Episode Intro by Host

You are listening to Coefficient, a podcast brought to you by TheMathCompany and I’m Nirupama, a journalist, researcher and podcaster. Currently, the Content Lead for Thought Leadership and Podcasts at TheMathCompany. In each episode of Coefficient, we talk to experts, and deep dive into compelling topics related to the present and future of data analytics and AI for business intelligence with experts.

This is the second part of my conversation with Sayandeb Banerjee, Cofounder & CEO of TheMathCompany on what it means to be a truly data-driven business. If you haven’t listened to the first episode, I urge you to listen to it first to get the most out of this conversation. Let’s dive straight in.

Episode Conversation

Nirupama: So regardless of the route—like you said,  it’s not just one route—I’m understanding that some level of investment has to be made: in infrastructure, personnel, a lot of these physical kind of investments. But also, I would guess that there has to also be some sort of cultural change in the organization to go through this journey. So what would that entail? What sort of cultural changes and maybe even organizational structure changes?

Sayandeb: Thank you for asking that question because, I think this is one area where I have seen a lot of enterprises failing at the end of the day, right? While they have made serious investment in infrastructure—in personnel, upskilling them, buying the right software, et cetera, et cetera, but, the org structure and the overall culture—you know, like you said, data-driven…When I said every decision (needs to be data-driven), it’s a big cultural shift for an organization, right? So I think that is one place where I have seen many enterprises…while many have done a fantabulous job there…there are some who have not put enough focus on that area.

So first of all, what needs to happen? You know, this whole data and data driven decision making, if we have to make that real, this has to be elevated within the organization. So what it means is that there has to be a CXO role created. So that it’s not under a particular organization, or the second level under that one. It has to be as important. As a marketing organization where you have a CMO, and in a finance organization where you have a CFO, you have a sales organization where you have a CSO many times, right? Or a Chief Revenue Officer. Similarly, you need to have a Chief Data Analytics Officer.

You know, a lot of our companies are doing that today. CDAO as it is called, or CDO. But first of all, this needs to be elevated so that this whole business of data is a discipline on its own and not part of something else. And then, if you do that, then the rest of the org follows; where there’s a whole CDO organization or CDA organization who, in different ways can interact with the other stakeholders because they work with all of them. But I think, which is even more important…because this still happens and it’s happening more and more now. Look at many Fortune 500 companies, there is a CDO or a CDA role, and they’re doing some really big changes, as far as automation is concerned.

But you talked about the culture, the culture that is needed when you have to be data driven is a culture of celebrating failures. And I’ll explain why I’m saying that. A data-driven culture means that it needs a lot of experimentation. It means that, either I have evidence based on which I’m making a decision or I don’t have evidence, so therefore I have to generate evidence, right? I have to generate data by doing certain things.

How will you know (for instance) that walking the right path is better than walking the left path? If many people have actually walked both the paths, you can see how many have reached the destination and you can do a very simple analysis and say the right path has a much higher probability of doing that.

But if nobody has walked the left path at all, and everybody has walked the right path, then you don’t even know what it is. And the only way to answer that question is by actually walking the left path, knowing fully that it can lead to failures; that I will have—out of 10 people—at least three people walk the left path so that I can collect some data as to what happens. There’s a very high chance that I’ll fail. But remember that this failure, as they say, is the pillar for the success. That now, in unequivocal terms, you can say that the right path is the best part to go. If you’re proved on the contrary, nothing like it.

Many times, the outcome of a data-driven work or an analytical piece is either to make a new discovery; that is, in this case, the left path is a new discovery—nobody had walked that path before. Or, it confirms a known fact—that you have already been walking the right path. The fact that it is the right path has now been confirmed with data, so you can now walk that path far more confidently.

And can say, “look, we have explored all the possibilities and this is the best one, and I’m doing it”. It’s very different from saying, “oh, I don’t know what will happen otherwise, but I have always walked the right path”. It’s a very different approach, so please keep in mind that if an enterprise has to become truly data-driven, they have to do lots and lots of experiments, and when they’re doing experiments, there is always…some experiments will succeed, some will fail.

Obviously you’ll have to fail fast. You can’t wait for the full length. After six years, you cannot say, “sorry, we tried all of this for six years”. When you fail, I think in less than six months you have to see enough indications that this is going to be a failure, (and decide) so let me change course.

