The 113th episode of Datacast is my conversation with Shruti Bhat, the CPO and SVP of Marketing at Rockset, a real-time analytics database for building data-intensive applications at scale.
Our wide-ranging conversation touches on her early engineering career; her shift to product management while getting an MBA at UCLA; her experience launching VMware’s vSAN; her time as the VP of Marketing at Ravello Systems driving them to an Oracle acquisition; her current journey with Rockset to make real-time data processing easy and affordable; her insights on data applications, the modern real-time data stack, and technical partnerships; her thoughts on hiring, go-to-market, and fundraising; the evolution of enterprise marketing; and much more.
Please enjoy my conversation with Shruti!
Key Takeaways
Here are the highlights from my conversation with Shruti:
On Her Upbringing
It is interesting how many people say they went into engineering because it was their passion and simply followed it. In my case, I did not discover my love for engineering until after I started studying it. I simply felt that I was good at and enjoyed tech, so I went into engineering to explore that.
Looking back, I realize that the learning I took away from that experience was more valuable than simply waiting for my passion in life to reveal itself. I went exploring and found my passion. Overall, my university experience was amazing.
At the time, computer science was a very popular choice in India, attracting many smart people. It was a competitive yet fun environment, especially for those who enjoyed engineering and computer science. I certainly did.
On Her Early Software Engineering Career
Choosing to work for a product development company instead of a services or consulting company was an interesting choice for me. At the time, most of the recruiting on campus was focused on those types of companies. However, my passion for product development led me to HP, which was an excellent choice for me personally. I learned a lot from my early years there, including the ability to make change happen early in my career.
One example was when we relied on a remote team at HP, and all of the hardware equipment was across the ocean. Whenever there was a failure, we had to wait for someone on the other side of the world to reboot the hardware. This was a terrible experience, so I managed to get funding to order the hardware equipment for us. The only way to make it happen was to roll up my sleeves and say that I would set it up myself, even though I knew nothing about being a storage admin. I promised that if they gave us the equipment, I would learn how to be a storage admin and set it up myself. This approach worked, and it taught me that if you take a step back and systematically remove roadblocks, it is possible to achieve big things.
Another lesson I learned was not to be afraid to experiment or automate things, even if it means automating your own job. For example, early in my career at HP, I did a lot of quality assurance and testing, which involved maintaining different operating systems on many different servers. To make this process more efficient, I decided to virtualize everything and automate how the operating system was maintained on these servers. This automation allowed me to complete the task that had previously taken months in just a week. Instead of losing my job, I was promoted because I had made the process more efficient. This kind of experimentation and constant effort to automate my own work eventually led me to VMware later in my career, where I became one of its early users.
On The Early Days of Cloud Computing
Back then, cloud technology was not a thing yet. It was all about data center technologies. At HP, I mostly worked on storage hardware — specifically, a storage array. It was called a Virtual Storage Array.
The state-of-the-art technology at the time was HP’s Enterprise Virtual Array, which aimed to virtualize storage. This was the first step towards virtualizing servers, eventually leading to the cloud. However, those were the early days when you had to sit in the data center, freezing while trying to reboot things. The hardware constantly failed, and you had to go in and pull out discs. I remember staring into these orange cables to see if they still worked. There was a lot of hardware back then.
Today, I appreciate how the cloud has changed everyone’s lives. I work with people on my team who have never seen or touched a server. It is amazing that we can build software without ever having to touch a server or storage array.
On Pursuing An MBA at UCLA
One of my most transformative experiences was when I learned that if you hustle, you can learn so much more during school.
For example, I landed internships in product marketing at Cisco and product management at Comcast. While studying for my demanding full-time MBA, I did an internship that everyone else did, but I also found a second one on my own. I did this purely for the experience.
I walked into the company and told them I would do product management for free just to learn. And that is how it started. Within two weeks, they offered to pay me and support me as a student.
So not only did I learn, but they also helped pay off some of my student debt. This experience taught me how to work hard, hustle, and work my way through school, which was a completely different experience for me. Of course, as part of the business school education, I also learned many business fundamentals.
I learned that while I enjoy technology, my passion is really the business of technology and understanding how to build technology that solves problems, monetize that technology, and take it to market. So my engineering background pivoted into more of a business side.
