I think the true way to add value to any company is to be proactive and is to be strategic. Data is only as valuable as the action that you can drive with it and the impact that you can get from that action. Welcome to Data Center and Karthik Co, Founder and CEO of Babbage Insight, where we are building our proactive insights engine. This season of data chatter is all about talking to industry leaders to find out how they use data in their day-to-day decision making.
So to start off with, can you tell us a little bit about yourself and also what job that you do? Yeah, absolutely. I'm happy to be here and excited for this conversation. Karthik, my name is Aswan. I lead the FB and a team here at Quadrics FB and a, for those that might not be familiar, is a team of financial analysts and managers who manage the short, medium and long term financials
of a company. My team is responsible for managing the expense envelopes, budgeting process on the cost side, but also on the top line side. So my team is always focused on data and is always looking for insights to help the company make better resource allocation decisions across the board. So yeah, excited to be here. And in terms of maybe a quick history, I've spent the last like 10 plus years in various finance roles across tech
companies. And prior to that, I've spent time in the private wealth industry. So really close to data across the board through my career. Thank you. So I assume that being a finance person you, my assumption is that you'll be using a lot of Excel, but can you talk through how you get the data, what kind of data you look at, what kind of analysis you do and at what frequency and so on as part of your day-to-day job?
Yeah, I work with a team of about 35 ish analyst and managers and at any given point that are probably 10s of analysis that are ongoing within my team that ultimately helps the company make better decisions. And the scope of the data that my team works with is pretty
expensive. So it could start all the way at the at the very end of, if you can think of the value chain of revenue, it could start with cash, it could work backwards in terms of Billings. It could look at like revenue, it could be about annual recurring revenue or ARR. And then you could keep going further upstream in terms of that data. We could be looking at customer transactions, you could be looking at new customers like upsell. We could be looking at number of
customers. And then even further upstream, if you think about in the tech industry, a lead to cash sort of value chain, like we go upstream into number of leads spend against marketing programs. So the data that my team works with is really expensive and it could be very varied depending on the time of the year, depending on the time of the month, depending on the specific member of my team.
A few examples could be that the business is considering a potential short term sales incentive to accelerate a certain portion of the business be a product or region. So my team is working through what could be the impact of an incremental dollar in sales compensation that we spend on that product or region and what is the impact that we should be expecting and what would be the ROI of that incremental dollar that's spent. So this could be a very quick
analysis that my team does. My team would also work on like very long term strategic analysis, like how much should we be committing to with some of our largest like cloud or server vendors in the next like three to five years? Because we want to look for long term contracts because we want better economics and discounts from our service providers. So this could be around sizing that and getting confidence
around that. So My, my, yeah, a whole variety of different data problems and challenges on different time horizons working with different business partners from product and engineering or like go to market, sales, marketing or even like our GNA functions, thinking about allocation strategy, thinking about like benefits. So the challenges really depend on the the time of the year. Planning means different
problems. Ongoing execution through the quarters means different kinds of problems that my team works on. Got it. And can you talk a little bit about your text stack in terms of like how you get the data, how you do the analysis of who are your stakeholders, to whom we have to present your numbers? How? Yeah, this takes through the
process. Yeah. So our CRM system today is we use Salesforce and the data from Salesforce is dropped into a database solution, I believe it's Redshift. And from there, we have plenty of analysts across the team who work directly with sequel queries, pulling the right data from the databases. We also have Tableau, which is the data sort of visualization reporting layer that some of the analysts on the teams use. And yeah, it, I, I wish that there was more consistency.
I wish there was more interconnectedness of the different data elements that are coming upstream from the different systems, which is one of the challenges that we struggled with today as a company on how spread out our data ecosystem is. And sometimes pulling together a simple analysis would require cobbling together different data sources to get the right insight.
The other thing that I'm working through with my team within the company is democratizing data and having various analysts across the team get access to this data versus a specialized individualized data team being sort of the stewards of this data and supplying that data to various organizations. So democratizing data is going to be critical for at least where I sit today in terms of getting the most amount of
impact. The stakeholders that we interact with is basically every business leader across the company who has goals that are financial either top line or bottom line driven that are either looking for how do they accelerate their business or how do they allocate their cost dollars in the most impactful areas. So that's any leader across the company who is wanting to push the envelope on top line or or bottom line.
So this could be the chief security officer, this could be the chief like IT officer, it could be any like senior sales leader, It's definitely the chief operating officer, the chief people officer. So I have analysts on my team that will partner directly with each of these leaders in terms of understanding where their biggest pain points are, where their biggest opportunities and challenges are.
So it's literally every single business leader across the company who works with an FPNA partner to make prudent financial decisions for the company. Got it. And the analysts of your team might imagine to all the data work themselves that do that. Basically you have in house the part of the data and seeing that you require for your job I assume.
