Experiential Tech: Happier Customers, Increased Profitability
JUSTIN THENG: Today's consumers not only expect a lot from brands, they also capitalize on AI, IoT, mixed reality, and other immersive and emerging tech. This puts tremendous pressure on marketing organizations to reinvent their operating models so that they can act in the moment. But if you think about today's consumer expectations, they're pretty hard to meet, and you haven't seen anything yet.
Tomorrow's level of customer experience and personalization will need to be even smarter, more immersive, and more trust enabling. The question is, are brands and consumers ready. I'm Justin Theng. I'll be your host for today, and I'm joined by Professor Matt Kuperholz. Matt was trained as an actuary, but has been practicing data science for the last 25 years, focused on AI for the last 20.
He's PWC Australia's chief data scientist and a partner in their consulting business. He was awarded by Australia's top analytics leader by their premier industry body, and one of Australia's top 100 knowledge workers by then prime Minister Malcolm Turnbull, and the Office of the Chief Scientist, and most recently, one of the global top 100 tech innovators. Matt, it's great to have you with us.
MATT KUPERHOLZ: Thanks, Justin. Thanks for having me.
JUSTIN THENG: Matt, could you for the sake of the audience, just-- I mean, I've given an outline. You've obviously had an illustrious career. Can you give us a bit about your journey to this point?
MATT KUPERHOLZ: Sure thing. I guess it started back in 1972 when I was born, at the same time as the first microprocessor was born on a chip from Intel. I don't think that was a coincidence. Instead of sleeping with Teddy bears and stuffed toys, I slept with toasters and irons. I've always been a massive fiend for technology. I still remember when my relationship with computers started in the late 70s, and has been a continuous love affair ever since.
As a child, when you're good at maths and you want to get into business, there's not too much choice back in the day. So I became an actuary. But I also studied computer science, and I was looking for a way to bring those two disciplines-- or actually, three disciplines. That's maths, computer science, and complex problem solving-- together, which I guess is what we call data science nowadays. But when I was put through University by an actuarial firm in the early 90s, we didn't know it as data science.
It's just the direction my career took first with the actuarial firm. Then jumping into an AI start up company that I was a general manager for. I had my own web development company. And I then started my own consulting firm in data science after we sold the AI firm, but quickly found myself with one main client, that being Deloitte, and started a great adventure with them, building an analytics practice in the early 2000s around the world.
I've since left Deloitte where I joined PWC as a partner, and have been on a similar journey with them. Obviously the market has changed substantially. I still remember Justin talking to clients in the early 2000s and saying, this business problem you're having, it's quite possible we can solve this by using the data that you've collected in your operations. Could you imagine when that was a completely novel idea and no one was thinking that way, to then everyone thinking that way, but no one being able to do it?
Like shooting fish in a barrel, the golden age of data science consulting, to nowadays where it's taken as a given that we are collecting and working with ever-growing-- in fact, exponentially growing-- amounts of data. And our customers and citizens expect that we're using the evidence at our disposal to best service them. So that's kind of the journey. Lots of technology, lots of nerding out. It's been great.
JUSTIN THENG: Yeah, you've been an entrepreneur as well as a consultant. And it's interesting that you said you've been on this journey for 20 years. I wish the listeners could see this. There's a visual behind you. You've actually got motherboards-- it looks like motherboards-- in the background just up on the beams.
MATT KUPERHOLZ: Yeah, I do a little bit of art in my spare time. I love the creativity merging with technology. Incredible global events like Burning Man make me very excited where you see technology as art, and I think that's one of our crowning achievements is the miniaturization that goes into electronic computing, and I find it quite beautiful.
JUSTIN THENG: Incredible. Matt, are you reading any books or blogs at the moment? How do you stay up to date with such a fast moving industry?
