Everything about entrepreneurship, the good, bad and the ugly of it, fascinates me. I take a keen interest on startups and venture capital firms and have written extensively on fundraises, M&As and business strategies. I can safely say changing tracks from engineering to journalism has been one of my best decisions. When not working, I indulge in almost every Indian's poison, cricket, playing or watching. I am a foodie and video game buff.
Anand Rajaraman has been there, done that, first as an entrepreneur and later an investor in startups. He tasted the highly-coveted but mostly elusive entrepreneurial success, a profitable exit, pretty early in his career. Within two years of its launch, Junglee, a comparison and price discovery startup he co-founded with Venky Harinarayan, Ashish Gupta, Rakesh Mathur and Dallan Quass, was bought by Amazon for $200 million in 1998, a stellar exit in those days.
Ever since, a pursuit to make sense of tomes of data—Rajaraman also teaches machine learning at his alma mater, Stanford University—and Harinarayan have been constant fixtures in Rajaraman’s professional outings. The duo started Cambrian Ventures, an early-stage fund backed by Jeff Bezos, in 2000. In 2005, they launched Kosmix, an ecommerce-focussed social media startup, that was bought by Walmart for about $300 million. Together, they have invested in several startups, notably Facebook and Lyft in the US and Urban Ladder and Snapdeal in India.
Their latest venture, a $40 million early-stage fund Rocketship, has made about a quarter of its 32 bets in India, including Fynd, PaySense, Mad Street Den, Moglix and Locus. A similar proportion of its corpus has been deployed here. In the midst of raising a second fund for Rocketship (he declines to divulge details), Rajaraman, in an interview with Forbes India, talks about the fund’s data-driven approach to investment, the need for an investor to identify an addressable market and lessons learnt from his India investments. Excerpts:
Q. How has online retail played out in India vis-à-vis the US? How many such businesses, both horizontal and vertical, can India accommodate? The US and India are not different. In the US, there is one leader in ecommerce, Amazon. In India, there are two businesses competing, Amazon and Flipkart. India as a market can possibly support one or probably two big ecommerce companies. The market is just not big enough.
There was a bit of over-investment and the valuations were out of sync with reality. You can expect that anywhere in venture capital. In certain verticals, there is room to build. There was a belief that certain verticals will go online before India was actually ready for them, furniture for example. India still needs an online-offline model for some of these things and that obviously takes longer and more capital to build. From a broad market point of view, if anything can be done purely online, you can expect the market leaders like Amazon or Flipkart to be dominating those.
Q. Did we overestimate the market in India? As an investor, how do you differentiate between overall market and addressable market? I imagine so. We are an early-stage investor. At that stage, it is hard to know the exact size of the market opportunity. We see if the business has product market fit, whether the early cohorts look good and if there could potentially be a big market. When I invested in Facebook, I had no idea what the size was. I believe even Mark Zuckerberg or Accel (another early investor) didn’t know the size of the market. Sometimes there are positive surprises and sometimes negative ones. People can say they saw a great market and went in, but that’s all hindsight.
Our median valuation for Series A has been $20-25 million. At that valuation, we can take those risks. At the later stages, however, investors are evaluating market size and price of the round etc. People coming in later have to be more cautious on the pricing.
There is always a little bit of a herd mentality in all venture capital. It happened in the US as well. A lot of it is the fear of missing out. Suppose I don’t have a self-driving car startup and they become big, I will look foolish. All these are eventually baked into the economics of the model, at least for the good funds. The bad funds go out of business. When the boom starts, a lot of investors are drawn in from the fringes who don’t necessarily have the expertise and when the bust happens, a lot of these people leave.
Q. With Rocketship, was there any particular gap that you set out to address? Both Venky (Harinarayan) and I understand venture capital and we had our wins and losses. We knew the old model and didn’t really want to do that again with Rocketship. The old model is a network-driven model. It works well when there are geographic concentrations. Most venture capitalists spend their time cultivating their network.
But entrepreneurship and innovation have gone completely global. How do you invest in these hot companies across the globe? The answer that we hit upon was, the right way to do this is to use data and algorithms. If we can build a big data set of startups in the world, then we can run machine learning models on it and we might be able to identify these companies and instead of waiting for them to contact us, we can proactively reach out and invest. We license data from many data providers, like Tracxn in India. We also crawl the web, we look at social media and it’s all automated. At Rocketship, we probably have the most comprehensive data set of startup activity in the world—more than 10 million companies globally.
