Purvi Shah

Rahul Abhyankar [00:01] Hello listeners. This is your host, Rahul Abhyankar. My guest on this episode of Product Leader's Journey is Purvi Shah, Vice President of Product Development for Enterprise Data Platforms at American Express. Prior to that, Purvi worked at Deloitte, where she led a team of consultants to work in the public sector with multiple state governments to implement programs for citizens in the areas of acquiring food stamps, child care and other interventions. I asked Purvi what she learnt about empathy while building applications for people who needed such services. That's where we start our conversation. Later on, she was very much willing to participate in a thought experiment of applying Newton's three laws of motion to enterprise data platforms. That was quite insightful and a fun brainstorm to have. With that, let's hear from Purvi Shah.

Purvi Shah [01:05] I was consulting for various public sector clients, really helping them think through how do we take public policies and make them real. So, for example, the food stamp policies, how do we create an ecosystem that helps the citizens of the particular state and make that process streamlined so that they can get the food stamps that they acquire and then they can go about their business very efficiently and quickly?

A lot of the learnings from that point of time, I would group it into three things that I still carry forward today. One, problem solving. You're always constantly having to deal with a new set of problems that you have never anticipated. So even in public sector programs that we helped launch, we were sitting with our customers day in, day out, whether it was multiple different case managers across multiple different counties, bringing them together, really thinking about how do we understand what the current processes, what the current challenges are, what are we addressing? And then, most importantly, why are we addressing it so that we're able to come up with the right solution? And then the last thing I would say is looking at the front and the back, because the problem by itself is not a problem, but you have to understand the end-to-end journey to be able to clearly come up with the solutions for it.

Rahul Abhyankar [02:23] Can you explain what do you mean by front and the back?

Purvi Shah [02:26] When I was working within a particular state, the problem statement was how do we optimize and create efficiencies in how quickly somebody can fill out an online form? Now, if you just take that as a problem statement, what you're missing is the prerequisites that need to go in before somebody can even get to that journey. The prerequisite was, they needed to make sure that they have the identity card, they needed to make sure that it's scanned. And back in 2000, it wasn't as easy to take an iPhone and scan it because iPhone was not even around. So you have to think about what would that journey look like before you even got to the problem statement. And then the after the fact is, at that time the requirement was, take the information and mail it in.

And now how do you take that case information, print it out in a way that there has a barcode and a scanner that they can append additional supplementary documentation to, and then take it to the post office or UPS to be able to mail it? So that's what I'm talking about when I talk about end-to-end, that you're not only solving for a specific problem, but you have to think about the journey as a holistic piece and then understand what area you want to focus on.

Rahul Abhyankar [03:43] That's beautiful, because there is always that aspect of what I love to call as before, during and after. You have to take that entire spectrum of things into consideration and not just solve for what people are doing during the experience of when they have that form, the form that you designed in front of them. And I imagine that building applications for people who are looking to apply for food stamp programs or childcare and intervention programs, these are such emotional scenarios that if the application process does not go through seamlessly, that adds on to the level of emotion that people trying to do those things experience. So did you have any aspect of just feeling that emotional level that users are at?

Purvi Shah [04:36] Rahul, I think you struck on the chord because that was such an important learning from my time at Deloitte, which is leading and solving with empathy. The reason why I say that is because, no matter how big or small the change is, there are a lot of emotions that everybody faces from different aspects. So it's important to, A, know that they're going to be there, B, acknowledge and recognize them, and C, address them in a meaningful way that the other constituents understand that you have heard them, you have understood them and you're trying to actively solve them. If you just look at data in isolation or if you just look at the process in isolation, you will miss out on a very important part of the product that should be there, that should be delightful, that should want them to use it. That was a really important piece that we learned, because we assumed a lot, at that time at least, in creating the journeys. Oh yeah, it's easy, you click on a button and the web page will load. Well, you have clocks that are loading and things like that, and how do you move through that? So it's important at that time.

Rahul Abhyankar [05:49] And we throw around this word empathy a lot as product people and understand where users are and so on. We can never, ever really be in the shoes of another person because there is so much context that we will never have. So I love what you said about, one, recognize the emotion that they're experiencing and acknowledge the emotions.

Purvi Shah [06:12] It's the basics.

Rahul Abhyankar [06:14] So let's jump into the topic of enterprise data platforms. From Deloitte, you came to American Express.

Purvi Shah [06:21] It was an easy transition because I was already working on a lot of big scale application implementation. I quickly moved into product. And then, within product, my natural sense was loving to solve problems, and the biggest problem that we have today at most over 170 years old company, our data, our businesses have grown in silos, but there is a need to bring that together so that we know, hey, this is Rahul, Rahul had this type of merchant relationship with us. So how do we truly create that 360 view on a customer, both individuals like you and I, as well as businesses, so that we're able to really serve them better? And then my current role, which is enterprise data platforms, in that I have the responsibility of leading product for our big data ecosystem, globally managing all of the data quality and data management tooling for the enterprise. And then, once the data is available, how do I empower more than 15,000 colleagues across American Express to actually use that data through our business intelligence tools? So that is my current function.

