Product Evolution: The Journey Of Humanizing Digital Experiences

Mansi Kamdar at UXDX USA. Video: https://youtu.be/mAxTE5M3f08

Readable transcript: edited from the recording's captions for readability (fillers and false starts removed, punctuation and section headings added). Wording is the speaker's own. Timestamps are positions in the video. Names marked [?] could not be verified against the audio.

A product leader's journey

[00:00:00] Hello UXDX, I'm Mansi Kamdar and I'm super excited to speak at our first UXDX conference today. I'm going to take you through my journey of how we're going to build products and humanize these digital experiences, which this pandemic has taught us is even more important. So fasten your seat belts and get ready for this ride.

[00:00:24] As a product leader, I've been fortunate to work across various domains in the past 15 years. I started with printing at Xerox, the Google Calendar that many of you may be using at Google, in the financial industry with JP Morgan as well as Fannie Mae, and now working for one of the biggest retailers, which is Walmart. In this journey, working through various domains, what has taught me the most is to offer and appreciate the importance of learning the skill and art, and valuing the importance of people, culture and ways of working. These then become the foundations to make this ride a little less bumpy and a lot more fun.

[00:01:10] As we get started, let me tell you a little bit about myself. I'm a principal product manager at Walmart, and what I love doing each day is creating and building products that touch millions of lives of people and associates. As we build, we build for scale, and that has taught me about being a fearless leader, a woman and a mother. And there's no joy in a ride without the music and the love for cooking, so those are my little go-to places when things get too rough.

What is product? A meal on the table

[00:01:46] In this journey I often get asked the question: what is product, and what is it that you do every day? It's the most sought-after field, it feels like, these days. Can you tell us what it takes to be a product leader? You'll often find analogies where people would say you're the captain of a ship, you're an orchestra director, or I've even heard one that says it's like a game of cards, where you need all the strategy to make sure that you can get it right.

[00:02:15] What has really worked for me as an analogy is that product, to me, is what I do every day. Think about putting a menu together for a meal at the table, for the day, for the week, when friends come together, or even the financial planning that you do every day in your lives. If you think about that process, what you're thinking through is: what do I have in my pantry today? Do I have something, or am I missing something on my list? Are my little audiences going to be happy with what I put together? Am I going to have to influence that idea, or am I going to have to put my foot down and drive a decision?

[00:02:57] I'm sure all of these sound way too familiar. These are nothing but setting our north stars, putting our goals together, thinking about our problem statements, managing our stakeholders, being customer-centric for our little kids in the house, influencing and driving that execution. Whether it's cooking that meal, managing your finances or building these products, these are the life skills that help you be a better product leader. What I've found is that taking these skills into work helps me be a product leader, but taking the learnings from work helps me navigate my own life better. That's where that whole journey comes together.

From requirements to data-driven product development

[00:03:41] How did we get here through this journey? I'm sure all of you have either been through, are in, or can resonate with one of these phases. You're either in one of the phases looking to move and evolve into the next evolution stage, or you've experienced them and you've gone through those phases together.

[00:04:06] When we started this journey, if you can jog your memory back, think about the days where everything was requirements. Business or subject matter experts told us what they wanted; they told us exactly the solution they were seeking. We took those requirements, we translated them, and then we provided them to our software development teams so that they could build the solutions to meet those needs. From that phase of requirements we then moved into more product-centric thinking. We started putting the products in front of us, at the forefront of our development.

[00:04:52] Very quickly we learned that we need to be customer-centric. We've got to think about customers, not just about the product, and what needs we are meeting for those customers. And now you see a lot of data being collected with this journey, and that then becomes data-driven product development. As we go through those phases, what we want to do is build these products to serve the needs of what people want and desire.

[00:05:17] What this visual is going to do is put that story together in perspective. I'm sure you all can resonate with the tire swing cartoon. You've seen something being asked for and not quite getting what you asked for. Since then you became product-centric, and you're probably wondering what the ketchup bottle is doing in the middle of all of that. That's an analogy that I love using. Think about the ketchup bottle with a cap on the top and the ketchup in it. It serves your need: you want to get the ketchup to your users, and you made that happen.

[00:05:55] But think about the struggle of the user when they're trying to get that last bit of ketchup at the bottom of the bottle out of the bottle. You change your perspective, from being product-centric to caring about the customer. You just flip that bottle over, and voila, you've solved that problem. That tells you the shift in the mindset that we have to make as we think about the products that we build.