Yeah, they they do the, the analysts do the most of their data pulling the analysis, the insight generation and the partnership to go convert that insight into action so the company can see impact. All of that happens in house. As I was saying earlier, I wish the the tools and the interconnectedness of the data was a lot better.
If it were, then the analysts would spend a lot less time pulling data, but be spending a lot more time on digesting the data into insight or even like using the insight to drive action for the company. Because ultimately data is only as valuable as the action and the impact that we can get out of it. So over time I hope for better systems and and data pipelines that helps the team move up the value chain into action and impact.
Can you elaborate a little bit about this interconnectedness that you've been talking about? So, So what is the precise problem here? Is it that you have fragmented data sources or is it that, like you, the team struggles to sort of reconsign data across sources? Because I assume that as a finance team, you guys must be fairly sort of you want stuff to be accurate and things like that based on my experience in finance and so on. So how does it work? So what are the issues that you
face right now? Yeah, Quadrics has been on a journey in the last like several years depending on how much you follow the news, which means that like going public, being bought over, then being taken private. The company's priorities from a data perspective have changed meaningfully over the years, which means that we have different systems, different implementations that does not
set up for success. Cobbling together data from different sources or even the movement of data from one system to another, say a Salesforce, which is our transaction system, CRM system to next week, which is our general Ledger. There's a lot of information that either gets lost or is mistranslated that makes data less compatible across some of these systems. And the the unique identifiers that you'd want to use at a customer level or at like a contract level could also be different.
That makes combining different pieces of data across the value chain very difficult. And that makes the insight that we can get from some of this data a little less intuitive. So it just takes a lot of effort to prepare the data from these different systems to then be able to gather or glean insights off of this data. So that's what I mean by
interconnectedness of data. It's a consequence of different systems, different implementations that were either done perfectly right or not not at all. That makes data from the past less compatible and less immediately useful, without spending a lot of time first preparing that data first. Got it. And how, how does this compare to your previous organizations? Because I think you've worked with slightly sort of more tech
companies in the past and stuff. So how does the data experience from a finance perspective compare? Yeah, I think it really depends on how Core Data is to accompany in how products are built and how systems are built and how
the company chooses to scale. I I genuinely believe that if you look at a company's existing data stack and their implementations and how and the tools that analysts across the company used to get insights from that data, you'll be able to tell how much focus there was on data in that company's
history. In my past, I have worked for companies that are really massive, that have invested like years, maybe even like over decades in cleaning up that data infrastructure and creating better end user tooling for analytics that makes using that data incredibly straightforward. But that's after like years and years and maybe a decade, over a decade of investment in that space.
I've also worked for like similar mint to large capital companies that have been built with the ethos of data and using data to make better decisions that make a lot more data accessible to a lot more teams, a lot more metadata about product usage or the customer usage or like a go to market like Rep engagement with a customer. All of that data can be helpful to make better decisions for a company. So I've, I've worked at companies that have done that right.
So yeah, it, it really depends on how much focus and investment that has been in ensuring good clean data. Ultimately, it also comes back to the business rigor and the financial, I would say acumen of a leadership team as well that creates the investment in data over like many years because it takes a lot of work. It takes a lot of work to retrofit broken or disconnected systems and data pipelines to intuitively and like immediately make sense of them.
So yeah, in my career I've seen very different levels of that interconnectedness of data and the readiness of data to be made into insights that helps decisions. Awesome. And talking of insights, I understand that there are two frameworks that you use to kind of when it comes to data. So can can you sort of mention that first and then like elaborate about them in terms of like how you look at data and data this is making? Yeah, the do for frameworks that I share with you offline.
The first one is I mentioned briefly and like one of the earlier questions, data is only as valuable as the action that you can drive with it and the impact that you can get from that action. So I, I spent a lot of my time thinking about with my team, how do you use data to get insights about the business problem or opportunity that you're looking to solve? And how do you use that insight to come up with meaningful recommendations for the business
to action? And then how do you action those recommendations? It takes a lot of influence to do that because having the best idea and the best data and the best insight does not mean that will immediately be accepted and action on. So there's a lot of influence that goes in a cross functional matrix large organization to get that insight into action. And once you have the action, then you have to track the impact of that decision that you made. Then you learn from those
actions. So the next set of insights that leads to the next set of action is going to be accreted to the company and that the company is getting better at this. So to me, it's very, very important thinking about this value chain of data to like impact.
I think there's a lot more focus on the rigor of data and like data analytics, but I tend to focus a lot more of my time on insight to the impact part of the value chain as well, because that's how you realize the impact of, of any sort of analytics or insight that you put together. That's one. The the second one and it's, and this one also sort of relates to the first one, is the impact that you can achieve with the data that you have.
