Real-time intelligence: How India’s Swiggy serves millions with Microsoft Fabric

Indians are embracing delivery apps, placing millions of orders every day for everything from biryani to cell phones. 

Swiggy, one of India’s biggest on-demand convenience platforms, received 923 million orders in the financial year ending March 31, 2025, up 22% from the previous year.  

Buying trends are hyper-local and seasonal. While groceries remain the top selling category in quick commerce, users buy everything from team jerseys and snacks during cricket season to gold coins during Diwali.  

That makes for complex logistics in an industry that competes on speed and service. 

“We are always on our toes to keep solving, keep optimising delivery times and the operational hiccups that come with them,” says Vipinkumar Tiwari, Swiggy’s Chief of Tech Staff. “Say all of a sudden there is rain and the roads are choked up and people are not able to get out. How can we ensure timely delivery and a good customer experience?” 

In the past, it could take up to 10 minutes to update operational data on dashboards, a significant lag when food orders are expected in 30 minutes and other items in 10.       

Swiggy has turned to technology tools, including those from Microsoft, to stay ahead. In recent months, it deployed Real-Time Intelligence in Microsoft Fabric, a platform to process and analyse streaming data—from inventory levels to road conditions—and give actionable insights in seconds. 

It also rolled out generative AI chatbots using Microsoft Azure OpenAI Service to communicate these insights to operations staff, customers and delivery drivers.  

A logistical web 

Swiggy started in Bengaluru in 2014 with food delivery. In 2020, it launched Instamart, its marketplace for grocery and household items, a segment which is doubling each year. It has also gone into the business of restaurant reservations and organising events. Swiggy went public in November 2024, raising $1.34 billion dollars. It now operates in more than 700 cities across India. 

“People now have the income to go and spend on some of these things,” says Madhusudhan Rao, Chief Technology Officer of Swiggy. “The number of people who are willing to pay for convenience is growing.”  

There are now 23 million monthly transacting users on the Swiggy app on average, who place 3 million to 4 million orders each day across food delivery and quick commerce. More than 690,000 delivery riders pick up these orders from over 260,000 restaurants, as well as more than 1,100 'dark stores' which are local warehouses stocked with the most popular items. 

Targeted discounts 

To crunch the billions of data points collected daily, Swiggy earlier this year started using Microsoft Fabric Real-Time Intelligence (RTI), which was launched just a year ago. One of the first uses was to weed out misuse of discount coupons, says Deepanshu Thakur, Co-founder of Mandelbulb Technologies, an IT company that specialises in deploying Microsoft Fabric. 

Like many retailers, Swiggy dangles targeted discounts to keep customers loyal and boost orders at certain times of the day. There have been cases where these discount codes are leaked on social media and misused by others, which translates into less money for Swiggy.  

Fabric RTI helps detect instances where there’s an unusually high number of people using discounts, and Swiggy can discontinue the coupon. The real-time data also helps update inventory on the app as well as customer expectations. 

In the past, if sanitary pads were about to run out at a particular dark store, there would be a time lag of five to ten minutes for that information to show up on the app, during which a user might unwittingly order the item. The result? Frustrated customers and harried store managers. 

With Fabric RTI, the seller is immediately alerted to add stock while the customer is told straight off the bat that it may take longer to get the pads to them from a warehouse that’s farther away. 

‘Where is my order?’ 

Swiggy is also using Microsoft Azure OpenAI to automate its contact centre and answer customer questions such as "where is my order?" without having to add extra staff during peak hours such as mealtimes.  

Fabric RTI also gives the back-end team visibility on potential bottlenecks. For example, if they see orders pouring in from particular neighbourhoods, they can direct delivery riders to head to the area. 

Previously, “the data had to be loaded in batches, refreshed continuously,” Rao says. “We weren’t able to be waiting for these insights in a business where orders need to be delivered in under 10 minutes.” 

Another AI-powered chatbot—Driver Dost—is aimed at delivery riders, helping them with onboarding, tracking earnings and planning delivery routes. 

Swiggy picked Microsoft after evaluating competitors because of the choice of AI models as well as ease of building and deploying solutions using Microsoft Foundry, essentially providing the “plumbing and pipelines,” says Tiwari. 

In the future, Tiwari says, Swiggy plans to use generative AI to analyse and improve its dealings with restaurant partners. Currently, Swiggy is reliant on the manual notes that sales teams upload, which can be hit and miss. 

It’s also looking at providing similar consumer intelligence to the brands it works with, so a laundry detergent brand, for example, can see trends in what’s selling, where and when.