Paytm Boosts Homepage Sales with Personalized Recommendations Using Amazon Personalize

Financial services company Paytm wanted to increase the sales and click-through rates of its ecommerce service, Paytm Mall, by implementing personalized recommendations on the homepage of its site. To create this recommendation model, Paytm needed a robust machine learning (ML) solution to analyze and propose recommendations to over 10 million daily active users who shop for retail goods on Paytm Mall.

Paytm turned to Amazon Web Services (AWS) and used Amazon Personalize, a fully managed ML service, to create a personalization model that generates recommendations for each customer. Amazon Personalize makes it easy for developers to build applications with the same ML technology used by Amazon.com for near-real-time personalized recommendations with no ML expertise required. Using Amazon Personalize, Paytm has increased its sales and the click-through rates of the Paytm Mall homepage while making it simpler for its customers to find items.

Paytm is a digital payments, ecommerce, and financial services provider located in India. Today, it supports over 17 million merchants and is used by millions of individuals daily to pay for utilities, groceries, movie tickets, and more. Paytm’s services also include travel booking, insurance, and retail—the Paytm Mall—that are all contained within the Paytm app. The company wanted to create an ML model to present personalized recommendations to its users on the Paytm Mall homepage based on their browsing history. Previously, the company did not have personalized recommendation services on a user level, but it did have widgets for item-to-item recommendations.

The company researched the option of creating an in-house solution, but ultimately Paytm determined that using Amazon Personalize would be quicker to implement with better turnaround time. Paytm already used AWS for its application deployments, so it was familiar with the environment. The company began implementing Amazon Personalize in May 2021, and the personalized recommendations model went live in September 2021. Paytm was supported during the transition by the Amazon Personalize engineering and support team. “We received excellent support from the AWS team, especially during the initial phases of development,” says Ankur Gogate, technical lead at Paytm.

Using Amazon Personalize and other AWS services, Paytm collects user data and runs it through the recommendations model to generate unique content suggestions for each of the over 10 million daily visitors to the Paytm Mall. To supply the data outputs that power the personalization model, Paytm uses an in-house Java-based application and Amazon EMR, a cloud big data service for running large-scale distributed data processing jobs, interactive SQL queries, and ML applications using open-source analytics frameworks. After the user data is processed through Amazon EMR, it is sent to Amazon Personalize to run through the personalization model, which returns personalized recommendation results that are pushed to the homepage. The recommendations create an individualized experience for each Paytm Mall visitor. “Personalization helps users get what they need in the fewest clicks possible,” says Gogate. “By having personalization using Amazon Personalize, we create a unique app for every user based on individual choices and preferences.”

Since adding the personalized recommendations model, Paytm has seen a 5.5–6 percent conversion rate from the Paytm Mall homepage. By comparison, the widgets that Paytm had been using previously for item-to-item recommendations had seen a 1.8–2 percent click-through rate. “This conversion rate is higher than with any other recommendations model we’ve had,” says Gogate. “Because these recommendations are on the homepage, people can get the item that they want right there. It’s helped increase our total orders and the volume of sales from the homepage itself.” Implementing the recommendation solution has also helped Paytm better measure activity on its homepage because it was not gathering metrics on homepage conversion rates previously.

Using Amazon Personalize has boosted Paytm’s agility as a business and has helped the company continue innovating. “After the user personalization went live, we realized how quickly and easily we could incorporate Amazon Personalize with other new solutions,” says Gogate. Paytm was considering creating another solution for showing brand-based recommendations. Instead of building something from scratch, the team was able to repurpose what had already been created using Amazon Personalize to get a beta up and running quickly. This new brand-based recommendations model, which is currently in testing, will give users recommendations based on brands that they have previously interacted with on the app. The beta was implemented in only 1 month using Amazon Personalize, whereas the team estimated it would have taken twice as long for a similar model to be created in house. “We’re seeing a very simple plug-and-play approach using Amazon Personalize,” says Gogate. “This is very beneficial for us as a company to be able to bring new recommendation models to users quickly.”

Paytm is looking to see where else it can implement personalization on its application in the future, beyond the Paytm Mall. “Having Amazon Personalize has opened up new spaces on Paytm to implement personalization,” says Gogate. “Now Paytm may soon start looking at various ways that we can use user personalization to get better results in other business areas, such as travel and insurance.” With more personalization, Paytm wants to help its customers efficiently get the services and items that they are looking for.

Ultimately, Paytm has benefitted from not only increased sales but also new opportunities. “Using Amazon Personalize has helped us innovate and rethink our whole approach to improving sales conversions,” says Gogate.