Podcast | The connected technologies behind connected experiences – a solution architect’s view on what’s holding us back.

Jean
Shin

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

30min Podcast
Episode 37 Guest: Anders Weijnitz, Director of Connected Technologies at Valtech

In this episode, we check in with Anders Weijnitz, Director of Connected Technologies at Valtech. Anders will help us define what ‘Connected Experience’ means when it comes to how businesses interact with consumers. We’ll also be delving into the latest technological enablers – and what’s holding us back.

Podcast Transcript

Jean:

Anders, welcome to the show. I'm really excited to talk about this huge topic of connected experience today, because it happens to be one of those topics, I think there is a lot of advice for these days, but not enough clarity. But before I get ahead of myself, to get us started, can you just tell us a little bit more about you and what you do at Valtech?

 

Anders:

Sure. Thanks for having me. I'm Anders Weijnitz. I work at Valtech as a solution architect. I also head up the practice, which we call Connected Technologies, which is broadly anything that has to do with connecting customer experience with technology and enabling the best customer experience through technology.

 

Jean:

I could not have asked for a better person to really tackle this. I keep thinking it's a humongous topic. I really want to start from some basic questions, so that we are clear about what we are talking about. I mean, our conversation is happening as we are, I guess, in the second, third, fourth wave of the coronavirus pandemic. It depends on which part of the world you are living in right now. I think we can safely assume that no matter how they actually call it, some kind of concept of connected experience is really on everybody's mind, especially on the business side.

But first, just help us really clarify what we really mean when we say connected experience, in the context of how business is interacting with a consumer.

 

Anders:

Connected Technologies, connected experience is of course a huge topic. And it's used in various settings, a little bit like digital transformation, which is also means everything and nothing nowadays. I see it along a couple of dimensions. The recent trend in connected experience comes from IoTs and more from a technical angle that connected devices, surfacing data from that.

But it's broader than that, because it's also, why are you doing that? It's not only to connect devices, it's to create an experience for the users. It could be for customers connecting the real and offline world. How do I take a customer when it's not online orders? How can I guide and help my customer on a user journey, which starts online with researching different channels, to a conversion offline in-store experience and connected technologies can help guide and enable that journey.

To me, looking a little bit beyond that, it's also in the context of something called quantum computing or related to ubiquitous computing where specially looking at smart phones and recent developments with machine learning on edge fingerprints, LIDAR, everything that's included today. It's the augmented technology experience, which allows in the end, this quantum computing, where the user is using technology in an implicit fashion. Technology is there, but in the background enabling the experience, but it's not something you open the machine learning app that you unlock your phone without having to do anything, because it recognizes your face or your fingerprint. If you talk about connected technology to talk about connecting up devices, it's implicitly going in the direction where quantum computing and ubiquitous computing, that's where it all ends up.

 

Jean:

I remember what it all meant like, okay, how do we make internet work? For like an online experience or something like this. But these days, when I go to, actually almost two years ago when I used to go to stores…there's so much technology there that I may not be really cognizant of, but I know more technology is being used there. Is it correct to think that now that the physical experience has a lot of technologies too, whether visible or not, and there's a whole online experience that we've been developing probably last couple of decades…something like putting that all together and making that seamless. Is that what we are talking about?

 

Anders:

I think as soon as we say seamless and the customer focused or customer journey focus development, and customer journey focus. I wouldn't say features, but it's implicitly also saying that the technology takes a step back and it's enabling, but it's about the customer. And it's about the experience, connectedness in some way, because it's connected, the data is connecting the customer to whatever their goals. They have the service they need, the help they need, the products they need. Ideally, they shouldn't notice the technology. They should be able to focus on their test, right?, And then what they came to do rent a car or book an appointment. They're not thinking that may use outlook for 15 minutes.

 

Jean:

Without being spooky though…I remember having a conversation about some of the authentication technologies and when things are just too connected and seamless, I was told that when there's no user interaction, people are getting little spooked and like, how did you just do that?

 

Anders:

Totally and especially, I'm based in Germany and part of Europe and GDPR and through history also Germany, you had some really bad experience with collecting personal data almost and using that in a bad way. So the data can enable a lot of things, right? And my personal data, who does it belong to and how is it used? How can I control how it's used? And I think it's about transparency and control and lack of bad surprises.

I was listening to a podcast the other week, which they discussed the seeing as the algorithm and what is the algorithm, and one of the interesting nuggets of information, which I picked up is that TikTok has showed us and the success of TikTok has showed us that it's not a social graph anymore. They are collecting hoards of information. I'm not saying that, but it's less personal because tagged content information and the algorithm, it's not my connection. So recommendations on YouTube, on Facebook, on Instagram, you follow people. You don't follow topics. You follow people and kind of trust, you compose the social networks, which you trust to surface topics or content. You want to get product tips. What are the social shopping is the thing that's interested through the social network. And TikTok doesn't do that. What they do is you look for content and you consume content than the algorithm, not the people you follow. You can follow people too, but actually that's-

 

Jean:

Well, yeah. Perfect for binge-watching generation, right? You start consuming certain content and you getting deeper, deeper, and you're getting all the adjacent things, and it is about what you're consuming.