[00:06:20] Very quickly, as we went through this journey, what we've also done is taken our ideas, ideated on them, prototyped them, taken our customers' insights, and then built those improvements through that data into the next development that we want to see in our products. That's where data-driven product development comes in. Think about it as idea creation, concept creation, forming your strategy, developing that product, testing it, launching it, getting it into the hands of millions of users, and then going back and rinsing and repeating that process. That's what the data-driven development process is.

The retail journey from store to online

[00:07:07] Let me use retail as an example. We've all been a retail shopper in some form or other. We see ourselves go in as customers to the store and look and feel and see the products that we're planning to purchase. If you think about that retail journey way back, customers didn't trust online or didn't even think about that concept. You always wanted to go into a store, you wanted to touch and feel your products, you wanted to know what you wanted to buy, and then walk away with that transaction.

[00:07:41] Then very quickly came a need where we wanted to leverage online shopping. We wanted to do that so that we could look at competitive pricing and at what the offerings are across the different stores, and then go in and complete that shopping still in the stores. The store was providing us education, inspiring us when we were trying to make a meal, pick a product and find other products that we were looking for, and we still saw our customers wanting to create that connection.

[00:08:12] But with this pandemic, more so, we've realized that we've got to enable and unlock online shopping to the extent that it feels like a human interaction. How do we take that leap, and how are we taking that leap from not wanting to go online at all to now wanting to shop online for a large portion of the journey that we take? How do we make that foreign, not-so-understood experience the most go-to experience?

Foundations: people, roles and ways of working

[00:08:45] As you think about these and start to build these building blocks to come to execution, what I found is that it requires the foundations to be in place. In order to create that shop and browse, how do you create the recommendations, how do you personalize? How do you think about a customer that walks in and gets inspired by what they're shopping in the stores, and recreate those experiences online? How we make that happen in a wholly digital format for our customers is what we're trying to focus on.

[00:09:21] In order to do that, the foundations that we want to build on are how we think about our teams, how we think about our roles, how we think about collaboration and how we think about empowerment. It's not just the tools, it's not just the tech, it's not just the thinking, but how we bring the ways of working into our foundation so that we can truly build bigger, better products. And how do we take our people through this journey that was foreign, with design systems that are usable, and put that thought process into being?

[00:10:01] What we find is people, culture and ways of working being that strong foundation. In order to succeed, we need to respect each of these roles. We want these roles to collaborate together, and we want to empower our people playing these roles so that they can make and drive those decisions and ideas. With this four-in-a-box concept, the ability to monitor and track our analytics gets us to build those better products.

[00:10:34] When you think about each of those roles, I'm going to spin through this little circle that you see. Our business leaders are providing us the direction. They know the domain, they understand the history, and they know the gotchas of the business. They're telling us what the problems are and what they've seen historically, which helps us better understand the problem on hand. Then we have our product leaders, who put the customer-centric and product-centric hats on and go in to problem-solve and plan for that launch.

[00:11:12] Then comes our design group, which puts in the utilization of design principles and design thinking: discovery, research, prototyping, collecting all that usability feedback, and then bringing that product to life. And last but not least, our engineering or technology teams. They're architecting, designing and building these products that not only serve the need today but are also built and scaled to hold strong for tomorrow.

[00:11:47] That's what has evolved and become our data-driven product development approach, so that our customers enjoy and can recreate those shopping experiences in a way that humanizes it for them. As we think through this journey, what I want to get your minds to is the tools and the techniques that are being used. How are you thinking about discovery, the concept design and review? How are we putting those in the hands of the people that can actually feel and touch our products, use them and give us feedback? How do we collect that feedback on an ongoing basis, optimize it, and then put it back into our product strategy?

[00:12:33] As you hear all these terms and processes, I'm hoping that you can either resonate with them, you have experienced them, or you're just itching to put these principles into your next big project or product development.

Humanizing search and shopping

[00:12:47] As you think about those foundations and roles, business, product, design, engineering, and the culture and ways of working, let's see how we can relate that to a retail shopping journey and how that's evolving. That's some of the work that I've been doing in my most recent projects, and I'll share a few experiences of that process and how it gets implemented into the work that we do.