MATT KUPERHOLZ: Yeah, look, I usually have a couple of books on the go. I have a bunch of websites I read regularly. It's interesting with AI. The field is growing so incredibly quickly. I think I'm at the cusp of-- well, certainly I don't have my arms around all of the detail, but I just feel like I've got my arms around some rough idea of everything that's going on. But with its exponential growth, it's becoming even more challenging to do that.
However, books, I guess pretty different. I'm a big fan of science fiction, so I'm reading a book called "Cryptonomicon" By Neil Stephenson. My girlfriend's been a practicing Buddhist for over 10 years, and I've just started chanting with her, so I'm reading an introductory book called "The Buddha in Your Mirror." and I've been playing a bunch of chess through COVID and lock down, so I'm rereading a classic called "Lasker's Manual for Chess."
JUSTIN THENG: Incredible.
MATT KUPERHOLZ: And also I read every year "How to Win Friends and Influence People" from Carnegie--
JUSTIN THENG: Dale Carnegie.
MATT KUPERHOLZ: Dale Carnegie.
JUSTIN THENG: Yeah, yeah. Good old Dale.
MATT KUPERHOLZ: It's a good reminder.
JUSTIN THENG: I saw a super chessboard the other day. Somebody posted it in one of the Discord channels that I'm in, and it's four rows of all the pieces. So black has 4 kings. White has 4 kings with 4 queens. And you have to checkmate all 4. That'd be incredible to play.
MATT KUPERHOLZ: Wow. Yeah, there's actually been this movement of chess computing getting so strong that instead you rearrange the pieces or try novel arrangements of boards to mix things up a bit.
JUSTIN THENG: Now that's actually quite an unintentional but quite a good analogy for what it's like to be in marketing at the moment, or even CX for that matter. Anywhere where you're touching customer data there are so many moving pieces, and the competition always seems to be talking about the best moves that they've got. And but in reality, when I speak to CX leaders and CMOs, the general feeling is that, wow. We're really early on in the curve. To me though, it's a really exciting time to be a leader in marketing. What are your thoughts on that, Matt?
MATT KUPERHOLZ: Look, I agree. A theme that you'll see me coming back to quite a few times in this chat, Justin, is that of the exponential times we live in, which is not my idea. It comes from Singularity University, and Ray Kurzweil and Peter Diamandis, two futurists that run it. And just quickly, the background to that is acknowledging that computing power is growing exponentially. We know that with Moore's law. But also that many other things, the storage of data, the communications, AI, a whole bunch of technologies are actually exponential in terms of their performance per dollar.
And their third point is that all of these exponential technologies support and reinforce each other. So when you're able to interact with the real world through IoT which is exponential, and automation which is exponential, and gather exponential amounts of data and crunch it using exponential computing power, and then find the hidden patterns with exponential AI, and then service individuals through a combination of all these technologies, it very quickly means the ability to market, or more importantly, relate, with individuals automatically in a cluster of one that understands their needs, aims, wants, and desires, is incredibly powerful.
And it actually brings that kind of win-win situation. With exponential technology you don't have to be robbing Peter to pay Paul, but rather you can actually have happier customers and a more profitable company. And I think that's the trend that we're on. Although when you sort through all the hype, I would propose, especially in Australia, that marketers are underutilizing the technology at their disposal to really take advantage of what is possible in a data driven world with regard to servicing customers.
JUSTIN THENG: And Matt, with so much exponential opportunity, why do you think it is that marketers are underutilizing the technology at their disposal?
MATT KUPERHOLZ: So again, Justin, it's really complex to sort through all the hype. The skills of doing this are in high demand. It's usually not a single technology out of the box that will gather all of the data, but rather an experienced data science led approach that says, in your organization you have latent assets in your data. Those assets are your operational data sets about the customers or the market that you're facing into, the customers that you have, the customers that you've lost, every marketing campaign you've run, every physical point of presence or virtual point of presence, every product you've sold, every price you've changed, every product your competitors sold, and marketed, and discounted in the process they've offered.
