Meet the startup that's helping Uber and Spotify see into the future
Look at Elon Musk, he says: Planning Mars expeditions? Building electric cars?
"Those are all sci-fi book things," Doshi says.
Today, Mixpanel is unveiling its own sci-fi inspired product: A feature that will let its customers - including companies like Uber and Spotify - see into the future.
Mixpanel, a data analytics company founded in 2009 and funded by the likes of Y Combinator, Andreessen Horowitz, and PayPal co-founder Max Levchin, was already most of the way there.
"The best prediction of the future is the past," Doshi says.
Predicting the future
The idea behind Mixpanel is to help web site and app developers by telling them exactly how customers are using their apps. Where competitors measure which pages you look or how many times your app has been downloaded, Doshi says that Mixpanel goes a step beyond.
"We measure engagement instead of pageviews," Doshi says.
Mixpanel can track metrics like how many times a video is getting shared from a specific screen or how many people are quitting the app before finishing the sign-up process. Armed with that knowledge, a developer can tweak their app to boost shares, signups, sales, or whatever.
With the new Mixpanel Predict feature, Doshi says that it's using some (very trendy) machine learning algorithms to use the data it's already collecting for developers in order to predict what customers are going to do. Mixpanel Predict can start telling the future 30 minutes after a customer turns it on, promises Doshi.
So in a real-world example, Doshi says that Mixpanel is working with a popular recipe website. That website wanted its customers to save recipes to their account. Mixpanel Predict helped that site identify which customers were most likely to hit that "Save" button, so they could accelerate the process by sending the a nudging email or mobile notification.
Other real uses include Spotify guessing which song a user is going to listen to next, so it can send them a notification to hop to it, or an enterprise software company predicting when a customer is going to start paying.
It's limited, because Mixpanel Predict can only tell you what's going to happen, not how to adjust for a more desirable outcome - if you're going to fall 1,000 paid users shy of your weekly goal, Predict can't really help you. But more insight into the factors that go into a prediction is coming down the line, Doshi says.
'Raising a Series A round every year'
Mixpanel's engagement-based approach that's resonated with customers.
Rather than offer an "amorphous" analytics offering, like the kind sold by other startups, Mixpanel has a solid product with a "tangible benefit," Doshi says.
In fact, he calls "predictive analytics," the technical term for the kind of math that Mixpanel Predict does under the hood, "the epitome of a hyped-up industry no one understands."
But by offering customers something with a demonstrable value, rather than a set of vague, abstract technologies, Doshi says, Mixpanel has stood out in a crowded analytics space.
Now, Mixpanel has 3,500 paying customers, including "bleeding-edge startups" and large enterprises like American Express and NBC Universal. Plus, it offers a free version to "tens of thousands" of users, Doshi says.
Early last year, Mixpanel had 27 employees. Now it has 230. In 2013, Mixpanel was dealing with 13 billion "actions," like playing a song or saving a recipe, every month. Now that's up to 50 billion.
Mixpanel isn't profitable, Suhail says, because it's reinvesting in building out its salesforce. But when it went out to raise its Series B round last year, the company was making $1 million a month, Doshi says.
"We used to call it 'raising a Series A round every year," he says.
By the time that round closed in October 2014, Mixpanel had raised $65 million at an $865 million valuation.
"We're trying not to grow too fast," Doshi says.