The schedule was pretty demanding since Anderson at UCLA has the quarter system, and the quarters are very short. As I went through school, I took a lot of extra classes because there was no limit on the number of classes I could take. Apart from my coursework, I sat in on many amazing courses, served as a teaching assistant, worked a job on the side, and did full-time on-campus/off-campus recruiting since I was graduating in a recession in 2010.
On Shifting To Product Management
I realized that what shifted for me was my approach to problem-solving. While many engineers jump right into solving a problem, I always started by asking why. Before delving into how to solve the problem, I would step back and consider why we were solving it, why it mattered, and what it meant for the customer. I would pull every thread to figure out why the problem was important. Once I had answered the why, the how became obvious. This big-picture thinking, always stepping back and asking why, made me the engineer that every salesperson wanted to bring in front of customers. I enjoyed talking to and understanding customers' problems.
As I learned more about product management, I realized that is what I wanted to do full-time. While at Anderson, I took entrepreneurship, technology and product management, and finance courses, which proved helpful. People often underestimate the importance of finance courses, but understanding the business and how it is run is crucial in product management.
On Launching Products at VMware
My proudest product launch at VMware was a product called Virtual SAN, or VSAN, as it is known today. It turned out to be one of the fastest-growing products in VMware's history.
I was so proud of launching VSAN because, at the time, EMC owned it, and we were a virtualization company going into the storage market, directly competing with EMC, the mothership. Of course, the technology was exciting. It was the first time that there was virtual storage. It was not hardware anymore; it was completely software-defined.
Software-defined means there is no physical storage array; it is all software. It takes a bunch of servers and turns them into storage. I thought it was fascinating; it was the first of its kind and category-defining. After we did it, many other such software products were released.
I am proud that there were many pricing challenges and political waters to navigate, given that EMC had concerns about VMware competing with them. We had to find our positioning, find a way not to compete, and still go to market in the most profitable way for VMware. Pricing and positioning were the two biggest challenges apart from the complexity of the technology itself.
But how do you take this really complex technology, price it so that you monetize it, and then position it in the right way so that it becomes the product that everybody wants to adopt? That's exactly when cloud computing was taking off. This was a time for us to say, "Hey, instead of just selling hardware, how do we make everything more software-defined?"
The big challenge was mainly around the go-to-market strategy because taking a category-defining product to market is very challenging. Creating a category is not fun, to be honest. People think you want to create a category, but it is very hard. So that was the biggest learning: avoid creating categories unless you absolutely have to. And if you do have to create a category, how do you go about it? That is where all the nuance was, and that is what I really learned: how to create a category.
On Serving as the VP of Marketing at Ravello Systems
Startup itch is something that's hard to ignore when you are in Silicon Valley. I happened to know one of the investors who had invested in Ravello Systems, which caught my interest. This is one of the perks of being in this area - I got to meet them through my personal network.
After talking to them more, I realized this was an opportunity I could not refuse. It had nothing to do with wanting to leave VMware. I was happy at VMware, but this was both interesting and exciting.
At Ravello, I saw a change in how customers bought products. Instead of being sold to, they wanted to discover products on their own. They were more informed and had direct access to their peers. This was the start of what we now know as product-led growth.
To succeed in inbound marketing, it is important to have a self-service product that can support the inbound demand. The technical audience is very savvy and does not respond well to marketing gimmicks. The key is to understand how they like to buy and what they like to learn and provide them with transparent and honest content.
At Ravello, we built an amazing product and trustworthy brand. This, combined with word of mouth, helped us spread our message. It was an exciting time, as people were starting to have direct access to their peers and could learn about products through reviews and online forums.
On Managing Oracle’s Cloud Portfolio
Oracle was an interesting experience. Going from a small startup to such a big company was quite a shock to the system.
What I found interesting was spending the first few months integrating the company and getting the Ravello acquisition off the ground. But then I got the opportunity to do some really interesting strategy work. I was intrigued by how technologies like blockchain, AI, and IoT could touch many different products in a company like Oracle.
This was my first experience with something called portfolio product management. Instead of just thinking about one product and its roadmap, I was looking at a portfolio of products and figuring out how to bring Oracle's AI vision into these different products. It was challenging because every product already had its own roadmap and a lot of different challenges it was working on.
