From Developer to Successful Machine Learning Entrepreneur: Aparna Dhinakaran, Co-Founder, CPO of Arize (Part 4)

Sramana Mitra: How did you generate inbound interest?

Aparna Dhinakaran: Cold emails. It works. I just told them that I’m building a company focused on ML observability. We spent months doing this. Every call we were on, we showed them our mocks. We asked them for feedback. At the end of these calls, people would say, “When is this going to be released?” We started to build up a list of alpha users.

Sramana Mitra: How many of these did you have?

Aparna Dhinakaran: A lot more than the ones who actually tried it out. Some people drop off. Some people switch companies. Maybe somewhere in the 50’s. These are people who just wanted to try it.

Sramana Mitra: How long did it take you from that point to actually releasing an alpha?

Aparna Dhinakaran: Six months.

Sramana Mitra: What does that bring us up to? What year?

Aparna Dhinakaran: 2020.

Sramana Mitra: Middle of 2020?

Aparna Dhinakaran: Summer. YC was 2019. Jason and I joined in the beginning of 2020. There were around 10 users that I was meeting regularly to create that feedback loop.

Sramana Mitra: You found that set?

Aparna Dhinakaran: Yes.

Sramana Mitra: What companies were these from?

Aparna Dhinakaran: Some of them were FinTech startups. Some of them were an AI team within a Fortune 500 company. A lot of this is iterative. You don’t land on a go-to-market strategy and say this is it. While we were doing the mocks and pitches, that was our way of learning who responded to LinkedIn messages. What kind of people didn’t have the budget or couldn’t bring in a new tool into the organization?

Sramana Mitra: In your case, there is a difference between the economic buyer and the technical decision maker.

Aparna Dhinakaran: Sometimes, they’re the same. An engineer could say, “My team has this budget. I can make this decision.” If it’s a central ML team, you’re talking to the central team. Go-to-market strategies are iterative. You’re learning new things as you penetrate the market. Who you think will be your competition when you start is also changing.

When we started, I thought that the platforms that did serving would be a competitor. There are platforms out there that are helping you deploy models into production. They’re not the observability solution. All of this is shifting. The fastest way to learn is by pitching and teasing out the target set that’s going to be able to make a decision and is worth your time to invest in at this stage.

Sramana Mitra: You mentioned something about financial services. Is that your primary target?
Aparna Dhinakaran: FinTech is probably one of the best spaces. The reason for it is they have models in production. Fraud models is one great example. Every time the models go down or aren’t doing what they’re supposed to be doing, there’s a financial impact to that.

Sramana Mitra: It’s mission-critical.

Aparna Dhinakaran: Yes. It’s something that people already have live in production. The ROI of what we’re building is very clear. Figuring out what is that set that you want to get in and start owning is critical.

Sramana Mitra: Finding a use case that is high ROI is really key. It sounds like you’ve found yours. From the middle of 2020 to the beginning of 2022, what has happened? Were you able to penetrate the FinTech fraud market?

Aparna Dhinakaran: Absolutely. Once you get into a vertical and you have examples of customers, it does two things. From a product perspective, you can go even deeper. You’ve taken a lot of product takeaways too. How can I make the onboarding better? How can I make the value-add better? The next one you onboard, those insights should make the second customer even better to onboard.

The second thing that it does is how did you land that company. Who was the decision-maker? Maybe in FinTech companies, if it’s small enough, just that ML team can make a decision. If it’s big enough, do they still have to contact other people?

You’re learning about the go-to-market. Every customer you onboard, there are go-to-market things you’re learning and there are also product things that you’re learning.

Sramana Mitra: What is the split – FinTech startups versus the larger ones?

Aparna Dhinakaran: A product like ours is horizontal. It doesn’t matter to us what vertical you’re in. If you’re in the fintech space, there are fraud models. If you’re in the AdTech space, you have click-through rate models. If you’re in retail, there are churn models.

Sramana Mitra: But those are different use cases. Which ones are you going after? Are you going horizontal or vertical?

Aparna Dhinakaran: The product is designed for horizontal.

Sramana Mitra: But the market is what matters.

Aparna Dhinakaran: Exactly. Things like how do you get your data in, how do you turn on monitoring, how do you decide when something’s not doing well – all of that is applicable. There are base things that are applicable across use cases. There are some things around onboarding, documentation, use case and specific tutorials that become more vertical-focused.

The interesting thing is that it’s not just FinTech that has fraud models. You need to figure out what are those right use cases for your product and rally around those use cases with the go-to market.

This segment is part 4 in the series : From Developer to Successful Machine Learning Entrepreneur: Aparna Dhinakaran, Co-Founder, CPO of Arize
1 2 3 4 5 6 7