Rahul Abhyankar [07:43] So when you embark on a Customer 360-like program, obviously there is an overarching mission of having that single source of truth for how American Express understands a customer or how any company understands a customer. But when you define the success criteria for an initiative, what metrics do you look at? What indicators do you track? How do you know that we've done what we set out to?

Purvi Shah [08:12] Customer 360 is unique in a lot of ways where you're never really done. But let me back up for a second. When we first started creating Customer 360, we had set success criteria. Number one was the two or three top use cases that we're really focused on improving. Were we able to bring together data across 18 different ecosystems?

We have 18 different ecosystems that held demographic information, individual relationship information. How do we combine those into one? The second one was, can we make sure that when these use cases are coming on board, we're creating them in a way that can be extended and augmented very quickly into different types of journeys? So, for example, the application journey, could I make it easier, by the fact that I know you as a customer, to pre-fill? That allows you to have a seamless experience. So it was a use case.

And then the third part of it was really understanding the quality of the data, because it's easy to say, okay, let me just take 18 different tables that I have across 18 different ecosystems and stitch it together. But the reality is there's a lot of noise in the data. And how do you go about cleansing that data, matching that data and then giving it a score so that you know this percent is highly accurate, this percent we're in the gray area, and then this percent we cannot even match.

And even at American Express we're always thinking about, what is our data landscape today? What are the regulatory considerations that we need to plan for? And then what does that tech stack look like? Because it's not a singular tech stack. It is a combination of things that come together, different types of toolings that solve the problem for the end-to-end data ecosystem, because it's not just one thing that will solve the entire challenge that you have with data, all the way from how do you procure the data, the lineage, the metadata, to how do you apply data quality solutions to it, to how do we bring that data at rest, how do we make it available in real-time ecosystem, what are the ways that the individual at the end of the day will use the data? Is it through visualization, BI tools? Are they going to do Excel spreadsheets? Are they going to do AI, ML models, for example? Are they going to use it for Gen AI? Are they using it to understand the customer that is on the other side of the webpage, to provide them with the most relevant merchant offers, for example? I think that's the important part. It's not necessarily one solution itself, but it's a multitude of solutions that you need to stitch together.

The ask is, what are some of the things that you're already good at and you build on, and where do you complement it with something that is more external?

Rahul Abhyankar [11:16] I was thinking about our conversation coming up and I was thinking maybe there is a way of looking at Newton's three laws of motion and see how that applies to enterprise data. So are you willing to go on this thought experiment?

Purvi Shah [11:26] Let's do it.

Rahul Abhyankar [11:27] All right, let's see how that goes. So Newton's first law of motion is about inertia and says every object continues in its state of rest or state of motion and an external unbalanced force is needed to move it. Now, when we apply this to enterprise data—you mentioned 18 different systems that had a view of the customer data—there is a lot of inertia in enterprise systems and data platforms and how data stays in an enterprise. When you think about overcoming that inertia, how do you think about that?

Purvi Shah [12:02] The first law is so critical, and the way, as you were talking about it, the way I visualize it and I brought it to life is, for example, with Customer 360, when we created it. Why did we create it? It was for two reasons. We listen to our customers and we listen to our colleagues, and that, in my mind, is the force.

Our colleagues were saying, hey, it's impossible for me to serve Rahul as a customer in the best way possible, and our company's goal is to provide the customer first, and if that's the case, having five different screens with five different applications running to be able to tell the information most effectively, efficiently, accurately, that's too much. I need to be able to have a singular place, easy place to do it. So our own colleagues are saying, we need something better, we need to break down these silos, we need to make it easier for us to achieve the objective that American Express has set out for ourselves. And then the second is customers. Nobody wants to wait 24 hours to get their decision back if you got the increase in the credit line that you requested or not. Everything needs to be instant. So in order to meet and delight our customers and continue to exceed their expectations, that was really a big force for us, a combination of customers and colleagues to move our data platforms forward.

Rahul Abhyankar [13:30] There is an aspect of identifying the business case and the benefit that you can create for internal as well as external stakeholders, and that becomes the tipping point to overcome that inertia. Beautiful. So let's go to Newton's second law of motion, which is force is proportional to mass of the object and the acceleration or rate of change of velocity. Now, when we think about this in the context of enterprise data, I see mass as just the size of the data, complexity of the data. So much data is available about any customer and millions of customers within American Express. And then when we talk about the rate of change, velocity, momentum. How do you look at those things and say, well, how does that apply to what I'm doing at American Express?