To apply AI to an existing product, I had to go deep into customer problems and understand how the new technology could improve the product. I did not want to go top-down and force AI into every product. Instead, I wanted to influence product managers by putting their interests first and showing them how AI could help them.
Navigating politics is not about getting what you want. It is about figuring out what others want and helping them achieve it. By putting other people's interests first, I was able to build trust and influence. And ultimately, it is all about people.
On The Founding Story of Rockset
While at Oracle, I attended many meetings with the CDOs and CIOs. They had all invested in streaming technologies and collected a lot of data because they knew data was valuable. However, their biggest problem was going from collecting data to building an application. Many were collecting data in real-time and storing it in S3 or some data lake. But data sitting there is useless, and S3 is where data dies.
Almost every customer kept asking the same question: How do we use this data to build an application? I could not give them a great answer, so I started talking to startups to see how they were solving the problem. This was when I met Rockset's other co-founders, Venkat and Dhruba, who were coming out of Facebook and had solved this problem.
Facebook's newsfeed is a data application that uses real-time data to make personalized recommendations. Venkat and Dhruba had experience building the system and wanted to bring this technology to the masses. Meanwhile, I was hearing firsthand the problems faced by CDOs and CIOs and the applications they wanted to build.
We put our heads together and built Rockset. From the early days, we knew we wanted to solve the problem of real-time analytics. The technology evolved, and we developed converged indexing and our own query engine, all driven by the initial pain we saw in the industry. We wanted to make it easy to go from data to an application.
On Real-Time Analytics and Data Applications
Rockset is a cloud-based platform for real-time analytics. Real-time analytics means analyzing data as it comes in, regardless of the shape or source of the data. It could be JSON, Parquet, Avro, or any other format, and it could be coming from various sources, such as Kafka or Kinesis streams, or even a CDC stream from a database like Oracle, Mongo, or Dynamo.
Real-time analytics differs from batch analytics, where data is transformed and loaded into a warehouse. With real-time analytics, queries can be written to show data that happened even a few seconds ago. Rockset is built specifically for the cloud, allowing for fluid and variable data streams that can be scaled quickly and cost-efficiently. It is highly parallelized, which means it can scale to hundreds of CPUs very quickly and bring it down quickly as well.
Data applications require fast data retrieval, which is not optimized for warehouses. Warehouses are built for data storage, and using them for data applications can result in slow queries and high compute bills. Rockset is built to optimize for data retrieval, making it a more efficient and cost-effective choice for data applications.
On Rockset Architecture
We already discussed data applications and the difference between developers and analysts building them. Developers focus on experimentation and iteration, which means they always try new data sets and queries. This presents a challenge for the data engineering team to keep up with these new requirements.
For example, the growth engineering team at Facebook conducted thousands of experiments, including the creation of the "like" button. They could do this because of their real-time analytics system, allowing them to run different queries and see what worked best. That is why we created Schemaless Ingestion with Rockset.
With Rockset, you can bring in new data sets in minutes because it is schemaless ingested. You just point the system at your data and start writing SQL queries on it. We remove the burden of defining the schema and doing transformations, making it easier for developers to experiment with different data sets and queries.
Furthermore, we give developers SQL access, which is a language they all know. We support JSON nested, JSON, Avro, and Parquet, and within seconds, you can start writing SQL queries as if the data was already structured for you. The system all automatically does this as part of the indexing process.
On Converged Indexing
When you search on Google, the results return in milliseconds thanks to the indexing process. Similarly, we have developed a converged index that combines search indexing with columnar stores to create a new index type that optimizes data retrieval rather than storage.
While a data warehouse optimizes for storage by compressing data, we index all the data and fields to make queries quick and easy. This allows for fast and compute-efficient queries, leading to significant cost savings for our customers. In fact, many customers have found that they have cut their compute bills in half simply by making the switch to our converged indexing.
So if you are looking for real-time data retrieval that's both fast and cost-effective, converged indexing might be just the solution you need.
On “The Modern Real-Time Data Stack”
The modern real-time data stack has a few key requirements:
It must be cloud-based.
It must use SQL for interoperability with other systems.
While these two requirements are fundamental to any data stack, what sets the real-time data stack apart is its focus on using the right tool for real-time data processing. This means avoiding batch ETL and instead using tools like Confluent, Kafka, and Kinesis for streaming data. Tools like Debezium and Striim can be used for change data capture (CDC) from databases.
