? statistically? Let me take you On a brief journey through the world of data science to explain, so I’m going to go a little bit geeky here for a minute, but bear with me. Hopefully, it’d be interesting. So let’s start with clustering. So here we have an illustration of our clustering algorithm at work. And that is training itself. OK, so the mathematics underlying our clustering algorithm are pretty complex. And the data engineering required to do that scale across millions or 10s of millions of customers really sophisticated, but the concept is really quite simple. So essentially, what we are asking the machine learning to do here is to find a natural clusters or segments of customers across multiple dimensions. So each white dot here represents a customer and this example, there were three dimensions we have recently a purchase frequency or purchase and average spend. So here you can see how the algorithm finds these clusters by assertively seeking at the best cluster centers across the world. Dimensions once these are fans, the the average characteristics of the dimensions for each cluster can be used to label the clusters to provide business meaning pretty cool. Next up is t data life scores in this algorithm distributional or mathematical assumptions are made regarding the purchase purchasing behavior of very large populations of customers. And these assumptions are then plugged into a model that predicts the probability a customer would purchase again and the amount they will actually purchase. It’s quite valuable information. This is really quite a sophisticated algorithm. Not only does it predict Hamilton individual customer will purchase in the future it actually varies his prediction based on the pattern of purchase for each customer. So in this illustration, which is based on real data for a real customer, you can see how the probability of next purchase represented by the blue line here Actually decays differently depending on the purchasing pattern of the customer. So it’s pretty sophisticated stuff. So next up is propensity. And I think you’ve probably heard this word many, many times. So, let me try, let me try to explain what actually happens under the hood. So propensity predicts the probability a customer would purchase a specific travel product or category of travel product in the near future. And it works by first identifying the factors that are most likely to drive the outcome across all customers. For example, if we are trying to, to predict the propensity that a customer has to purchase a Beach Hotel, then things like the proximity of their home address to a beach is actually quite important. So once the optimal input factors or features are determined, a model is the inbuilt. So let’s take a look at this in action. So here’s an illustration. Of how one of these models is actually built up, you will probably recognize this as a decision tree. But in this case, the branches in the tree are being built by our machine learning algorithm. So eventually, the algorithm finds the optimal tree in terms of prediction performance. And really, this can be quite complex. And to avoid memorizing the training set it was built on we actually build a forest of trees, they’re all slightly different and use majority voting to set the probability so this is quite an established method in the field of machine learning. And it’s following the principles of like a committee of experts and here it is for specific customer their propensity score for purchasing a Beach Hotel is 92%. Okay, so look, it’s great news that you know, these deep customer insights are now available to you as travel retailers as the box and without the need to hire and retain expensive data science and data engineering teams. Our goal really is to democratize these technologies for travel retailers and make them accessible and easy to use for all business users. But ultimately, the key question is, how can these insights transform your retail business? The Tech is great, but what about your business? And I look there’s many studies that have examined this and some of these are focused specifically on the travel industry. So they tell us that customer centric retailing strategies and personalization deliver in almost every case. increase conversion rates through more targeted personalized offers more effective marketing through personalized messaging, increase revenue per trip through more cross sell, powered by insights and propensity and improve repurchase rates powered by increased satisfaction and loyalty. So this is really where every travel retailer needs to be today. And you can just look at Amazon as an example of this in action in other areas of consumer retail. So data is really the foundation. It’s in some respects the glue that holds a customer centric or personalized retailing strategy together. But remember, it is all connected. And Karen mentioned this earlier, as did Brian and Brian spell the name differently, though platforms like t retail, anti social, as well as your email campaign and programmatic advertising tools are needed to action, these insights so I’d open Joe we know this. So we built TJ API’s with API’s that hook the platform into the broader retailing ecosystem. Here’s what I mean. So this is an example of TJ working together with T retail and digital experience. We have T data on the left generating customer insights. And using the T Data API is these insights are delivered to x distributor in TV. So x distributor triggers business rules that can generate real time personalized offers. And finally, with digital experience, the user experience is personalized. So it varies for every logged in user based on their team data clustering profile, their tea data live score and there TJ to propensity scores. So this truly believe is the future of travel retailing. So that’s it for me. Thank you for your time. And I believe it’s now time for coffee.

Leave a Reply

Your email address will not be published. Required fields are marked *