And then you've got geospatial and temporal overlays as well as the opportunity to enrich with social data and external data. We're talking about a very complex environment. The most successful marketing analytics campaign I've ever run, which was still many, many years ago, has not been bettered in terms of the response and the value delivered was for a 200-year-old North American bank. It was the most successful campaign they'd ever run.
We used AI to consider 17,000 metrics for each of their 10 million customers. So to actually find the right offer for the right customer, we were simultaneously considering 17,000 different things about them, which is an incredibly high dimensional data set. And it's challenging to find the signal in the noise in such a large number of dimensions, but my personal belief evidenced through years of practice is that generally the more data the better. The more I know about you and are able to compare what I know about you with someone else, the better.
Which is why also in a digital world, the ability to experiment and test many, many different permutations and combinations on subsets of our customer base, and learn from what succeeds with different customers and double down, essentially A/B testing across the whole alphabet, is another way to generate and work with data and technology, but something that's also relatively nascent in most companies marketing operations.
JUSTIN THENG: With 17,000 data points I think it was that you said, how does a human trust that the outcomes, the decisions that are being made by the AI, are appropriate, that maybe the model that was used is appropriate? How do you go about that?
MATT KUPERHOLZ: So let's not talk yet about my latest obsession being responsible use of AI, which is appropriately governed, ethically sound, and the performance is unbiased, fair, transparent, explainable, robust, and secure. Let's put that to one side. Imagine I have some black box, and in this case, it's an AI that has built a model looking at the interrelationship between 17,000 variables, and we're trying to predict an outcome, which is your response to a certain campaign through a certain channel.
A very old trick, and not just used with AI, is that idea of withhold testing and cross-validation. Which is if I randomly choose for my 10 million customers, 1 million customers that the model is not allowed to see when it learns. So let's say in this case, I learned from 9 million customers. And in the remaining 1 million customers, I have some known outcomes, campaigns they responded to positively or didn't respond to.
I then disguised that from the model, rewind carefully in time to what those customers looked like before that campaign, passed it across the model, see what the model thinks, and compare it to the known outcome. So you can always use historical data, in a way, to test and validate the models you've trained on, as long as you're very careful and very scientific about avoiding something called information leakage. But essentially, to sum it up, you've got data where you know the outcome. You use that to test your model.
JUSTIN THENG: That's obviously more than just technology. That's a thinking process. How do you go about leading others, Matt, to think in this way? I mean, there is a balance, I believe, between this behavioral science gut instinct, knowing how people tick, knowing what to look for and what questions to ask. And then, like you said, the huge data sets. Now how do you go about leading a team, for example, your own, to thinking this way or thinking through a challenge?
MATT KUPERHOLZ: It's not just our team. We actually have a service now that most of our clients have analytic teams, which is uplifting their analytics maturity as well. And all of this is very technology independent, Justin. It's not about a particular approach, or piece of kit, or even data set, or industry or problem domain. The answer is, you have a standard yet flexible process, an analytics process, and we use one that's actually over 20 years old.
But whichever one you use, you'll find that the good ones all have the following in common. They don't start with an analytical problem, or a data problem, or a technology problem. They start with a very clearly defined business or societal problem. We need to know-- and you mentioned it in your question-- we need to be very clear on what the question is we're actually seeking an answer for. And then we need to be equally clear at the tail end of this whole process, and what are you going to do with the insight.
I often say, OK, so your question is, I have a problem with customer churn. I would like to know which customers are likely to leave me next. And I say, well, OK. Here is a magic wand, and this magic wand represents everything in the middle of that process. Finding the data, engineering the data, running the models, running the predictions, tuning the models, et cetera. This magic wand is now predicting with 100 percent accuracy, sensitivity, and specificity exactly which customer is going to leave you next month.
How do you bank that money, Justin? What do you do next? Then you have to realize that your question was different. Which customers am I able to save with the offers at my disposal at the right investment in terms of their potential lifetime value if saved? My question is not about who churns. My question is about profit maximization through savability. So you realize that when I put it to you that the analytics is not the hardest part, that I will give you-- if I give you this magic wand, you still are miles away from banking the money.