 

Anders:

Algorithm, whether it not [inaudible 00:07:49] so you're subscribing implicitly through your connection to topics, not to people, which is another form of sure, it's data gathering but it becomes less interesting for the supplier to kind of hoard your personal data and all your connections in the name of your friends and their friends and so on. It becomes more tailored to your current interests because it fluctuates over time and it's, what you're interested in and the content, the kind of content you want to consume. It has an effect on how you design, especially a mobile experience. TikTok only allows you to see like one video at the time it's full screen video, as opposed to other media or Facebook or other apps where you have more like cluttered experience. You can say that's for clarity for the user, but actually it also helps the algorithm, this recommendation argument, because you can take these videos and you know that the user is not watching something else.

The focus is on this topic right now, and all the topics is classified, so the algorithm knows doesn't do like frame by frame analysis and image recognition. It's just metadata that's attached to these clips. And like, there's a dog in here, there's music, there's dance and that is used to then feed the algorithm. And that's what they meant by seeing as the algorithm. The user interface it's actually partially made for you, but partially made for the algorithm. So you would take, because you couldn't do the algorithm that good, which is so that you need to design the interface to let the algorithm pick up on what you, like you're interested in, what your preference is, which is and then enabling your user experience, to feed you your preferred, to filter all the billions of clips that I have to filter out this stuff you're in.

 

Jean:

So it's a design to help them help us better, as well as our consuming interface. Now, I'm going to bring closer to the connected experience. I think there are many, many different ways to do this. As a business, we sort of know we want to do this, but if there's a big Why, what is it? I mean when you started putting first online shops on the internet, you kind of understood why you were doing that, right? So tell me, what is that big Why? What is the connected experiences trying to do that other digital, conceptual frameworks weren’t able to do before?

 

Anders:

In the context of commerce, the connectedness there is that the users, so they don't choose the channel to shop in. They go multi-channel, they hop between channels. So connected experience there is in the online commerce context is around following and supporting, not of stalking following. And that's importing the user on that research, augmented reality, for instance. So we're downloading the model you're interested in to see the sofa in your room, in the context of your home that's connected experience. Post-purchase, it could be sporting while in a negative case with returns, make that partly see or change something if it's piece of clothing or clothes. Connected experience and one of my favorite topics, which is some machine learning on edge or AI on edge, in-device machine learning, help produce returns by ensuring a better fit, for instance. Is it fitting clothes? Or is this piece of furniture fitting in my home? Is the car going to fit in my garage?

 

Jean:

Which by the way, thanks for bringing up a topic that made me go ‘oh my gosh!’ What looks like just a common commercial problem is actually linked to even a bigger problem. For example, because the fit wasn't able to be more accurate, there are many returns…thinking about this COVID time we’re living in and the volume of returns…is really causing an environmental problem, not just a business problem. So you brought up a real recent topic here. Are we getting any better using AR and those things to try to help this along?

 

Anders:

Absolutely, I'm optimistic. And in this case, because there is gains for everyone. So it's not only thinking about the environment and waste and minimizing waste. It's also minimizing costs for the business. If you sell something, you don't want any returns, then you don't want the goods to go out to the customer who buys all of the sites above and sites below, just to try them on, send it back. And there's a lot of product being tragic because it's cheaper to destroy the returns then to repackage them for our new sales sometimes. In my experience is that it's increased in the last year alone, the machine learning, not on the device, but it's not only size, it's fit, right? So is it narrow fit or is it for which kind of body type size finder. Let's say previously, it was just a table, used to be a table where measurements and you could kind of see.

Now, it was using a product. So I don't know how it works behind the scenes, but clearly using machine learning to reduce your chance. So you want the algorithm to minimize for that customer, the amount of return. So you can train it on the returns to predict the size, because sometimes I don't know, small is maybe I should have a medium or large, and that information can then be used to advise the customer and that's classic machine learning topic, but some rules, but much occurred as you get returns flowing in. You can also see what did they order, what did they, in fact, in the end, buy.

 

Jean:

That really depicts some of the things that we are talking more about in terms of enabling the experiences to be connected. You really have to have data and process and the communications all connected to make this happen.

