Podcast | Tapping into the data lake and human intuition to make predictions and improve human interactions

Jean
Shin

Director, Strategy & Content | Podcast Host of Mobile Interactions Now

22 min Podcast
Bas van Leeuwen

In this episode, we continue our conversation on data with Bas van Leeuwen on data. Building on what we learned about the role of data, we now delve into how to apply data to improving the entire customer journey. Check out his insights on the limitations of making predictions based on historical data, and why human intuition is really important even with a big data lake.

Podcast transcript

Jean:

Bas, welcome back to the show. In our previous episode, we talked about how to really think about data, the role of it, and how to work with it. Now, let’s look at the entire customer journey, are we seeing more trends where data is being used more in our customer journey? Or any places where it's lacking. What are you seeing overall when you see the entire journey?

 

Bas:

Thanks, Jean and great to be back on the show. I think there's a lot of work being done in transactional data. So we getting specific transactions up and running, using data and improving customer experience is really important points. For example, education, right? My experience a decade ago with education institutes, it tends to be complex to get to the right room in a specific building and a specific time, to get my lesson going. And it wasn't an acceptance per se or an exception so to say that we had to go to another room because it was booked and somebody else was there and it was really complex.

So I was annoyed because I just wanted to have an in-depth discussion on a specific topic and not travel around with my bag to another location because it wasn't working. Right? So using data, which is available about education planning, perhaps with the maintenance roster, perhaps with sick or attendance of teachers, combining those data into a better model to plan education experience is really valuable.

And that's all transaction data. This is of course interesting. It doesn't per se improve my life if it doesn't let me down. Right. So I think it's okay because it's a hygiene factor and if it's running nicely, it's okay. But using data to predict outcomes is really interesting. And we see it a lot, especially in retail for example, if you bought this one, perhaps this... if you bought these shoes, perhaps these trousers will fit you also. Or if you bought these shoes half a year later, a new model will arrive. So perhaps this is the one you want to go for also.

But using data and improving the customer experience, really predicting what the next step will be, even prescribing what the next step will be. So automating procedures based on data is really interesting and it's just something also a lot of companies are looking into. This is where the AI hype is about right?

First of all, it was big data. It was data science. And then AI was the new thing. If you organized an event and AI was in the name, I think you get twice the amount of participants. So we're a little about AI, but AI is really about data science or data being used in a specific model to predict an outcome. If you look at a transactional procedure, then data can be really interesting in providing even more efficient outcome in this case. But it's also the limitation because human intelligence is really usable in not confined environments, right? Getting the human intelligence there, thinking about interaction about different possibilities is really important.

So if you look at customer experience, if you're using, for example, direct messenger or chat environment with your customers, you could say, "I'm going to automate this with a chatbot." I think it's really interesting to do so because a lot of questions are asked regularly, how late at what time does my flight depart? Where's my parcel. At this point, you can use a chatbot to answer that. That's all data, that's all describing or prescribing evening and the next event, which should take place.

But there's a line which you cross at some point in which the question and the answer don't have to rise probability to be correct. So if you're below the 80%, which is with the answer being possible correct, you have to switch to a human interaction for example. So the chatbot can generate a specific answer based on data, consumer data, which use your application and send it to some representative or employee of your company who can review the answer and push the answer forward, or even take up the conversation.

And there's real smart interaction between repeatable most transaction interaction, which you can use data and you can even automate it with AI. The first is this is a specific problem which has a broader scope, or it's more complex in which the human can step in and make the customer experience better using the baseline of data which is available, but adding the human intelligence above technology.

 

Jean:

I’m sometimes hesitant to go there, but I'm going to go there. I really think that, in certain contexts, it is helpful to think of it as a part of the human society and there is a role that AI needs to play. And it is sort of a willful decision, how are we going to train this to get to a certain level of confidence…that you're talking about.

And now we have gotten to a point where the data comes in, as you said, every transaction, and then a machine is doing the analysis part of that, and actually giving you the meaning of it. And now it becomes, can it actually make the decisions as well.

But if I look back, the human experience, I don't know about you, but I had some experiences with people whose gut feeling is pretty good at predicting certain things. And I think many of us now and then turn to these people…and they wouldn't be hesitant to do a gut call.

How do we deal with something as qualitative as that? Sometimes it enriches our understanding, and it can be a huge thing.