You've got cultural changes. You've got marketing execution challenges. You've got possibly legal, ethical, or moral challenges. You've got staff challenges. You've got all kinds of different things that mean it's much greater than simply analytics. But nevertheless, the answer to the question is, this whole thing starts with a process of getting the question right and at the tail end knowing what you're going to do with the answer. And in the middle there's a very structured way of thinking about taking data on a journey from the operational systems to an analytic data map, choosing the right technique, applying that technique, rigorously testing it, for example, through the cross-validation I described earlier, and then deploying and executing.
JUSTIN THENG: In your experience, what is the level of-- I want to say data capability-- of the marketing people that you talk to? What level do you think that they're at? Because I know that from a more customer analytics background or data analytics background, the leaders tend to lean more into that. But you mentioned a lot of soft skills. What do you see out there in your conversations?
MATT KUPERHOLZ: So Justin, I'm a pretty harsh judge.
JUSTIN THENG: So am I.
MATT KUPERHOLZ: I don't score many companies in the world, or governments, as making full use of the incredible value of the data at their disposal because, in fact, the value from data is unbounded. The questions you can answer when you join it together and find patterns in the right way are theoretically unbounded, and practically levels of value that are very rarely attained. The best examples of attaining them are your AI first entities. Your Googles, your Facebooks, et cetera, that are built on a data and a data mining model.
Everyone else, in my opinion, is falling short. And that includes almost every Australian company that I look at them and I say, hey, whether you are digging stuff out of the ground, putting stuff on the shelf, or moving stuff from a to b, I can find you incredible efficiency simply by looking at this great asset of your data. So I don't score anyone as doing anywhere near their full potential. Now within that harsh criticism, I must say that marketers are even further south of there. But that's OK because it is early days. It is comparatively early days.
But on the other side of the coin, your customers are very soon going to be overwhelmingly represented by digital natives and those that have grown up expecting treatment at an individual level-- relevant treatment at an individual level. I mean, I buy something online, and then I find from the cookies with that thing I keep seeing ads for the same thing for weeks afterwards. And I'm like, what a fail is that. I've just bought that set of, in this case, electric drums for my daughter. Why are you showing me ads for electric drums? You know, that's just a Mickey Mouse example of a technology getting it completely wrong.
JUSTIN THENG: Yeah, absolutely. There's an anecdote that one of our customers shares of his. And I won't say the brand because I don't want this to become a product demo. But his wife walked into his own stores, bought an iPhone, and left the store, and was being retargeted with iPhone ads. Exactly what you just described with the electric drums, and she actually complained to him and said, I just bought an iPhone. Why are you spamming me with-- and there is a role for that real time piece there. What's something in the last 12 months, Matt, that you think might help our listeners?
Now I want you to be completely honest because I'm going to be the black sheep here first. I don't like the phrase marketing as a standalone function, quite frankly. I have always seen marketing and sales as part of the revenue generation pathway, and now customer service is also a part of that because we're looking at customer lifetime value, or average order value, cut size, that kind of thing. So when it comes to data-- and I'll let you answer the question in a second-- but when it comes to data, I think you've touched on it a couple of times now, it's a bit overwhelming.
And I feel like marketers, one, don't have access to the real tools that others might, like in predictive data analytics fields for example. And two, I think there's this kind of, I want to say, the legacy version of marketing which is very gut feel, very what makes people tick anecdotes, that kind of thing. And three, the upcoming generation of marketeers, if I could call them that, who are data hungry, who are consumers themselves, they don't have as much of a feel for what makes people tick.
And we saw that during COVID. When all the data became difficult to use, more and more difficult to use, and things stop working. Funnel stopped working. Facebook stopped working. Google Ads stopped working. What do you do next? So there's that skill gap there. So what's something in the last 12 months that you think might help our spread of listeners?