You have to calibrate how much effort or time you spend trying to solve that problem or opportunity. You might be familiar with a Pareto principle of 8020 that generally holds true in more situations, 8020% of your customers bring 80% of your revenue. 20% of X give you 80 gives you 80% of block. X is the import and like Y that is the output of the outcome. I feel very strongly about that in terms of data analytics as well.
For most business problems or challenges that my team tackles today, I believe our 20% effort will get us 80% of the impact on that answer and majority of the situations that is going to be enough to drive the right decision. We in, in the work that we do in FBNA, we don't aim for precision. One of my finance professors from grad school, like 10 plus years ago told me once that in finance, like all data is wrong. Like you forecast something,
it's going to be wrong. Like there's no point obsessing about precision and accuracy. So for the most of the work that my team does, it's 8020. It's like put 20% of your effort into this, get 80% of the outcome. Let's move on. But that is a scale for me and that that's the framework that I talked to my team about. If it's 20% impact effort for 80% of the impact, if you put 10% effort in, you'll likely get 60% of the impact. Or if you put 2% effort in, you'll get 20% of the impact.
So it's, it's sort of a sliding scale of how much effort should you put into this problem. And it is super important because any data problem or challenge or opportunity can be solved in 5 minutes or five hours or five days or five months, or you can even take five years, depending on how academically rigorous you want to be about that problem or how accurate or precise you want that solution to be.
An example would be like, if you're launching rockets, yeah, you want, you want to spend five years. You want to be absolutely sure if you're, if you're saving lives, if you're like a neurosurgeon, like, yeah, you want to like put all that effort into like practice because you get something wrong, like it goes wrong. So you there you want to have a 90 9% effort for like 99.999% of
the impact. But it actually matters in finance a lot because I am almost always pushing my team to calibrate less on the effort scale. I'm usually asking my team for like, hey, give me a 2 to 5% effort that gets me 20 to 40% of the impact. An example might be a business leader wants to know, hey, if I want to invest 10 resources in this problem, can you tell me what it's going to cost? You could probably solve that with mental math of average, like cost per head.
And then like when do you invest? And like you could have a response in 10 seconds. And for that business leader that is going to be sufficient. Or you could take the entire next two days to go pull specific like cost at like a specific level, like what location, what like level. It's not going to be helpful for
that business leader, right? So it is really, really important to understand and calibrate how much effort you put into solving data problems for the impact that is possible to achieve in answering that problem because that is a cap and any investment of effort or time beyond that is not going to be useful. So that's the other like framework. And I've tried to make that a part of my team's terminology on when I ask a question or when I
ask someone to do an analysis. I'll usually say, give me a 5 to 10% solution on this, which means that they know that they don't need to spend days and that I'm looking for something quick. Something quick means that I also need to give them grace if the 20% answer is slightly different, which is OK because the answer can be a little bit different and we'll tweak as we go.
So that's the second framework that I found really helpful in working with my team to make sure that they're allocating their time and resources in the best possible places. That's awesome because my experience with finance has been that I will say the cost is 100 and they'll come back and say
no, it's not 100, it's 99.8. So because I think some finance fee will come from the, I don't know if it's because they come to the control side or if it is like when you're preparing your financials, you look at numbers from a certain kind of precision. And for FPLA you probably don't need that because like you, what you need is a sort of a broad direction in terms of like what the company needs to be doing. It's more of a strategic role, right?
Absolutely. And sometimes I'll even say, hey, can you give me a sizing of this cost? Like ±5 million is fine. Like, because beyond that, I don't care. I don't care for it to be more precise than that. Or I might say, oh, this one's important, so give me something that's a little more accurate. So you're absolutely right. Like there's no point debating 99.8 versus 100 because there's no value.
There's no value. We're not going to make a different decision as a company whether it was 99.8 or 100 or even 98 for that matter. So, yeah, yeah. Slightly changing tracks, how do you sort of like look at, let's just call it a data analysis, But I mean, FPNA job can be broadly described as data analysis in some sense being proactive versus reactive. Because what I've seen in the past and also in the present is that like analytics teams can
come in two flavours. 1 is where in your job it could be a team coming to you and say how much do you think we should be able to we should spend on this new project? Was this your team proactively monitoring the numbers and then going back to some team and saying that OK, this project is tracking as per expectation thing so and so how do you see the balance between proactive and and reactive insights when
it comes to data? How does it work with your team and other teams that you that you work with and so on? Yeah, I think the only way to do finance is to be proactive. It depends on the maturity and the capabilities of an organization and where they are on that journey. But on a long enough timeline, I think the true way to add value to any company is to be proactive and is to be strategic. And by strategic, I mean you are looking to influence the future. That to me is strategic.
If you combine proactive and strategic together, any finance person will be able to bring so much value to the business that they support. To do both, you need a deep understanding of the business. You need curiosity, you need empathy for the organization, the team, the business leader, or the function that you support.
