Zomato, Swiggy using AI, machine learning to drive more growth3 min read . Updated: 19 Sep 2019, 10:33 AM IST
- Both Swiggy and Zomato boast of handling over a million orders a day across more than 300 cities
- Each customer order is now being influenced by the customer’s own previous history of order preferences
Bengaluru: When food-tech company Zomato let go of around 540 of its support staff last week, it said improvement in its after-sales technology forced its hand. The automation made up for almost 10% of the workforce in certain support roles across Zomato’s customer, merchant, and delivery partner teams, redundant.
While routine jobs will continue to give way to automation, food-tech companies such as Swiggy and Zomato are increasingly turning to machine learning (ML) and automation to drive their businesses, using years of data accumulated from food orders and user-level consumption patterns.
Both Swiggy and Zomato boast of handling over a million orders a day across more than 300 cities. And each customer order is now being influenced by the customer’s own previous history of order preferences.
Swiggy boasts of more than 1.3 lakh restaurant partners on its platform, while Zomato claims to have added around 1.5 lakh restaurants. With such a large supply base in place, both food-tech apps are now primarily using data to tap demand.
Swiggy, for instance, banks on its history of order-level data and real-time feet data to reduce customer wait times and for retaining customers. “We are processing around 40 billion messages (or data points) per day, which are unique data points purged from either customers ordering from our app and from drivers delivering orders. And if I look at that, scale, will probably touch 100 billion messages within a year," said Dale Vaz, head of Swiggy’s engineering and data sciences department in an interview.
According to Vaz, Swiggy is using data analytics to individually curate the customer landing page—list of restaurants—to each user’s taste preferences rather than just curating on the basis on the customer’s location. According to him, food is a personal choice, and cannot be generalized on the basis of the location of the customer alone. The company said that it’s building a concept known as “food graph" which breaks down a food dish by recipe, cooking style, ingredients used, calorie value, and variations of the dish.
By combining the food graph with a customer’s previous food preferences, Swiggy is able to derive a personalized restaurant feed at the app homepage itself. “For example, a user may prefer an Andhra style biryani over a Lucknowi style biryani… so we are trying to get to that level of precision in our understanding of the customer," added Vaz.
Unlike, e-commerce, where data sets are derived from customer purchases in the hyperlocal segment, the delivery fleet, restaurants, and customers together generate massive amounts of data points. This also means that restaurants can make sense out of the order-level data and past history of customer preferences.
Zomato currently provides a business dashboard for restaurants on the web as well as the app, according to a company spokesperson. “We also help restaurants understand customer behavior by educating them on purchase funnels such as awareness (metrics) like the number of visitors on the page to Intent (building a cart) to purchases (placing an order). We have now started tracking dish-level ratings and are working towards adding more features to the dashboard," the Zomato spokesperson added in a written response.
According to Ankur Pahwa, partner at consulting firm, Ernst & Young, e-commerce companies “are also trying to increase retention of the same customer, so that the lifetime value of the customer increases. The same thing is playing in the foodtech segment".
However, some experts believe food delivery platforms still have a long way ahead. According to Manish Singhal, founding partner of pi Ventures, India is still a price-sensitive economy where most customers will only order a dish from the cheapest platform available to them.
According to Singhal, it is very difficult to model human nature accurately when it comes to food. He concluded, “Artificial Intelligence (AI) might be able to do a good job in predicting and bringing back the customer to the platform but it will only make an incremental difference--it will not change the world for these companies. But this is not because the AI models are not mature enough--it’s more because of human nature."