Foodtech startups have revolutionised the way Indians consume food. Customers can order whatever they want to eat, whenever they crave food, and from any place of their choice. With a valuation of $3.3 billion, 1.4 lakh restaurants covered, active 2 lakh+ delivery personnel and a presence in 500 cities, Swiggy is one of the largest homegrown startups in India that delivers food to the customers.
From letting customers select a food item, getting the order ready, and finally delivering it on time is not an easy task. Dale Vaz, Head of Engineering & Data Science at Swiggy has empowered all stakeholders in its business ecosystem — restaurants, drivers, and customers — with technology to make the entire process smooth and seamless.
Applying Tech to Food
Vaz is determined to turn Swiggy into an AI-first company. The starting point for this has been to build a large data lake. Crucial data with respect to food preparation, delivery, and transactions is accumulated in the data lake, and actionable insights derived from it.
“We’ve done a lot of work to gather the data in a single place and build this data platform that allows us to build insights and models on top of it,” says Vaz.
The in-house software developed by Swiggy brings the restaurants, drivers, and customers all on the same page to ensure complete visibility and efficiency into their business processes.
According to Vaz, timing is the essence. Keeping a track of time is imperative at every step — when a customer orders, the restaurant starts its preparation to finding the right driver to deliver it.
With machine learning algorithms, data from previous orders is processed to fetch insights on the precise time drivers will take to travel from the customer location to the restaurant or the restaurant to the customer location, taking factors like timing of the day, festive season, weekends and so on into consideration.
The company is also able to predict how much time the restaurant will take to prepare food.
Operations optimization models have also been built, which enable Swiggy to assign the best possible driver to a customer order while meeting constraints of delivery time and costs in real-time.
Though the company has a mix of different machine learning and deep learning models, Vaz calls the system that predicts travel time at the minute level key for Swiggy as the company provides a precise promise to customers.
Treating every Customer as Unique
For enhancing the search experience, Swiggy has built a custom AI model that ranks all the restaurants uniquely for every single customer. The list of restaurants is unique for each customer, matching his preference.
“The AI-model looks at not just your past orders but also things like affinity that a customer has towards a certain restaurant. So we can say on the basis of the kind of food you have purchased in the past whether you would prefer a south Indian meal over an Italian meal or vice versa,” avers Vaz.
The company has also wrapped intelligence into categorizing food on the basis of which it offers recommendations around dishes and restaurants that serve similar food.
Helping drivers and restaurants
Vaz’s team has also built ML models for restaurants, which help them in managing inventory and forecasting supply for Swiggy orders. Vaz explains they do predictions on identifying how many dishes of a certain type will be sold in a given day. This information is shared with the restaurant to plan for the demand and optimize preparation.
To help drivers get more orders, a tool called Heat Map is created which predicts from which area the next orders will come in.
“With close to 90 percent accuracy, Heat Map allows us to guide the waiting drivers to head for those particular areas where more orders are expected to come from”, adds Vaz.
Swiggy has also launched an ‘Access Program’ that helps restaurant partners in deciding and opening kitchens at new locations. Analytical models have allowed the company to identify locations with enough customers with the preference of a particular dish.
Creating a new revenue stream
Advertising is another business that Swiggy focuses on. The company has built an in-house tech stack which is a highly scaled system that serves ads to millions of customers.
“Advertising benefits restaurants by allowing them to place targeted relevant ads for customers. It is a great stream for us because it allows us to monetize our traffic and gain revenue,” says Vaz.
“AI-based model ensures the ads seen by customers are relevant to them. We don’t show any spam or try to sell meaningless ads to customers,” he adds.