Blog

The Indian intelligence behind UBER’s AI

Vidya Duthaluru Head of UBER Engineering for customer care at an Economic Times event

Uber came to Bangalore India in 2014, five years after it launched in the US in 2009. It’s hard to understand the impact that a company providing cabs on request has made across the world. I recall the old way of having to make a cab booking the previous night for a 4 a.m. ride to the airport the next morning. Invariably, the lines would be busy at around 9 pm, because several other customers had the same idea. Then providing directions on the route home followed by an SMS with driver details.

The next morning, if the driver was more than a few minutes late, panic set in because alternatives at that early hour were not easy to find. If the driver did not respond, call the cab company and then, they would try and find an alternative as the minutes ticked by. Several calls back and forth later, the problem would eventually be solved with unnecessary stress. You get the idea. Until Uber came along, we had no understanding of the intensity of human involvement in a simple cab booking chore. Now, of course, we simply whip out our phones, enter the destination, confirm our ride and watch as the little cab icon weaves its way through the streets towards the pickup point. 

Uber is no cab company

Booking cabs
Photo by Charles Deluvio on Unsplash

There’s this quaint notion that Uber is just a giant version of a local cab company. Nothing could be further from the truth. Uber is a local transport facilitator globally. Until Uber came along, cab companies in India were small businesses and fleet sizes of 400-500 made them ‘big’. There were several manual operations to be managed, right from booking and cancellations to driver co-ordination, cash payments, permits, and maintenance. Scaling the business was practically impossible. The US did have nationwide chains like Heinz and Avis. But India had nothing comparable. 

From Ryan Holiday’s Growth Hacking post on Quora, this is the playbook that Uber followed worldwide: Uber, a car service start-up founded by Travis Kalanick and Garrett Camp, has been giving out free rides during Austin’s SXSW Conference for several years. During a single week, thousands of potential Uber customers—tech-obsessed, high-income young adults who cannot find a cab—are motivated to try out this service. One year Uber offered free rides. Another year, it offered BBQ delivery. Instead of spending millions on advertising or countless resources trying to reach these potential users in their respective cities, Uber just waited for the one week a year when they were all in one place and did something special. That’s thinking like a growth hacker—it’s how you get the most bang for your buck and how you get it from the right people.

App without a user manual

Intuitive design that onboards UBER customers
Photo by Dan Gold on Unsplash

Uber used technology and existing global infrastructure to grow into a behemoth. But they had to ensure that enough cabs were available when customers searched right from the beginning. The app had to be as easy for drivers to use as it was for customers. It did not come with a user manual which was a given for practically every product launch. Do you remember making your first Uber booking? There was little you had to do apart from entering your destination and confirming your ride. And drivers would focus on figuring out what they would earn every day. Uber’s genius was in tying up lots of little threads with technology and working out what needed to be done next.

The small things that we have seen over the past few years in India. The fare details displayed before you hopped in. The time it would take you to arrive at your destination calculated in advance. The masking of your number when calling the driver. The introduction of messaging to communicate with the driver. The payment options that made it easy for you to simply board and exit without paying cash. Of course, you can’t expect that drivers would not be smart and figure out ways to ensure that they did not take up rides beneficial for them. But the rating system ensures a minimum level of civil behavior on both sides. Customers don’t want to be locked out of the system. Drivers don’t want to lose their job. So, some level of seesaw fairness prevails.

Indian intelligence that drives Uber worldwide

The lady steering Uber customers worldwide

Uber’s global engineering team for customer care is lead by Vidya Duthaluru from Bangalore. The engineering teams from San Francisco and Palo Alto report to her. The size of the operation? This from a report in The Economic Times: “Uber had $65 billion in gross bookings in 2019, a 35% year-on-year increase. It has clocked 7 billion trips and has 5 million drivers on its platform”  Can any ‘cab company’ dream of hitting those kinds of levels without technology and machine learning?

What are Vidya’s credentials? She spent over 2 decades in the AI field, developing early conversational AI at Nuance, a company in the customer care space that worked on understanding customer intent. It has grown to become one of the leaders in the field. At the core of Uber’s operations, it’s about managing support worldwide in order to deliver a customer experience by deploying machine learning and supporting various functions within Uber. The size of the operation?  80 plus countries, 800 plus cities, every city with its own traffic patterns, supply constraints, and rider demand issues to deal with. 

Building systems that operate at a global scale

Responding to complex routing, matching and traffic patterns
Photo by Osman Rana on Unsplash

Data drives every decision within Uber – from matching customers to drivers, analyzing trip details, understanding how promotions work and when they should be applied. While demand and patterns can be predicted, individual behavior cannot be. Riders can cancel, drivers can cancel and that cascades through the system since all of this is happening in almost real-time. The moment riders make a request, machine learning is applied to show the rider a set of local destinations booked in the past so that it becomes easier to make the booking. The history of the rider is checked to ensure that it is not linked to past fraudulent transactions. 

Once the booking is made, decisions on the fastest route to apply depending on local traffic patterns and time of day factors. That has to be matched to the numbers of cabs available within a certain distance so that the rider does not have to wait too long for the cab to arrive. This happens millions of times a day across the world. And just to correct the impression from the Quora quote that Uber doesn’t advertise, the company spends around $ 500 million in the US alone on advertising and promotions annually. Brands don’t get built or sustained without the oxygen of recognition and recall.

Read Next: Starbucks stars briefly in GOT

Or this: India’s No.1 deodorant does not have a website

Author short bio: I head Ideascape, an agency that I started in 2004. I have over 35 years of experience building brands in businesses as diverse as payroll services, software, cycles, HR services, hospitals, hospitality and project management.

We’re a boutique creative agency but we provide the full range of branding services in partnership with several associates in digital marketing, web development, and event management. This blog is a collection of my experiences and my point of view on marketing and advertising

Related posts

  • India's engine for women entrepreneurs

    A happy confluence - the rapid expansion of India's telecom network, the explosion of cheaply available data and the price drop of smartphones led to a startup that could redefine women's emancipation in India

  • Something's Tookitaki about this!

    Starting out as an adtech player and pivoting to a fast-growing sector, Tookitaki has made big strides in regtech, making anti money laundering detection and transaction monitoring the core