Once data is ingested, it must be transformed, stored, and analyzed. Real-time SQL transformations can be done at ingestion time using tools like dbt. For storage and analysis, Rockset is an excellent option for real-time data. SQL visualization tools like Grafana, Tableau, and Preset can be used to build dashboards and visualizations on top of the data.
The most exciting part of the real-time data stack is the ability to build data apps using SDKs and machine learning. Examples of such apps include fraud detection and anomaly detection, which can trigger alerts and other automations. In contrast to dashboards, which can be tedious to monitor, data apps allow for automation and real-time monitoring.
On Technical Partnerships
Product strategy all comes down to what the customer wants and needs. The problem that most people face today is how to build a data stack that works together seamlessly. From the customer's standpoint, they need the entire stack to work cohesively. It is vital for a data company to consider the whole stack and work well with other pieces of the stack to make the customer's job easier. This is why it is important from a product strategy standpoint to make it easy for the customer to have end-to-end visibility into their stack and optimize it for themselves.
From a go-to-market strategy standpoint, it makes sense for companies to join forces and accelerate the transformation in the industry from batch to real-time. This is all driven by understanding the customer's jobs to be done and what pieces of the puzzle they need to build their applications. Ingesting data, transforming data, storing data, and analyzing data are all important pieces of the puzzle, and understanding the lifecycle end-to-end is crucial in identifying the right products to bring together for the customer.
It is essential to listen to the customers and follow where the market is taking us rather than putting darts on the board. By doing so, we can identify the best vendors to work with based on customer demand and ensure we work well with that ecosystem.
On Product Vision
I believe we will accomplish two things. First, we will speed up the transition from batch to real-time data processing. Second, we will make it more accessible and affordable to analyze data in real time.
To illustrate this point, let me give you an example. Think about how many items you used to receive within two days ten years ago. Did you always opt for two-day shipping? Probably not, because it was expensive and the technology was not there. Would you pay $60 for two-day shipping on an inexpensive item? Most likely not. People settled for slower delivery options because faster options were too costly or complex.
The same is true for batch data processing. People settle for slower processing because faster options seem too expensive, difficult, or complex. However, we are finding that once people realize how easy and affordable real-time processing can be with Rockset, they want to use it for more and more use cases. Our goal is to make real-time data processing so easy and affordable that anyone can do it, and as a result, the entire industry will transition from batch to real-time processing much faster.
On Customer Use Cases
There are many fun use cases for real-time data. Today's generation, including employees and customers, demand real-time information, and it is easy to see why.
For example, a company in Alabama (called Command Alkon) builds software for tracking cement mixers used in heavy construction. Cement mixers need to keep spinning and be tracked in real time so they arrive at the construction site at the right time, even if the crew is running late or the weather is bad. The company uses Rockset to provide this real-time tracking, which their customers demanded, and their employees also quickly adopted it.
Other already digitized industries, such as e-commerce and gaming, also benefit from real-time analytics for personalization, recommendations, and leaderboards. Fintech is another big use case, particularly for fraud and anomaly detection.
Regardless of the industry, people are becoming accustomed to instant experiences and expect real-time information as a result. Providing real-time data can reduce risk, increase customer satisfaction, and improve performance in many areas.
On Hiring
At the end of the day, a startup is all about the people. While technology and product are important, it is the people that really make a difference.
The first few hires are crucial because they attract the next set of hires. So, we always keep our bar high on integrity and trustworthiness. It is important to us that the people we hire have a good reputation and can be trusted to bring in the next group of hires. This is how we maintain a strong culture.
We also place a strong emphasis on diversity. Diversity does not just mean gender or race but also diversity of thought. We look for people who can bring new perspectives to the team. Someone who thinks differently or has different experiences or backgrounds can bring a fresh approach to problem-solving. This is how we build a great company.
We use metrics to evaluate our hiring process to ensure that we have a diverse pipeline of candidates. If we find that we are not getting the mix of candidates we are looking for, we will go back to our pipeline and adjust our recruiting strategy. This may mean expanding our recruiting efforts to different campuses or working with our recruiting team to ensure that our screening process is fair and inclusive. As long as we are measuring the right things, we can continue to improve our diversity metrics and build a stronger team.