Once you have that understanding and that empathy, waking up in the morning and thinking about what can you do today as a finance person to help this business be better or help this business leader achieve their goals? To me, that is being proactive. And that's the only way to do it because a reactive mode is like you're, you're solving for like what is in front of you. You're prioritizing the things that are on your plate. Sometimes you, when you do that,
you miss the bigger picture. You miss the opportunities that are not in front of you. There's a lot of availability bias. When you're reactive, you're at that point like you're already on the back foot. You're, you're prioritizing or like you're saying no, or like you're thinking about the value of the work that you do.
So stepping out of that mode and getting into a proactive mode because once you start thinking about the ways that you can add value from the data to insight to like action to like impact, the better position you will be to bring strategic value to an organization. So to me, as I said on a long enough timeline, not just like finance, I think about that like
any function. If a business leader be ACEO, be it a Chief Operating officer or be it AVP of sales or Director of customer success, they're looking for partners with the capabilities of translating data to insight to tell them what they should do differently. And that to me is proactive and strategic. So yeah, I think that's the only way to bring true sustained value, in my opinion.
How has AI changed your role in the last couple of years and how do you see it changing your role in the next few years? I think there's a lot more change coming in my role in adjacent roles, especially as I think about data to insight, like inside the action, I think is going to take a lot longer, especially because these aren't like system actions. These are like business decisions that need to be
contemplated in a broader scale. Coming back to the data to Insight, AI hasn't changed that very much just yet because a lot of the AI evolution that we've seen in my opinion in the last like 3 years has been on language models and that has to do with language writing content like documents, structures. The data capabilities of LLM models is non existent. I don't think that's their purpose. 10 years ago, I worked on a machine learning predictive forecast for a large tech
company. And I don't think the machine learning or predictive modelling technology has meaningfully changed in the last three years. And but I do think that there's a lot more that is coming in AI that is going to evolve that significantly. The other vector to think about is when the company was incorporated, what is the text stack look like? Again like interconnectedness of
the data. If you want to put an agent or another like SAS, AI driven like SAS application on top, if you don't clean up the foundations, if you don't have a tight foundation of data, it is not going to work. So I think majority of the companies today have to retrofit AI into their existing landscape, which is going to be a humongous effort. And I think that's going to change as well in the next like 2 years.
I think there's going to be evolution and technologies that makes it a lot easier, more autonomous, more agentic. So I talked to my team a lot about how AI is going to upend the FBNA job in general. I don't have a sense of how that is going to happen yet. I keep my eyes and like ears to the ground on the evolution in FBNA use cases for AI, but I haven't found any that are incredibly sticky that are like
off the shelf. I can deploy with my team and have them use because of all the constraints that I talked about previously, but I think we're going to get there. So it's going to be a pretty big priority for me and my team in the next year to really think about how the FBNA job function is going to evolve in the next few years in the new age of AII know it's not a it's not a a clear answer, but I just know that it is. I just don't know how.
Yep. No, the one thing you mentioned about how AI is all, I mean, LLMS are all fundamental language models. I mean, it's in the name. And so yeah, if you ask them to do any arithmetic, they're absolute shifted. But but the one thing that we have found that we are leveraging in Babbage and quite a few others are also leveraging, is that what you can do is to get them to write SQL queries. So instead of you LLMS can do math. But what they are very good at is writing code that can do
math. And so based on that, for example, what what we are we are doing here at barrages in terms of like how do you use LLMS to sort of like proactively query data to find out insights and then proactively find out why those things are happening. So for example, in an FDA kind of a situation, you might have a situation where you have, let's say, I don't know, let's say you have a budget for a particular
hit. And then like you have the actual spend to to track the spend versus the budget, how it is going, where whether it's going as to plan in all parts of the business or if there is something going along somewhere. And if, if, if something's not working, why it is not working, it takes so and the entire stack in some sense can be built using just getting LMS to write word in some sense.
So, so from that perspective, I understand that like I assume that you guys have get to use any of the AI for analytics kind of tools of which the word 100 I would say so. So you just wanted to mention that? Yeah, absolutely. I completely agree with you. I think maybe maybe the content definition that I use for LLMS is definitely like broader, like coding is definitely a good
example. And and to be fair, like I encourage and almost like insist that my team start with AI for any use case that they have, whether they're writing a document about like root cause analysis or they're pulling like data together and they want to write supervise to me, I think there's a lot of progress in those step one or like horizon
one use cases for AI for sure. But I do think the, the next step is going to be an evolution more than a Step 2 in terms of how AI can truly change the DNA function. So that's what I, I, I, I spend a lot of my time thinking about. But there's definitely like several use cases for AI that my team leverages today that gets us like incremental efficiency, efficacy in how we do our work. OK. Ashley, thanks a lot for your insights today. Absolutely happy to enjoy our conversation today.
Thank you.