[00:13:15] Think about a customer trying to search for a product. Where we started was keyword searches. Let's say you were trying to buy a cocktail dress for an occasion. You're getting ready for a party, you want your best cocktail dress, you want to look your best tonight. As you're thinking about that, you're going to put in search terms: I'm looking for a blue cocktail dress or a red cocktail dress. Am I going to look for something that's off-shoulder? Am I going to look for a certain contemporary look? We use these keywords and surface the products that our customers are looking for.

[00:13:55] Then we evolve through that process and we start recommending to our customers. You may be looking for an evening outfit or a day outfit; is there a time factor in when you're going to utilize that product? Are we thinking about customization, their branding, their personalized needs? Are we thinking about their size, about what brands they shop with? That's where the evolution came in, and we started understanding our customer, the customer 360.

[00:14:27] Then we start to bring more intelligence into the shopping experience. I previously talked about walking into the store, where it was educational and inspirational. You walked in and saw an aisle full of pasta, but then I told you, oh, I need the sauce too, and by the way, do I have all the spices and the seasonings to go with it? Similarly, we want to do that online. How do we make that happen? We not only recommend another dress in the category that you've selected, but we take it further and say, hey, you probably need some sandals and some shoes, you probably need a purse, you need some accessories to go with it to get you that perfect look for that night.

[00:15:13] How do we start intelligently offering our customers those experiences that they would have seen in the store, and replicate those experiences no matter where their journey starts? That takes you through that evolution: now you don't just shop for the dress you were looking for that evening, you've bought a whole slew of accessories and things to go with it, so you can add them to your basket, your cart, and walk away with that perfect look for that night.

[00:15:45] That's taking it even further, humanizing the shopping experience. And where are we leading to now, beyond this? Now we start thinking about shopping assistants. When you walk into the store, you're meeting other humans, you're meeting our associates, and they're guiding you as to what you're going to need, and you're getting that feedback from them. Now what we have are shopping assistants that interact with you and ask you questions that allow you to hone in on that perfect buy, also using 3D models and putting that dress on you so that you know how you may look. That's where we start to bring the human element into the shopping experience.

Machine learning and AI as tools

[00:16:30] In order to create these customer-driven digital experiences, the tools it takes to make that happen bring us to what machine learning and AI truly are. That's our next leap: how are we using these tools to enable and unlock these experiences for our customers? When you think about machine learning, AI, data science, you're hearing all these terms in the industry, and you're probably thinking, how am I going to put these to use? Am I forced to use them because they're the next best thing to do, or am I truly looking to solve the problems at hand?

[00:17:09] When we put these tools to use, what we really want to do is take a step back and come back to our core principles. Are we respecting the roles we play? Are we collaborating and empowering each other to play our roles better? And are we solving the problems at hand? When you find those foundations are in place, that's when we use these tools to make that launch into a better experience as we digitize our presence online.

[00:17:43] If you think about the journey that machine learning has taken, I'm sure we all have experienced rules-based systems. If you had X, then you did Y, else you did Z. I'm sure we've used that if-then-else in all forms and manner. In many cases those rules-based systems still work. It's not about leaving a style behind and moving forward; it's about taking the best of the tools that we've learned and putting them to use.

[00:18:20] For instance, with a fixed set of rules with a fixed set of outputs that you're trying to drive, there are advantages. They're easy to debug, there's no training required, and they have high precision. But the flip side is that they have moderate coverage. You've got to think through all the different rule sets that you need to have in place as you're making that determination. You also need to have at your disposal a lot of subject matter experts who know and understand the domain and know exactly how to define those rules.

[00:18:57] Then, as we think about the tool explosion that we've seen in the industry supporting our product development, the next big explosion was big data. Now we had all of these rules, we started collecting lots of customer data, and we have lots of data all around us. What do we do next? We've got to analyze all of that and put it to use. That's where the big data explosion came in, and one of the things it required was that we were extremely reliant on manual analysis. We had all this data, we drove some insights out of it, and then we had to analyze it so that we could actually derive some type of results.

[00:19:41] The next logical progression, because that's not as sustainable, was the progression into machine learning and data engineering. What that does for us is we start to put models together and predict behaviors using the large amounts of data that we have. As you can see, you're not leaving a tool behind. You're taking the rules and building on the data, and taking the data and automating its behavior and predictions using advanced learnings and tools, to achieve the outcome that is desired.

[00:20:16] A great example of how this comes together is automated voice response systems. I think we all love them, don't we, because we want to talk to somebody on the other end that tells you punch one, punch five, punch seven, pun intended. The rules-based system asked us to press one for payment, two for billing, XYZ. That was cumbersome and painful. We all wished we never had to talk to that machine, because it never understood what we really wanted.