MATT KUPERHOLZ: Yeah, I'm going to give you a couple of observations I've made because I think they are ultimately exciting and heading in the right direction. The first point is that end to end view not just of your customer, but of your position in market and your entire relationship not being about revenue maximization from marketing, but being profit maximization across the whole value chain. From acquisition through to retention, it's actually about a present value of the lifetime profit, not the revenue, right? Because it's a trade off.
And you may actually realize that you have a different relationship with the household, not just the individual, and that you should consider the individual in their business context as well as in their personal context. So all of these are-- it's important to look wider than that, than just marketing. But instead the whole, as you said, get the provisioning right in the first place to stop complaints in customer service.
Build a relationship where they trust you and they want to interact with you, and then your marketing is a completely different proposition. In fact, it might be pulled and pushed in a different way. So that's one idea, and I've seen that greater sort of horizontal integration. I've seen a bunch of interesting AI first contact strategies where you start to build a relationship with an always on customer service AI agent that's starting to blur the lines between marketing service and just relationship building.
I've seen some really interesting advancements in what we call privacy enhancing technologies, which means if you're organization A, and there's organization B, let's say it's a Telco and a bank, and you would both benefit from sharing data about your mutual customers, and in fact your customers would benefit as well, but there are privacy constraints and other reasons why you can't simply hand over your customer data to each other, privacy enhancing technologies, which are relatively new, allow you to get the benefits of sharing that data without actually sharing it through things like home and morphic encryption and zero based proof.
So that's an exciting development because you remember my earlier remarks about the more data the better, even when you share it between organizations, especially with the right ethics and permission from your customers, that's even better. And the final advancement, which I think is hopefully a trend that will continue and will solve many of the challenges you were talking about, Justin, which is we've seen marketing being taken more seriously as we've seen CMO's join the boardroom table. What today's companies are going to massively benefit from is CDOs and CAOs, chief analytic officers and chief data officers, being elevated to that executive level and considered part of the most senior contribution to a company's ongoing livelihood.
When you take data seriously enough to elevate it to the board level, guess what? It's going to permeate through marketing, through service, through operations, through employee wellness and safety. By recognizing its relevance and putting it up at that senior executive level-- obviously I'm biased making this observation-- but you'll see a few major Australian companies going in this direction already, we will start to see the benefits more fulsomely captured.
JUSTIN THENG: And who owns the narrative, Matt? If a CDO or CAO gets elevated to the board level, who owns the narrative of what the data is saying?
MATT KUPERHOLZ: Well, they must be in tune with the needs of their fellow board members, right? So what are the questions that the chief strategy officer needs to answer to pull the company in the direction of its strategy and purpose? What are the questions that the chief marketing officer needs answers for? What are the questions that the chief financial officer or the chief operating officer need answers for? The chief data and analytics officer is going to furnish them with answers to those questions, and there's only four types of questions we ask in the world, and in fact, in analytics, Justin, which is broadly, what has happened.
Help me accurately report on the past. What is happening? Help me segment, classify, and understand the present. What will happen? Help me forecast the future. And what should we do about it? Help me optimize between choices of scarce resources. And any and every business and societal question can be put in one or more of those four buckets. And any amazing analyst can answer the question with data.
JUSTIN THENG: Do you know what? In that short phrase, and I know you probably saved that silver bullet until last, that answers a ton of questions. If a marketer or-- well, let's say if anyone who's looking at what's best for the customer from a lens of profitability as well as customer satisfaction, those two together, if they were to just ask these four questions. What happened? What has happened? What is happening? What's going to happen? And how can I do something about it? That's brilliant. Matt, it's not-- sorry. Were you going to say something?
MATT KUPERHOLZ: No, no. Just to agree and to reinforce that all of those questions can have hard, and reliable, and verifiable answers when the right analytic technique or techniques are applied to data that's engineered in the right way.