But those six months, if it gets penalized like, “oh, you did something which was completely wrong”, then it breaks everything else. No matter how much you invest in infrastructure and people, everything else goes for a toss. So how do you celebrate that failure? That that was a failure for a right reason, which will lead to many more successes in the future.
I think that mindset, that culture, it obviously has to start top down. It has to start from the leaders in an enterprise. But I know it is easier said than done. That’s something that I have seen is a big differentiator between the leaders and the laggards as far as this space is concerned.

Nirupama: So , again, something that I wouldn’t expect to be great advice, right? Celebrate failure. Thank you for sharing that. So my next question actually is to you as the CEO of one of the fastest growing companies in this space. We have been talking about what it means to be a data driven business. So would you say MathCo is truly data driven? Of course maybe there is no such thing as one company which has like arrived at that stage. But what steps have we taken or, are we taking right now towards that in the last few months or years?

Sayandeb: Yeah. Again, fantastic question. You know, the question is do you eat your own dog food or not? Right? Look, I think you already alluded to it, right? Let me first tell you the whole summary. We can always do better. And that is coming from someone who preaches to the whole world how you should become data driven.

So for me…if I say that we really, really kill it, I think I’d be lying. But having said that, I think from the get go, we have been very conscious about this fact that while we preach to the rest of the world, that you should be leveraging data, (that) data should drive your decisions, it cannot be a 180 degree shift, you know, a complete shift on the other side, as far as our internal working is concerned.

So what we have done is one, again, any major decisions in the company… which starts from forecasting demand. “This is the revenue that you’re earning last quarter. This is what you have earned two quarters back”. There is lot of good data-driven analysis that goes in to try and get to that prediction as well as you can. I’ll give you another example. I think we have now made it almost into a science of how a sales funnel for procuring new clients or, acquisition funnel as they call it, how that converges into a real closure. How many do you need at every stage of that funnel to have a client? We have numbers and data around it. We have trends on how it changes over time. Therefore, looking at a pipeline, I can say with certain degrees of confidence that this is what—six months down the line or one year down the line—total book revenue would look like.

So this is classic sales funnel analysis that we do for our clients, and we leverage this. I’ll give you internal examples. The operations team, the people in the operations team that we have internally, they conduct their internal meetings with data. There’s not a single meeting that happens just saying “oh, I have this observation, I have that observation”. If you have an observation, there is to be a data that backs up that observation. Even if that observation is about, “Hey, you know what, I think there are certain pockets in an organization where we need to do better engaging our colleagues”.

But that statement cannot be made in that meeting, without any data backing that up. And again, we have leveraged not just data driven decision making, but the accompanying technology with that. So, one classic example in the organization is that you have to listen to people to understand what’s really happening in people’s lives.

So while we have created this organization called People Success Organization that talks to people understand what’s going on. A big part of the People’s Success Organization, is a chat bot, a virtual agent that we internally call Amber right now. Again, leveraging the power of technology.

We collect tons of data through this chat bot and that data gets analyzed to help us understand what are the key pain points or what are the some pockets that we need to address and do better on. I can proudly say that a lot of, you know, initiatives that we take…So, there is a program right now that we are doing, which is…so you know, new joinees gets onboarded; there is an onboarding process for the first week. Now we are doing an onboarding when they start working for a particular client in a particular engagement, we are now onboarding them as a separate process for that account. This again, came up through some data, right?

We collected data and we realized that that is one area where there is some discomfort that is getting created. I’m saying there are, I think many areas where we are pushing to ensure that we are talking with data, talking with evidence. But can we do better? Absolutely. We can do better. So I think that answer will always be the same.

Nirupama: Right. I think from what you’ve answered, for a company that’s going to be seven years old, right? I think it’s a great place to be in.

Sayandeb: Absolutely.

Nirupama: Yeah. So, you are somebody who has—correct me if I’m wrong—around 25 years of experience in this space, right?

Sayandeb: Yes.

Nirupama: So, in your experience, what have you seen are the common mistakes that companies make, or business leaders make when they’re going through this journey of trying to become data driven, and what would your advice be to them?