[00:20:52] The next progression came in and said, how do we make that a little more human-like, a little more intelligent? The next wave of improvement was that there was someone speaking to us on the other end, which was still automated but had a dialogue with us, understood what we were asking for, and then navigated us through that help journey. But it still had its hiccups. It didn't get our accent, it possibly didn't have options that we had thought about. You were looking for something but didn't quite get the help, or it took you way too long to get the help when you wanted it instantly at the first step.

[00:21:28] Where are we now in this process? We're now in a human-like conversation. Think about the Google Assistant that you use, which converses with you as if there was a human in front of you, understands your natural language, processes it, and then communicates information back to you so that it can navigate you through that journey. Bringing all of these concepts together, you have the foundations of people, you have the foundations of collaboration and empowerment, you build on the product journey that you take by putting your customer at the center of it, and then you think about the tools within it that actually enable that journey and make the ride a little less bumpy.

Solving problems beyond what we ask for

[00:22:18] As you think about machine learning and the intelligence behind it, you then go into how that's unique in the retail world. We throw all these tools at our problems, and then how do we solve for the problems at hand in our retail world? As we talked about, in the shopping experience we were looking for a dress, we wanted to go to the cocktail party, we put our search terms in, we got all the recommendations, we personalized it, and then it gave us recommendations for other accessories.

[00:23:00] How do we take it even further? Now we're on Instagram, we're on social media, we're finding pictures of the dress that another person wore, and I want the exact same dress. How do we take those images and enable searching through those images? We're breaking those traditional approaches of searching and bringing them together in the problem solving. In the traditional approach, we threw the product problem, the tech problem, the business problem at our tools, and we got a solution. But what machine learning and AI offer back to us is not only that traditional approach but also models that allow us to solve problems beyond what we ask for.

[00:23:46] As a retailer or as a customer, we never asked, could you cross-sell me something else that I didn't even know I was looking for? I was there, I was shopping for it, and you gave me the bag and told me I could buy something else that would help me complete my purchase and my look for that night. Are you helping our businesses that think about how to optimize pricing, how to make it easier for our customers to get a better-value product? How do I fulfill these items in a much more efficient way? I have one store that has the dress and another store that has the purse. How can I ensure that our customers can get that perfect look where they want it and when they want it, and how can I solve those fulfillment needs?

[00:24:37] What machine learning and AI provided to us was not just solutions to the problems that we threw at it, but learnings from that data that allowed us to improve for problems that we didn't even ask about or know we were trying to solve. That's the progression that we're taking in that journey.

Challenges: jobs, bias and learning models

[00:24:57] No progression in a journey comes without a set of challenges. If you're part of an organization that has gone through that maturity, to the point where you're using data science to solve many of the problems, I'm sure you often get asked a bunch of questions. I sure do, and I've put up some common ones that I often get asked in this work. Is artificial intelligence going to take over humans? Is this going to take over all of us and replace us with these machines?

[00:25:31] My view on that is we're offloading the grunt work to these machines. Data crunching in large numbers, which we would have had to be reliant on ourselves at the time we were building these solutions for analysis. Can we offload the grunt work of stocking shelves, where we could use robotics and make that more efficient, and then use the time of ourselves, our customers and our associates where they can bring the greatest value?

[00:26:04] The next question I often get asked as a challenge is that these models are biased. They're biased against gender, they could be biased against mindsets. What I often remind myself and others is that models and machines are not biased; what's really biased is the mindset, the human mind. When we have the right mindset and the right culture to solve these problems in an unbiased manner, that's when we're going to solve them in the true manner. How do we remove that bias from the data? That's where we need human intervention. We want to make sure that we're consciously removing biases as we continue to train these systems to solve the problems at hand.

[00:26:52] And last but not least, how do you know these models are truly learning? Are they learning, are they evolving, are they taking feedback, are they improving? Are they going to continue to make predictions that are accurate? We can't do that without human support. We need human intervention to look at that data, look at the predictions, understand the change in environment, understand the change in behaviors, and then train these models to make better, bigger and much more efficient predictions.

[00:27:25] What I'm going to share with you today is, I hope you've had as much fun and learned, and at least nodded your head a few times as you related to these. My ride comes to an end now, but I hope you've had a good time. So unfasten your seatbelts, gear up to take your next ride through your own journey, and then go off and create new stories and awesome products. Thank you.