How Uber Depends on Data Analytics to Deliver Extreme Customer Service – Face To Face With Uber’s Chief Data Architect

From a simple limo hailing app for friends to the world’s go-to taxi app. Uber’s growth in the approximately 7 years of existence can be described by one word, “Phenomenal”.

But there’s another way to define Uber, one that not many have given thought to. Uber is a Big Data company, on the likes of Google and Amazon. It not only uses existing data in its banks effectively for its business operations, but the process of gathering data – data from drivers, data about drivers, data of passengers, data about passengers, data of traffic systems around the world, transactional data – and analyzing all of it in real time, continues.

BigInsights Principal Raj Dalal

caught up with

Uber’s Chief Data Architect M C Srivas

on his recent trip to San Francisco. In the course of the hour-long conversation, among many things, Srivas spoke of what data analytics means for Uber, and how innovation in data is being used to further what is now popularly known around the world as “the Uber model.”

This inaugural post in a three-part series examines the role data analytics has played in Uber’s success story.

Raj:

I have been tracking for a while now how data can be used to drive “extreme customer service”. Uber has done some exciting stuff, matching supply and demand and estimating pricing. Where do you see Uber going in the next 3 years where data and the enablers are concerned?

Srivas:

Well, it’s already the main ingredient. Let me give you the high level picture here. The way Uber makes money is when its drivers are happy and also the passengers are happy. That’s what makes Uber happy. The driver is happy when his taxi is always hooked. Essentially, he wants a passenger in his backseat all the time in order to keep making money.

For the passenger, he is excited when the price is the lowest; the least amount of money he spends to get from point A to point B. That, too, in the shortest amount of time. That’s what makes him happy.

That’s the real part of data analytics, it’s amazing.

“If you look at the last two years, its impact on the bottom line, you will realize that the average trip ride price paid by the passenger has actually has gone down by 50%. The average amount of money a driver make has actually gone up 30%.”

Yet, in spite of the fare going down, the drivers are actually making more money. That’s the amazing thing about matching these two factors – both sides should be happy. And this is where Uber has an edge over others; that’s our secret. This is where all the other guys who tried this got beaten back and lost. There are many companies which think of it as just a car hailing app where one needs to make these two ‘people’ get on the way and make some money. No, you don't make money, you got to make the other people money first. And doing that efficiently is very difficult.

“So analytics is like the core of the business. And there’s a barrier to entry for anybody else who wants to get into this business. It's very difficult.”

Raj:

Uber has got a few years’ lead. It’s capturing lots of data with every ride of its cabs. But why do you think others have not been able to catch up?

Srivas:

It's two things. The first is data and how well we do this. How well does Uber understand traffic pattern across every city in the world – in fact in 70 countries? More importantly, our ground operations are phenomenal, the quality of the cars, the quality of the drivers is ‘Uber’, that's why the service is called Uber. It's very high quality. So doing this low cost-high quality thing is the ground up mission of our every city unit ground operation.

Factors that are considered are - how do you inspect that cost, how do you go and get 500 drivers before even operations start….you need 2000 or so passengers before the launch. The drivers will not operate unless there are passengers. The passengers won't come unless there are drivers. Then, there are city-specific regulatory hurdles that have to be overcome, too. So, to get that momentum going for every city, it requires an amount of ground effort first. And for that you need extremely strong teams. And the inclination to take on local regulations, head on.

Raj:

Obviously, Uber's got incredibly deep pockets. So how did you manage to convince the investors community that it was worth going head on and the returns will come?

Srivas: Uber is really urban transportation done right by private citizens.

In the last hundred years we've seen so many attempts at urban transport. Governments tried to solve UT problems; it was a disaster. There's probably only two cities in the world that urban transport actually worked, that, too, in a limited fashion, and that is London and Tokyo probably.

Uber is really urban transport done correctly, and doing it using citizen themselves who want to augment their incomes. Because it harnesses the urban instincts of everybody. I mean this is really the next way of how transport is done - Self-service Transport. Therefore, the deep pockets happen automatically. The investors understand that this is really the next way of how transport is done. They know either somebody else will do it or we can go and do it. The rewards are very high, the stakes are very high for whoever starts dominating it.

Raj:

Back to your role... you've obviously involved since the beginning of Big Data analytics at Google, and then as one of the founders at MapR...as chief data architect at Uber what are some breakthroughs that will help Uber to go on to the next phase?

Srivas:

Good question. See, unlike Google or Facebook or Twitter, every bit of Uber's data is a monitoring transaction. There’s livelihood linked to that; getting that exactly right every time. Nobody cares if on Facebook if a ‘Like’ was wrong, or if you miss someone's picture. Big deal. Or in Google search if you miss a website, no problems. The next time you'll get it probably. So, it's okay to lose data in those situations. With Uber, every piece of data is directly impacting somebody's pocket. So the need for quality is much higher than that of any of the standard websites. We're not here scrapping the garbage and trying to figure out how to make money from them, or doing ads…every one of them is a marketing transaction.

It's more like the stock exchange. There are regulations, things are auditable. Eg: How many taxis ran today, and so on. So the standard Big Data rules, like approximation, machine learning, while those do apply, you cannot lose any of it. and you cannot lose any of it at this big a scale.

And that poses a whole different set of challenges. It’s not like the standard analytics that most companies have - either weekly report or daily reports or something like that. At Uber, in fact, pricing is instantaneously, and that pricing directly affects the bottom line, how people are charged. So that's very different, in fact, it’s first of its kind.

It’s like the stock exchange - based on demand, the pricing changes, and the volume, maybe not as the stock exchange high-end volume, but certainly more like say an airline’s pricing. Airlines have this traditional pricing model that it is based on demand, season and so on. But Uber has to detect.

“Uber is in more cities probably any airline,

Uber goes to more cities than any airline in the world than the biggest airlines. Uber has more passengers more than any airlines in the world.”

So there you go, that's telling you something. Uber has this dynamic pricing and dynamic demand and dynamic supply....dynamic everything. And that's only partially under our control, in fact we don't control it all. The only thing we have in our hand is pricing and money flow, that's it. We don't control any of the drivers, we don't control any of the passengers. So it's not like I own the plane like an airline does. It's a bigger challenge that way.

In Part 2 of our Uber series, we will bring you an insight into Uber’s plans in the coming days.