 

Bas:

Well, I think it's a really interesting question. I think our gut feeling was really important when COVID happened. Because if you make predictions using data, using the pattern which existed from, I don't know, 2000 until 2019 was going to continue on the same path for 2022, 2021. And we all know the world changed, right? I don't think any model which had best AI running could have predicted that. This is a really important point. Data can be used to predict the future by using the past. But if there are some major changes ahead, we have these brains above which you can use to think about the next best solution, because there's no competition right between an AI model and a human.

I think if you look at specific transactions and doing something faster, a specific AI algorithm, if it's applied within the best software context, and if it's linked to specific outcomes, which can then address specific actions can do things more efficient in a repeatable manner, just like energy usage in a server at work.

But if you look at creativity and about solving problems, which are new, humans intelligence is really important because when COVID happens all patterns more or less disappears. And we have to think about new ways of working in some cases. And we really scaled up online working versus going to the office. Companies have to think about new ways of producing goods, because if you're an event business and COVID happens, well, you better go fishing or something else or think about any product to develop, because now events are happening.

If you are making tailored suits, I talked to a specific suit company, nobody's buying suits these days, everybody's buying, I don't know, training outfits and with perhaps some nice upper half and just some relaxing shorts beneath, right? So you have to really think about I'm going to change my model fast or my production fast to get my company up and running, to get the best experience, to continue as a company or as an organization.

I know AI model or data is going to help you in there. The gut feeling about what's the best decision is probably the best one if the environment is rapidly changing.

 

Jean:

Well, how should we think about separating noise from the actual signal that you can embrace and get your action towards. Because, here's what I mean, there's a huge debate…here we are, you're sitting in Amsterdam, I'm in Germany and we are all working from our home office. So there's a huge debate going on right now for commercial real estate, say, hybrid working back to the office. Is there a better way to dissect, recognize what is unnecessary noise and what are some of the signals that you do have to factor into?

 

Bas:

Yeah, I think if there's a lot of information available, think about running hypothesis and challenging the hypothesis and fact is and see what the outcome is. I think one of the main challenges we have is that we are too much focused on efficiency and we think efficiency in any case is the best way to work. Just looking from an efficiency point of view, where people working together, you have to build a team and you don't build a team if everybody's sitting at home all day.

In any corporation, which has a total work from home policy, people are really anxious to go to the office again. Because they want... people are built for interaction with each other. Basic human values if you think about using data to get things more efficient, but also challenge it with what pleasurable and what's on the longterm if I'll go from my corporation.

 

Jean:

As we talked about, the data can point you to many different directions, whether it's about recommending things to people, earlier in the marketing funnel, whether you're coming up with a completely different way to buy things, your entire transaction experience and whatnot. And it happens in different funnel stages. The reality is that we still deal with a lot of silos. Are we learning more about how best to combine and connect the data that is happening in all different parts of the journey and using it for the same business outcome?

 

Bas:

Yeah, I think there's a lot of interest there. I feel look at Google trends, right, which you can just encourage on what is of high interest of people, you can see that the data lake search is I think five times bigger than artificial intelligence. So people are far more interested in how can I create a data lake or more active on how can I create a data lake as in search for artificial intelligence. So I think it's important to think about if we have got these silos, how can we combine those silos towards one environment in which they can... where all the data is gathered and in some ways cleaned or even comparable, so they can run better algorithms or better models, or get a better picture of a specific customers or employees, which is also important.

So building a data lake, which is, should combination of getting all this data resources into one environment is really important. And it's also about making choices not to have a specific solution for every problem, but think about how can we make a solution bigger to solve more problems. And there's always a challenge between both, right, because the specific solution in some cases is the better solution for a specific problem, but it gets you into trouble with total customer entry image. So there's a bit of a interchange or competition between both topics and really getting your data into one environment is helpful.

But again, if you focus only on the data lake and I'm coming back, I think it was in the first episode, I said, it's not about collecting data because data itself doesn't want to do anything. Perhaps start with a smaller environment, run experiments on that, and then broaden your environments and make the lake better. Because if you only focus on the data lake and getting a lot of data in, just harvesting crude oil, which doesn't want to really power your car or your plane in this case.

 

Jean:

Agree, wholeheartedly. We started out the first episode talking about some businesses that are literally losing customers because somebody else is providing more digitally connected experiences that is very much needed these days. So now, let’s think about the person who’s charged with that task, say, don't lose the customer to that competition…now the person is looking at the situation and go, okay, how do we get started?