Q. Is dependence on machine learning a full proof method to identify potential winners? Are the algorithms evolved enough to identify, for instance, fudged data? The first set of (machine learning) models is a screen. The initial models, of the millions of companies in the data set, identify about 500 companies a year. We talk to them and get the second level of data from them and then we invest in 10 of those.
There is some element to this that is data-driven and some element that is our expertise. The algorithms will say whether the pattern looks interesting, but whether it is fudged or not is something that we will have to look and see. The important thing to realise is that it is not the algorithm or machine learning model making the decision. It’s a combination of those models and us as venture capitalists with 20 years of experience. We are combining these things. I don’t believe in that (completely automated decision-making) model. My strong belief is that the future is about humans and artificial intelligence (AI) working together to make better decisions and not in isolation. You train the AI on some data, but it is completely blind to what else is going on in the world. Humans are good at broad vision. It’s a combination.
Q. While everybody talks about data being the new oil, do all companies know how to do something meaningful with the data they are sitting on? This whole data thing started way back in the 1990s. There was no World Wide Web, but companies were using their own data to make better decisions. At the point, data was internal and private. The second phase happened when World Wide Web came around and there was so much more public data available. That unleashed a wave of innovation. Companies were built on publicly available data and the biggest success story is Google. The third era was when social media came out and then we had social data, which is kind of neither public nor private data.
The new set of data that companies are trying to get are through trained machine learning models.
There is an interesting dynamic called network effects, which create lock-ins that make it hard for competitors to enter. When you are an early entrant, you might put out a very ordinary product. But since you are the first product, people start using it and when they do, you gather usage data and you can use that to build better models and improve the product. This is a network effect and when this starts rolling, it becomes very hard for a new entrant to come into the market because they don’t have this usage data to make their products as good. That is what exactly happened in case of search. The market leaders who have this network effect going have an edge.
Q. You were late in entering India. Why, and what interesting opportunities did you spot here? India started happening for us around 2017-18. Jio had launched. We didn’t consciously say we wanted to be in India. As a fund, our algorithms pulled us here. They were showing a lot of companies in different segments with good adoptions. Companies in the short video space. A lot of growth in fintech and vernacular languages. Rural ecommerce is another interesting idea. With Jio, it looks like the market can be much bigger.
In terms of sector and spaces, serving the second and third tier cities, serving the non-English user base, both in terms of entertainment and communication, commerce, interest us. Fintech is going through some issues, but we are keeping a close watch. Health care as well. We are looking at B2B software as a service for Indian companies with a global market.
Q. Data may guide you to sectors where a lot of companies are coming up or money is going in? How do you make sure that you don’t enter a bubble? We are focussed on specific companies and their numbers and execution, as opposed to the overall ecosystem. For us, the themes come bottom up. We look at the companies, get excited by them and then we see if the company is part of a theme or whatever. Many investors are thesis-driven. We are the other way round. We fixate on specific companies as opposed to themes. We are not thesis-driven at all.
Q. What are the lessons learnt from investing in India? In India, things always take longer than you think. The regulations change, so companies have to adapt to those things, which doesn’t happen as often in the US. Sometimes, the regulatory changes create opportunities. That’s an interesting learning that whenever there is some kind of a change, that can be an opportunity for an interesting company.
Q. How comfortable are you with the valuations here? I don’t take private market valuations seriously because it is one investor’s belief of what the company is and not that of the overall market. It may or may not reflect any underlying strength in the business. Massively funded companies with big valuations do fail as well. In the first wave of ecommerce, many companies raised lots of capital and they didn’t end up going anywhere. Private market valuations should be taken with a pinch of salt. Only when a company goes public, you know if the valuation is real or not.
Q. Are investors more inclined to back business model innovations while others get sidelined as science projects? There are companies with good technology that have been built. It is right to say that technology alone is not enough. In the early stages, I won’t know the size of the market, which is okay. We want a technology to be innovative and the adoption has to solve a real customer pain point and cause disruption in the market. Clearly, there are going to be science projects and there are going to be some that will solve interesting pain points and those are the ones that we will back.