On Interviewing
I have found that having an open mind is crucial not just during the interview but even during the resume screening process. It is all too easy to glance at a resume and think, "This person does not seem like us." However, doing so would be a grave mistake.
The first step is to approach every candidate with an open mind. I try to identify what makes each candidate unique during the interview process. I look for their superpower – that one thing that sets them apart from the rest of the team.
When I interview someone, I am really thinking about how they will fit into my team. What superpower do they bring that no one else on the team has? By building a T-shaped team – where everyone has different strengths and weaknesses – we create a diversity of thought that enhances our work.
It is easy to focus on weaknesses, but we need to focus on strengths. Where does their superpower lie? If they do not have one, that is a problem.
It is not enough to recruit diverse talent; you have to retain it. That means creating an environment where people can do their best work and shine. I believe it is crucial to create a culture that fosters success for everyone on the team.
On Finding Design Partners
From the very beginning, we focused on bringing a real-time data stack to our customers. This intense focus on solving a narrow problem helped us find early adopters. We looked for customers where the pain was urgent and where our solution was the right fit. This approach helped us move toward product-market fit.
We did not have an amazing product that solved all the problems right off the bat. Instead, we launched early with an MVP and worked with early adopters who gave us valuable feedback and input. We treated them as design partners to help us shape the product.
To find those early adopters, we maintained our focus on the narrow problem we were solving. We said no to other customers or problems that came along the way and did not fit our focus. We looked for big burning problems that we could uniquely solve in a creative way.
We preferred working with small teams in the early days because they moved faster and gave us more feedback. We focused on use cases where real-time data was important, such as building low-latency applications. Coming from Facebook, we understood these kinds of applications really well. We identified which companies were building these applications, whether small or large, and spoke to them to understand their problems. From there, we worked with them to address their needs by making our product schemaless and SQL-based.
As we learned more and more about their problems, we continued to work with them to improve our product.
On Growth Strategy
What we discussed earlier is very similar to what people nowadays refer to as PLG or Product Led Growth. Essentially, it involves having a narrow focus on a small set of problems that you can uniquely solve and creating inbound demand to ensure that those who need your solution can discover it.
For instance, we provide real-time analytics, so we understand the needs of those who build data applications. We start with the end in mind and figure out what our customers are searching for on a day-to-day basis. They might not know about Rockset, but we ensure we understand what they're Googling for and work backward from there to ensure they discover our solution at the right point in their journey. This is the inbound game.
We also complement inbound with a self-service product. As a cloud-native platform, we focus on "surprising simplicity" and strive to make it so simple that anyone can use it. We have an internal metric that the time to first query should be five minutes. This means that if someone starts a trial of Rockset, they should be able to bring their data in, look at the schema, and write their first query within five minutes.
Our growth marketing team is structured around inbound versus outbound. We generate inbound demand and support the outbound motion as the sales team grows. Within inbound, we focus on different personas and how to position and message our solution to appeal to them based on who they are. Data architects might look for different things than data engineers, and we ensure they find the right content that appeals to them at that particular time.
We work with partners to ensure that we reach our target audience effectively. Our growth marketing team experiments, tests, and is super data-driven. We are very focused on a few different things that we have discovered along the way, and we strive to be good at many different things to ensure our success.
On Fundraising
I believe that success ultimately comes down to the people involved. While having a recognizable brand name like Sequoia or Greylock can provide instant credibility, what really matters is the people on our board and the investors who support us.
In our case, our board members Jerry Chen and Mike Vernal are trusted partners who we've worked with in the past. We have great alignment with them, and they share our vision for the company. They've been extremely supportive of our approach to building the company.
Jerry and I worked together at VMware, while Venkat and Dhruba worked with Mike back at Facebook. These relationships have been very helpful to us.
My advice to other founders is to focus on finding the right investors who believe in your long-term vision and will support you throughout the journey. While having a recognizable brand name can help, it's the people behind it who really matter.
On The Evolution of Enterprise GTM
The IT industry has undergone a complete transformation from the days I worked at VMware. Today, customers' buying behavior has changed significantly. In the past, centralized IT teams made purchasing decisions after extensive POCs. Now, it is much more decentralized.