 

Bas:

Yeah. Just think about maybe make us a simple comparison to people going to a shop, right? So if it's a physical shop I'm going towards the shop, I'm looking at the entrance. If it's blocked or empty or the windows are broken, I'm not going to go to that shop. I think I'm not going to have the best solution right? And it's also about how's your user interface look? Perhaps the customers which are coming in are the right customers but if there's an efficient rates of 100% or 90% things before entering the... I'm going to leave because you're advertising is going wrong.

So it's about getting your analytics really on the shop front, which is available in most cases from Google Analytics, it's an easy data platform, of course, for a lot of companies to use. If you look at the specific customer journey, you can look at which products are clicked on most, which products are best used if they go to the charts and wanting to purchase the specific products, is everybody backing down at the payments? So is there something wrong with the customer interaction on the payment part? And it's also good to challenge some people around you.

This is not per se, a scientific way of working, but to ask even your grandma to experience how the digital flow works, because for even your children, they can really get you a new perspective on or your parents, how to use a specific environment, a digital environment, and where people are challenged on how to use it. And what you can also do is just look for trends. So if you look at a trend query, like Google trends, you can see what are people looking for at this point.

So if I think about my own product, how should I do a search on that? And what's important for people to have a look at. If people bought trousers in your clothing shop, they're bound to be interested in more than trousers only. So if you use data about trousers and you think we have to sell them more trousers, and your only offer them trousers, and don't ever go into to buy sweaters. I think really with large platforms like Salesforce, for example, which has this on-time proposition, which has a lot of also chatbot interaction can really help you get started on more intelligent solutions using data. So you don't have to build it all yourself. There's a lot available on the market at this point.

 

Jean:

Excellent. Given the space that you work in, is there something that you're really excited about, that really excites you out of what’s emerging, what you're looking out for.

 

Bas:

What I think it's really interesting, in the past CRM platforms were focused on segmenting in specific larger groups and now you see interaction with customers has become far more individual. It's easier for customers to interact with bigger platforms, such as Facebook and WhatsApp the social channels or mobile, to get some interaction with a company which is also aligned with the CRM platform to get specifically tailored response back, and also a specific offering. And if you look at, I don't know, five years ago, for example, looking at a CRM platform, it was more or less, this is a specific large group and they have to have this specific offer because we think on average, it'll suit them best.

And if you look at the interaction now with the usage of data, and so your advanced forums and specific interaction methods like mobile and chats and your website, offerings can be made far more individual and even solving a solution can be made far more individual. Which is great because then you see people as individuals and not so much as larger target groups who are the same.

So I think that's really interesting developments technology evolving to the point in which technology acknowledges in some ways, which we built of course, that are individuals and not so much groups.

 

Jean:

Yeah. I remember the era where the word personalization just meant changing the name on the top of the same letter and I'm like, well, that's called a mass personalization, which is an oxymoron. You could be a mass or a person. So I think we are at a stage where we're looking at real personalization, that's exciting space to be in.

 

And this is kind of a nosy question, but can you tell me, so that we get to know you better, what you do most on your mobile phone these days? Three things, I know you have a lot, but three things will do.

 

Bas:

Yeah. Just thinking out loud. If I travel to the office in the morning, I always check the traffic. So that's a really important one for me because traffic can be dense in the Netherlands, although it really has improved. So to say strange comparison. But when COVID entered, it really improved on the traffic density on the Dutch highway. So estimating how long it will take is important part for me.

Music is of course really important. So using Spotify is one of the things that I really like to do. Also with the recommendation engine beneath it. So it really gives me new music ideas to listen to based on my individual needs. So that's an important one. I'm using a WhatsApp a lot to communicate with a lot of people and just easy communication getting in touch.

And I think it's easy, even more important since COVID happened, because we have some limitations in the Netherlands in which you can't visit your friends, right? You can't have people coming over. So using traditional phone, but also even chat environments are really important for me to keep the interaction going. But there's a group of my friends, which in some cases I haven't seen for six months or so, and only I had some interaction with them using chat on phone. But it's still highly appreciated and very important for me. So those are three, I guess.

 

Jean:

Excellent. I think that is a lifestyle statement more or less. So thanks for sharing that. And it was absolutely lovely getting your thoughts on some of the tough topics and I just absolutely loved how you can put analogies we can all relate to. That was really helpful. Thank you very much.

 

Bas:

Thank you very much for having me. I really appreciate your questions and it's been a fun talk.