Purvi Shah [14:23] It is so applicable, because what you said is so true. We're producing a lot of data within American Express ecosystem from the experiences that we have, but I think for us, the biggest thing is two things. How do we protect the American Express brand and make sure that the data is secured, is used in a permissible format, that it does not unintendedly get used in an incorrect way? That is a big focus for us. Having a good data ecosystem that allows us to capture that information and then leverage it in the right way is one big reason that we're doing what we're doing and protecting the data as it grows very, very quickly. The second one, again, is the company's vision. We want to grow double digits. That is a big goal for us for next several years, and data is at the heart of it.

So, in order to understand the data that is rapidly evolving, and how do we provide the best products and how do we evolve our products and the strategy to meet the customers where they are is the second part of this. How do we make sure that, if you're traveling, that I provide you the insight into, hey, you are just checking into JFK. We have American Express Lounge because you have a platinum card. This is a five-minute wait. Go ahead and go get over. You have two hours before your flight takes off. Oh, your flight is delayed. We're automatically calling you. So how do we take all of that data to stitch and provide the best experience in the most delightful way is something that we're using data for, and how do we go through that area over and over again to make sure that we're able to delight?

Rahul Abhyankar [16:12] Let's stay on that second law for a little bit. When we talk about the force that needs to be applied to create the change of motion and the direction of motion, obviously Customer 360 enterprise data platform initiatives are non-trivial, significant effort. So that's the force that you're applying in this case. So how do you estimate what force is necessary? How do you start to estimate the sheer size of the effort and the complexity of the effort?

Purvi Shah [16:43] It's a really great question, and maybe I take you down a little bit of a different path and bring to life how we decommission something. So we had a 25-year-old platform, a data platform that served us really, really well, and our expectation was, great, we're moving, we're introducing the new data platforms and these things will go away. And then what ended up really happening is we had two forces that really helped us. One, top-down, who basically said we're going to put a foot down, this is the date that we're going to do it, but here are the resources. We're not going to compete with other priorities that you have in terms of growth and things like that to really clean up our data platforms. That was number one. And then number two was just regulatory changes and the expectations on how we govern and manage our data. If you had to manage both of the platforms, or three platforms, it becomes really really impossible to do. So that helped us decommission something. I just took you down a little bit of a different path to show you a different flavor.

Rahul Abhyankar [17:45] I think that detour was useful because it also relates to the inertia, because you've got legacy systems that you expect will be retired at a certain point in time when everybody jumps on the bandwagon of the new systems and starts to adopt those. But that migration doesn't necessarily happen in the way or on time that you expect it to.

Purvi Shah [18:11] It's so true and you hit the nail on the head, which is everybody wants to move to the net new because it allows you to do things that your old never allowed you to do. That's what we saw even in our experience, that a lot of new ways of solving problems, new use cases, how do we use more machine learning, became more relevant in the new, and everything that was legacy reporting and things that were working fine, that just kept running. So it is true that everybody migrates, but not necessarily all of the old processes that have been running in the data platforms migrate at the speed and then the rate that one expects.

Rahul Abhyankar [18:53] So when you went through the process of decommissioning these 25-year-old legacy systems, what are sort of key advice that you could give to listeners who may be in similar situations, trying to manage the old and the new, and what are some lessons that you have learned that you could share?

Purvi Shah [19:11] There are five key lessons that I would, Rahul, say were the most important ones. One is, set the tone, get the buy-in from the top. I think that's the most important thing that you can do. We decided, we as a company are 100% behind it. This is the date. We're not changing it. We are keeping it the way it is. So being very consistent and unwavering has a huge benefit, and so for programs like this, that's the number one thing I would say that we need to do. The second thing is provide the resources that is required so that it does not compete with other initiatives that become more important. You don't want to compete with growth, because I can assure you, migration will lose out to growth in terms of value, but there are a lot of benefits for doing it. So having a leadership team as well as your stakeholders understand that we're not compromising on growth, but we're having dedicated resources focused on this effort, is an important thing.

The third one is, I would say, communication, communication, communication, and I don't underestimate that, because as you would do with any new product, decommissioning needs to be served in. This would be the benefit of it. Same thing needs to happen with decommissioning, and in our case, it was an 18 month long journey, but we communicated through multiple different modes, whether it was through big TV screens that we took over in our offices with a countdown clock, to emails that we sent, to town halls that we attended, to one-on-ones, to bi-weekly meetings, you named it, we did not leave a single stone unturned on that aspect. So the communication is super important.