 

Anders:

The packaging of technology and making things. And let's say machine learning, which used to be just a couple of years ago, more like research topic. Whereas today, it's very quickly becoming just another tool in the developer tool belts, classic relational databases many years ago were also new. You had like a database admin or database expert on every project that needed a database. And that was the guy who kind of sets up the database so that it would work in the first place. Whereas today, the database for normal size run of the mill project, you don't need the database admin guy to be there as an expert on databases. It's expected for a developer to know how to use one. And it's very well abstracted through drivers and other things. Now with the cloud, of course, it's on tap to just say, activate database and you have one and you get like API key, whatever, and you can create, integrate the database in your code.

The machine learning and AI is rapidly going the same direction and even more so it's coming in these low code, no code environments as well. I don't know if you've seen teachable machine from Google. It's a great website. For anyone who didn't go to teachable machine, go to teachable machine. It's super fun to use and no programming skills needed. And it's really combining. It's so inspiring. It's combining again, my favorite topic right now, machine learning in the browser and on edge where you can take precooked datasets for instance, audio or image recognition, but also posture recognition or body segmentation. So this is, let's say facial recognition or say you want to make a small app or webpage that takes you selfie when you smile. Within five minutes, you have your data set trained, no programming required, and it can really be creative, mix it up and cover a whole lot of ground.

 

Jean:

Can I tell you something though? I remember recently, I think it was just a few weeks ago when Microsoft was doing a product launch and talking about how they basically built how businesses use relational database, and I remember in grad school, like trying to build a decision-making system and back then it was just absolutely rudimentary. You are doing the thinking part, right? Basically, trying to make the decision and then make the machine execute it out. But one thing they said that really captured what I was feeling is that…remember those days where you had to look at the data yourself and you had to think about what that meant and you kind of figured out what to do with it. But now more and more things are going in a way that the data and the application is telling you what to do.

Are we there where if you are dealing with a customer, you know this customer who orders those things…and when he purchases like this,  he tends to return the product…so maybe you give him a second recommendation or something like this. Are we at a place where 20% of people who bought it in this color turned out to make a lot more returns, maybe there's something going on and then connecting all this data and making those decisions or changes along the way?

 

Anders:

Yes, one person's data might help, but in order for machine learning to work, you need a lot of data and it's not like there's 20%. The algorithm will tell you where the machine learning works is like given unders vector or data points, what do we predict? Will you make a return on this order or not? And for that to work, it needs a lot of data points for other purchases and other returns, some individuals or some consumer patterns are probably, more likely to be the type of customer that returns some goods than others. To make an individual recommendation, you need the data from everyone.

 

Jean:

I like this way of thinking, what problem are you trying to solve? Because there are many discussions in terms of how machine learning really is data hungry and the more data we feed it, the better it gets. And that opens up for having to connect many different platforms for certain problems. But there might be other ways to solve that problem. Any exciting project that comes to your mind that really embodied the connected experience?

 

Anders:

For me, it was one of these touched everything that I feel it's the center of connected experiences. And it starts with IOT and it starts with the industrialization of agriculture or the next level, the optimization of agriculture. So, you see agriculture farmer on TV in the advertising is like small scale those farmer just picking some ripe tomatoes and carefully packaging to go out to the customer. That's not how it works. So of course, it's industrial scale and very much data-driven, but it's data-driven on a next level where today you plan depending on what kind of soil you have in different parts of the field. So it's not like on this field, I'll have that, on this field, I have that. It's really mapped out with geo data, the fertility of the ground and all the tools you have, and they are all connected and they all know where they are.

So more or less, there's the SIM card in them and loads of sensors. So you can plan first. You want to use your fertilizer toxin. So in an efficient way, you want to optimize it. So if some part of the field doesn't need so much fertilizer, it will actually just reduce the amount there so the nozzles will just spray less. They ask there generally from the manufacturers, the end consumer farmers and the industry. So there are laws and regulations around this, especially in EU and how you use fertilizers, because you want to avoid environmental impact and so on. You need to produce a lot of reporting to prove that you are compliant as a farmer. The connected experience goes to this level. We talked about in the beginning of the podcast where the customer experience and the customer journey.

So you need to, of course the planning, all the fields, it doesn't take place on the phone. It takes place on a desktop by, I would say, not necessarily a computer savvy or even interested user, they want to get the job done. They are farmers. They want to farm the land and operate their farm, not sit in front of the computer. So it's lots about interfaces, presenting complex data to the user making sense of that, and also on the mobile experience and making sure that only the right data is presented at the right time.

 

Jean:

So, let's talk about this. I mean, the choices are there, technologies are there. But if we really think about it, what is really holding us back in terms of more companies really deciding to make bigger waves into connected experience?