Sayandeb: Sure. I think I have alluded on that a little bit already, but I’ll repeat that and I’ll emphasize on that or, or illustrate that a little more. So first things, first, one mistake that I’ve seen is where people are trying to become (from data ignorant to)—what you just said—data-driven overnight.

Sometimes I’ve seen some enterprises feel that there’s a magic wand that if I move my wand, suddenly voila! everything happens. So it never happens that way. It’s a journey. It’s a long journey. It’s a difficult journey.
So we have to be prepared to, you know, go through the right stages, take the whole organization along and things like that. So that is one very important part which I’ve seen people not doing right and therefore failing. The second one is what I mentioned about, this culture of celebrating failures.

While we all celebrate success, if you are not able to build a culture where you celebrate failures, it’ll be extremely different, difficult for that organization to really become data driven, because there will always be this sense of fear that,  “Oh, in order to become data driven, I have to try three different things”.

Some of these are new, “what if I fail?” right? And a part of that culture is also to give some amount of time and runway. And you can’t do an experiment once and say, “oh, sorry, it succeeded, or it failed”. It never happens that way. You know, in statistical terms, that is a problem of small sample, as we call it. So if you have a only few samples, that’s not a good conclusion that you’re making. So that’s definitely a pitfall that I would advise enterprises to avoid for sure, to become truly data driven.

So, as I said, don’t expect it to be a magic wand. It’s a long, hard journey. Now, let me put it that way. It’s a hard journey. It’s not easy, particularly for large enterprises. And second is this, you know, whole culture of celebrating failures… There’s another thing where I have seen organizations fail which is kind of related…is when there is a belief that, “well, as long as my team or my subordinates are dealing with data and understanding that, we as leaders don’t need to get into the details of the matter”.

Because particularly when it comes to data, sometimes it can tend to get a little bit technical, right? The technology that is used, the methods. But I think some amount of awareness. I’m not saying the CEO of the company needs to be a Python programmer, no. I know there are, really, really advanced companies in terms of data driven decision making where some of the execs can write probably the best codes in the company. So that happens. But I’m not necessarily saying that. But at least to understand the different data structures, what does it mean to have data on premises vis-a-vis in the cloud? Know what are some of the details of that, right?

A little bit more. How can data be leveraged? Why feature engineering, data munging is important? What are the different categories of operative models? What a linear model does. Why do you need a neural network at a certain stage? I think some amount of awareness and understanding also has become important because I’ve seen it drives a company many, many miles in a very accelerated pace, in that direction.

And if it is not there, then you are sometimes in trouble. Because while a lot of good work gets done, it doesn’t really go up and make a difference at the right level. So what I’ll also ask enterprises to watch out: Is your leadership taking enough interest into some amount of details when it comes to data and data driven work?

Nirupama: Okay. I think that was great advice. So we come to the last question. This is just something that comes to my mind. We’ve been talking about being data driven. I want to understand—by making a case for becoming data driven and by saying data will sort of dictate what I do, what are we making the case against? Could we possibly be making a case against instinctive decisions, you know: my gut tells me this is the right thing to do. Or, is there a possibility we could be unknowingly making a case against creativity and innovation and, some of the crazy ideas that might come out of it. What are your thoughts on that?

Sayandeb: So, it’s a good question to think about. So the first part that you said is…clearly we are making a case against that. When I talk about data driven decision making, It only means that you know, please don’t ignore evidence before making a decision.
Which means that just because your instinct tells you something or sometimes it’s not even instinct. Somebody may feel that, “look, I have for the last 20 years been doing this”. So, you know what is called gut-based decisions. “You know, my gut tells me that this is how it should be done”, and the gut also comes informed with all this 20 years of rich experience. I think we are still making a case against that because even if I have 20 years of experience, the way the world is changing, believe me—19 of those may not be even relevant. So then, instead of just relying on your last one year and…there are many examples around that…

Let’s take the world of marketing. If you look at marketing 20 years back, right? It’s very, very different from the whole digital marketing, social, you know—being present there; all of these were not even part of the marketing, playbook. At least in the enterprise side. People were a lot more focused on “how do I leverage TV, how do I do radio, how do I do print?”
Now, the bulk of that conversation has moved to digital channels and whatnot. So definitely making a case against them. Are we making a case against creativity and innovation? Honestly, I would say there is a danger of that. If you become very blind, you know, in terms of being data driven.