And then the last thing is being able to celebrate milestones and victories and giving the platform for the users that needed to go through that journey to celebrate. That became really important for us because, as you can imagine, while we own the central data platforms itself, the use cases that we're running were all federated and a lot of different teams were running these processes. So how do we make sure that, if a particular group went through and went through the decommissioning process, how do we give them the platform to say, hey, not only did I do it, but I saved us X amount of dollars, or I improved the customer experience in a different way? So I think that was another big thing, which is taking a moment to celebrate, not waiting to pop the champagne at the end of 18 months, but throughout the journey. So I think those were the huge lessons that we learned that I think can be applied by all of your listeners.

Rahul Abhyankar [22:05] Wonderful. So let's come to Newton's third law of motion. Every action has an equal and opposite reaction, and so I think of it as opposition that comes up to some of these initiatives, not intentionally, but just unexpected opposition, unexpected situations that come up that might derail these initiatives. Any reflections on that, and what can you share?

Purvi Shah [22:46] Two or three things that come up. Let me break that down. In the effort of decommissioning that we went through, not giving ample time for training and upscaling and investing in colleagues, can have a detrimental impact on your journey. So I would say expect a lot of whining when you're going through these types of changes, whether it's decommissioning or a new product feature, and then be ready and prepared with the tools that you need to make it available to the enterprise. So in our case, it was how do we make training easy? Can we have cohort learnings? Can we do self-serve learning? Can we make sure we have office hours? Can we make sure we have train-the-trainer programs? Those are the types of things that we did to make sure that we were able to achieve our target. So that's the first and foremost thing. That is the opposition, which is, hey, this is too hard. I've only learned how to do this in this language. Are you telling me I have to quit my job? That's not really the case. We're giving you the tools to upscale you and, as a company, American Express is investing in you. So that narrative and that effort to upscale and uptrain is really, really important.

And then the second thing is overhead. With any data platform changes that we have done, introduction of new processes, there is a thought process that, oh my gosh, I have so much overhead, you've just increased my work. If I double-clicked on that a little bit, as our data governance and data management programs have evolved and as regulatory bodies expect companies to do more with data, protecting the data, with GDPR, or the Reserve Bank in India, or even our California laws, how do we make sure that we have the right set of governance around the data that we're securing? So we're expecting our data producers to go through additional steps so that we can catalog it appropriately, so that we can have the right level of metadata, so we know this table has these types of data and, more importantly, we can drive consistency across the company. With that, the perception is, oh man, this is so much overhead. But I think you need to keep telling the story that, while this may feel like it, here are the tools that help you accelerate, but then here's the benefit of doing it in the right way that helps us protect the brand, helps us serve the customers better and help serve our colleagues better.

Rahul Abhyankar [25:26] Purvi, you talked about the success criteria for Customer 360. Similarly, are there success criteria for enterprise data platform initiatives? And maybe one example could be, there's so much data available within American Express. How much of it are we really using? Is there a way to put a score on that?

Purvi Shah [25:49] Actually, that is the first and foremost thing that we're doing, which is, when we're thinking about bringing data into our ecosystem, we're thinking about, is it going to be leveraged and is it meaningful data that can serve multiple purposes? I think that's an important part, the phrase multiple purposes, because what we want to do is we don't want to keep deriving data sets over and over again. We want to create the curated data set once and make it available for multiple different use cases. But in doing that, the pre-work is, can I tell that this data has been leveraged by multiple different users, because we have a lot of the logs that helps us understand that. And if it's not used, if it's used by two users in the whole year, is it really worth bringing that into the ecosystem? Because there's a storage cost, there's an overhead cost, there's a compute cost across the board. So we were being very surgical and deliberate in terms of bringing the right set of data that is going to be used. So that's one of the big criteria that we're focused on.

And then the second one is what type of new value that we can unlock that we have not been able to do across the old ecosystem. In the old ecosystem we were using a lot of data at rest that came in back jobs that were stale, that were not real time. By thinking about the new data ecosystem, are we able to leverage more real-time events, streaming data that can allow us to quickly tell what is happening in the marketplace. Or it could be, providing the experience that allows us to say, Rahul, you were just visiting our website. You seem like you're interested in this particular product. How can we provide that product to you more seamlessly with an offer and things like that? So, really using that as the litmus test to say, are we able to unlock net new things that helps us grow revenue, helps with our growth targets across the board.

Rahul Abhyankar [27:56] Awesome. Well, Purvi, this has been a great conversation, really insightful. This topic is vast and we could spend hours and hours talking about it, but really appreciate what you've shared. You've been willing to go on this thought experiment of how to apply Newton's three laws of motion to enterprise data. So really appreciate you being on Product Leader's Journey and thank you so much for your insights.

Purvi Shah [28:21] It was great. Thank you for having me, Rahul. I really appreciate it.

Rahul Abhyankar [28:25] Hi there, this is Rahul Abhyankar. Hope you enjoyed listening to this episode. You will find the notes at productleadersjourney.com. Subscribe to the podcast and, if you like it, share it with your friends and colleagues. See you soon.