 

Anders:

I think one of the things that's holding back, so this is a technology. It's not a trend, it's effect, but it could be accelerated. And sometimes I get the question, "what's holding us back if we have all these machine learning?" One is very detailed and that's XF information. If you take a picture with a camera, there'll be metadata in it that says, is it portrait mode or landscape mode? And that's simply often wrong because the sensor, it doesn't recognize it or you have, if you're like me, I have it locked to portrait mode all the time, because I don't want the interface to just jump around. For myself as a, summer project side project last summer, there's a game where you can guess where you connect license plate registration. So in Germany, they have a prefix so you can see which state they're from. Like I said, from Bavaria, is it from. If you're on a long trip, you can play a game in the car with the kids, like who is sees the next from that one. I don't know these codes because I'm from Sweden. And then they don't tell it so I was like, I wanted to win. So I made a small app also again, like with machine learning on it, like we use a TensorFlow to just... My day was like, I'll just point the camera at the cars. It will cut out the cars, recognize the cars, and do OCR on the license plate. And I will say, if it's from Baden-Württemberg or Niedersachsen or whatever. Before my kids or the family-

 

Jean:

Now you tell me, I lost so many games.

 

Anders:

Anyway, I made that and it kind of worked, but I ran up to exactly this problem. You could have a car that standing still right in front of you. You hold up the camera and it, the algorithm just says, there's no car in the image.

 

Jean:

So did you do a bad programming? Did you do bad coding or did that app didn't have enough data to learn?

 

Anders:

Where you as a developer run up against the wall more or less is when that happens. So, you point your algorithm at a data, it returns from a warehousing system or in my case, pointed at the car and the training data. And you need thousands of pictures of different cars from different angles. You don't do that yourself. You download pre-cooked datasets, you need a lot of source material, in different lightings, in different conditions to make an accurate finning data. When the data wasn't trained on the kind of images you feed it, then there's no debugging you can do as a developer to fix that. It's just not going to work.

Jean:

So the fair takeaway from this conversation is connected experience is really happening, but it needs a lot of data. And that is at the end of the day, once you enable these technologies, how this is going to get better, because that is in essence the learning mechanism.

 

Anders:

A lot of data, but also the right data. There's a debate about bias in data. And it started with racial bias because some auto exposure cameras cannot deal with different skin colors or tones, but there's actually always a bias in data. The algorithm needs to see all cases and have enough data in each case. And if you happen to have a little bit more in some case, then that will be, that's the bias in the data. You can measure bias. You can statistically do mathematical transformations on it to find out the properties of your data. But I wouldn't say it's data hungry. Yes. But it's, what's more important is probably the right data and trying to guard against bias saying in the data.

But there's another financial reason, and what comes to mind to me... And this, I see this in many projects or many conversations with customers, is the cost of app development. On one hand you have a low code, no code tools, where you can... They are great for small campaign apps or very, I would say, not long living enterprise applications. But as soon as you step outside of that realm, quickly the app development to cover Android and iOS and all the variants with tablets and so on becomes much more expensive than customers are sometimes expecting or budgeting for. Because they underestimate the effort, also the ongoing maintenance efforts, there are updates to operating systems, there are new recommendations from Apple and Google. And that's a little bit of a failure on the industry.

So when apps came, it was expensive because it's new technology and people were loving... It's just an app, they were thinking website and getting presented with the cost of maintaining an app is a computer program and an application. And it's a specialized skill, you cannot just drag and drop website. And it's a usable app with a native experience. And that has never really... That expectation gap of cost versus actual costs has never really been closed, the way I see it from a technology standpoint. The customer expectation is still that it should be quicker or that you can reuse a lot of functionality between platforms. But that's not really the case.

And I think that's one of the... It's not lack of standards, they are standard platforms, just different. Two different platforms with Android and iOS, with their own sets of challenges. And that's really, if that would be in a web development context, where it's much more standardized. And the browser as a platform has enabled development to go much faster, I would say. And the deployment of new technology within the sandboxed case of a browser. But yeah, that's something I'm seeing. If it was simpler to develop native apps, we would see more adoption of new technologies on mobile quicker. I think, yeah.

 

Jean:

That really comes back to the age old problem of total cost of ownership.

 

Anders:

Absolutely.

 

Jean:

And that really rings true. I see that when we provide business solutions for WhatsApp Business API. Part of the reason why these platforms are more successful is because they handle the device problems. So it is about going into a place where they don't have to worry about all these different operating systems and devices, that’s certainly part of the solution.

 

Anders:

There are hybrid technologies, et cetera coming out, but they're still not really closing the gap fully. They have their use cases as well. And progressive web apps, which you can have on your phone and have an offline experience, they're not really taking off. Customers are not used to saving a webpage as an app on their phone, and it's not in the app store. So there are technical possibilities maybe, but it's not really yet taking off, I would say.

Part 2 of the interview with Anders Weijnitz will be released in two weeks, following this release of Part 1.