Developers are the ones with the power. They try out free trials and make their own decisions about what they like and why. This decentralization is good because it eliminates the need for a massive sales team that solely caters to the buyers' whims. Today, the sales team can directly approach the person consuming the technology, which is a huge shift.
Our sales team works closely with the technical personas, such as users, developers, data engineers, and data architects, which is crucial. The buying process has shifted from centralized IT teams to smaller teams and individual developers who choose what is best for their specific use case.
The second significant change is in data. Marketers now have access to real-time data, which is all digital. At Rockset, our marketing team is entirely data-driven, and we use real-time data alerts to monitor the buyer's journey. We can track intent data, research behavior, problems, and purchasing decisions. Savvy marketers need to use this data to show up in the right place at the right time.
Show Notes
(01:49) Shruti shared her upbringing in India - where she studied Engineering and Computer Science in the early 2000s.
(03:11) Shruti reflected on her early career as a software engineer at Hewlett-Packard and IBM.
(07:29) Shruti recalled the early days of cloud computing.
(09:01) Shruti reflected on her time pursuing an MBA at UCLA Anderson School of Management.
(11:55) Shruti explained her shift from software engineering to product management.
(14:19) Shruti revisited her years at VMware as a product line manager for cloud infrastructure - owning all aspects of go-to-market strategy and execution for VMware's entire software-defined storage portfolio.
(18:30) Shruti talked about her time as the VP of Marketing at Ravello Systems - growing the business from zero customers to a successful multi-million dollar acquisition by Oracle.
(23:07) Shruti went over her time as a senior director of product management for Oracle's cloud portfolio.
(27:20) Shruti recalled the founding story of Rockset - where she is a co-founder and Chief Product Officer.
(30:40) Shruti explained the concepts of real-time analytics and data applications for the uninitiated.
(37:31) Shruti unpacked the high-level design of Rockset architecture - which brings together cloud-native architecture, schemaless ingestion, converged indexing, and full-featured SQL.
(40:23) Shruti elaborated on the concept of converged indexing.
(42:43) Shruti dissected the technology requirements and the key layers of "the modern real-time data stack."
(46:17) Shruti talked about the role of partnerships in Rockset's product strategy.
(51:29) Shruti highlighted some of Rockset's customer use cases.
(56:06) Shruti shared valuable hiring lessons to attract high-integrity and diverse people for Rockset.
(58:51) Shruti shared her take on interviewing on strengths over weaknesses.
(01:01:32) Shruti shared the strategy Rockset used to find design partners in the early days.
(01:05:12) Shruti shared the tactics to combine the power of product-led adoption with sales-driven growth for rapidly scaling Rockset's business.
(01:08:45) Shruti shared fundraising advice to founders who are seeking the right investors for their startups.
(01:10:27) Shruti described the evolution of enterprise marketing and GTM strategy in the past decade.
(01:12:59) Closing segment.
Shruti's Contact Info
Rockset's Resources
Mentioned Content
Articles
"Building Data Applications Powered by Real-Time Analytics" (May 2021)
"How startups can create a culture where women can win" (May 2021)
"Streaming Data and the Modern Real-Time Data Stack" (Nov 2021)
People
Barr Moses (Monte Carlo Data)
Jay Kreps (Confluent)
Book
Competing Against Luck (by Clayton Christensen)
Notes
My conversation with Shruti was recorded back in June 2022. Since then, a lot has happened. I recommend looking at the resources below:
The launch of compute-compute separation for real-time analytics (March 2023)
This benchmark on top real-time analytics databases in 2023 (Feb 2023)
This talk on emerging architectures for real-time CDC (Dec 2022)
About the show
Datacast features long-form, in-depth conversations with practitioners and researchers in the data community to walk through their professional journeys and unpack the lessons learned along the way. I invite guests coming from a wide range of career paths — from scientists and analysts to founders and investors — to analyze the case for using data in the real world and extract their mental models (“the WHY and the HOW”) behind their pursuits. Hopefully, these conversations can serve as valuable tools for early-stage data professionals as they navigate their own careers in the exciting data universe.
Datacast is produced and edited by James Le. For inquiries about sponsoring the podcast, email khanhle.1013@gmail.com.
Subscribe by searching for Datacast wherever you get podcasts, or click one of the links below:
If you’re new, see the podcast homepage for the most recent episodes to listen to, or browse the full guest list.