So, I’m saying data driven decision making—that is, leverage data to make a decision. I’m not saying let data make that decision. There’s a big distinction between the two. I am leveraging data to make that decision. There is still a human in the loop, as they call it. And that human’s gut and experience are very, very important. I’m a big believer of that, and I’ll continue to be a big believer of that.
See, there are certain things which are very transactional, very operational. The same thing gets done again and again. I think a machine can make a decision on that. I’ll give you an example. In order to approve a loan, all that a bank needs to see, at least in some of the advanced economies is what is your (credit) bureau score? And sometimes suppose the decision making is unidimensional. That if your score is above 750, you are approved. Below 750, you’re declined. Now this is a decision that a machine can make. You don’t need to have a human in the loop. But most of the time it is not as simple. Even if your score is below 750, because of who you are and some of your past records and other considerations, I may still be better off approving that loan. So there is a human component to that.

I think the case we are making here is: are you at least using data to know what is a baseline? And then there are many reasons why you might be making an exception. But do you know that you are making an exception from what the baseline is, and are you recording that? So that you can go back and analyze that every time I have made an exception, maybe all the time, it has worked out quite well, then there is something wrong with the baseline or whatever the data is saying. So you go back and look at that. But you might see that maybe 30% of the time you’re still right, but 70% of the time you’re not right. So that realization can happen only when you look at data, So again, I’m not saying data to make the decision.

I’m saying use data to make the decision, data driven decision making, and definitely not a case against creativity, because creativity absolutely comes into play because data gives you a baseline. And then it is your creativity, your innovation, where you know, you experiment and you try new things. And by the way, that becomes a new data also, right? So now you can compare what this data and the previous data…put them side by side and see—it’s so much more powerful. So let me do that: use data to fool creativity as well.

Again, creativity without data is like a boat or a ship without a sail. Like, it’ll just be moving here and there, right? In water. But if it is guided by a proper sail with a mast and all, it’ll go…while you are being creative, you still want to go towards a destination. If you’re a musician, even if you’re very creative, you still want to create good music, be it a nice tune, be it a good song, whatever it is. And for that, nowadays, by the way, there is a lot of data that gets collected that, kind of (tells you) when I do this combination of tones or notes, then what happens. If you are not being disciplined about it—and data is nothing but another word for discipline that builds into your overall approach—then you might not land up with anything. You’re very creative, you’re trying to create music. But ultimately, at the end of the day, even after a few days of effort, there is no music that has come out. Is that desirable? I don’t think so. Creativity, innovation has to lead to some outcome or interesting thing.

Otherwise, the creative person will also lose interest in creativity. So that’s how I’ll put it, that we are not making a case against creativity, innovation. In fact, we are leveraging data to make it even better. But definitely making a case against instinctive, pure gut-based decisions like “I know what it is; you don’t need to tell me”. I’m not telling you. Data is telling you; evidence is telling you and use that as a baseline, right? You make your call, deviate from the call, now that becomes also data and then look back and see is your deviation powerful? Was that effective? If it is effective, then you have a great discovery here, which tells you that this data was probably not right. It was not telling me the right story. So we have to constantly fight that battle with data as well and not become subservient that “okay, because data is telling me, that’s it. I’ll close my eyes and forget about everything”.

Nirupama: Right. One thing I realized from your answer is when I was asking the question, I used the word dictate—data-dictated. So that’s not what we want, right? What we want is data-driven.

Sayandeb: Yeah. Data should not dictate your decision. Yeah. Not at all.

Nirupama: So, yeah, if used well, data can empower creativity and innovation as well.

Sayandeb: Absolutely.

Nirupama: Okay. So that brings us to the end of the conversation. Thank you so much for taking the time, Sayan. It was great talking to you and hopefully we’ll talk to you again, in another episode.

Sayandeb: My pleasure. Thank you.

Episode Outro by Host

Thank you for tuning in to Coefficient. Hope you found this conversation insightful. In future episodes, we will be discussing more specific topics including retail media marketing, revenue growth management, and enterprise AI. Do subscribe to the show on Apple podcasts, Spotify or YouTube—wherever you get your podcasts—to be alerted when new episodes are released. Goodbye, and have a nice day!