JUSTIN THENG: Brilliant. And that's not necessarily the head of marketing's job, but the job of the head of marketing is to ask the right questions, know where to find the answers, and know how to get the answers.
And what to do with them. Ask questions that are actionable with that magic wand. Pretend you've got the magic wand, and now show me how you're going to bank the dollars.
JUSTIN THENG: And let go of all the vanity metrics. All the stuff that doesn't matter.
MATT KUPERHOLZ: 100%.
JUSTIN THENG: Now it's not every day that I get to talk to one of Australia's top 100 knowledge workers and the global top 100 innovators, so I have to ask your views on the future as we start to wrap up. Are there any industries or developments, big picture-- you even mentioned societal questions earlier. I'm very interested in that. Are there any developments that are inspiring you about the future?
MATT KUPERHOLZ: Yeah, absolutely. These exponential technologies when pointed in the right area, and I fundamentally believe humans are good, mean that we start generating even more value from even scarcer resources. So look at the fact that we knocked out a vaccine in under a year. Look at the fact that AlphaGo, that was originally a Go playing computer using a new form of generative adversarial networks, has now solved protein folding. So just quickly, protein folding, a notoriously difficult NP complete problem.
And if you start to solve it using AI, that means you are going to make incredible leaps forward in drug discovery and personalized medicine. So we're going to do away with horrible things like cancer. The leaps forward we're making in environmental sciences and renewable energy now just make good commercial sense, not just good moral sense. I'm incredibly excited about those technologies and what they offer the world. AI in general, I think, is one of our crowning achievements as a society if we step up to the plate and acknowledge that it brings with it a whole new risk profile which we need to address responsibly as well. But on the whole, I think it's a very bright future for us.
JUSTIN THENG: Brilliant, and I couldn't agree with you more. And what I love about our chat today, Matt-- and thank you for sharing your insights-- what I love is that we went very technical, but at the end of the day, technical in terms of process. Or at least for some of our listeners who maybe aren't data scientist, probably were like, wow. How do I achieve all of this and how do I get my head around all of this? But I mean, as you summarized it earlier, the four steps, the four questions, I think summed it up perfectly.
But it is true that the same advances that we're seeing in our technology for marketing is actually an overflow of the advances that we're seeing more broadly. And you're right, AI is a crowning achievement. It brings with it a whole bunch of questions that we need to answer at a societal level, and we need to answer at an organizational level. We also need to answer it at a personal level. What information am I willing to give over? So that's a very fascinating topic, and probably a podcast all of its own.
MATT KUPERHOLZ: Sure. Well, I've spent the last couple of years focused on that, so I'm happy to pick that up another time, Justin.
JUSTIN THENG: Great. Thanks for joining us today. Guys, thank you for listening and joining us on Reimagine Marketing. This has been Justin Theng. I've been joined with Matt Kuperholz, and I look forward to seeing you next time. Now Matt, if you could nominate somebody to be a guest on this podcast next, who would it be?
MATT KUPERHOLZ: Wow. So Justin, I've got a couple of mates who runs something called Future Crunch. Angus Harvey is a political scientist, and Tane Hunter is a data scientist focused on cancer research. And what their idea is is that if we ever look at the news today, we are bombarded by bad news. The news engine exists to keep our attention. And there's an old saying, if it bleeds, it leads, right? So bad news is psychologically something that we're attracted to.
But what we're missing out on is all the amazing news, all the fantastic stuff that is not puff pieces, but it is genuinely happening in the world. And what these guys mission is to bring that good news to people. So have a look at Future Crunch. I think they are a fantastic way to bring more health into your information diet and stop the junk food of the news that's just designed to shock you and make you depressed.
JUSTIN THENG: There you go. For our listeners, stay tuned. We're going to try and get Future Crunch on here. Thanks, Matt.
MATT KUPERHOLZ: Good luck. Thanks, Justin.
JUSTIN THENG: Thanks, everyone.
MATT KUPERHOLZ: Nice to chat to you.
