Enriching travel experiences with data and machine learning
Increase in click-through rates of proactive campaigns
“Thanks to Databricks, we have been able to unify our data and analytics infrastructure to unleash the power of data so we can use the right insights to better connect with customers.”
As India’s leading online travel company with an impressive combined desktop and mobile MAU of 45 million, MakeMyTrip sees machine learning (ML) as a key driver for continued growth by delivering personalized experiences for every customer from recommendations to in-location engagement. As the amount of data received increased exponentially, so did the complexity of data and analytics at scale. MakeMyTrip turned to Databricks to unify data and accelerate ML innovation to deliver customer experiences.
Siloed data across the business creates campaign complexity
Known for their service quality and a wide product range, MakeMyTrip often found themselves either inundating customers with too many options or offering too few. With siloed business verticals — hotels, flights and cars — each inputting their consumer data differently, cross-sharing data for a complete view was near impossible. These massive and diverse data sets led to complex, inefficient data pipelines that incurred significant technical debt due to inconsistent and duplicate data.
“It was not possible to efficiently prepare data to effectively train our models,” said Piyush Kumar, head of data platform engineering at MakeMyTrip. “We had a jungle of data pipelines that were extremely difficult to maintain and slowed our ability to support our data scientists.”
Having migrated their complete data platform stack from on-premises to AWS cloud, MakeMyTrip also needed a way to ensure seamless migration for data deployment and orchestration.
Another issue was the lack of cross-team collaboration between data engineering and data science. Real-time featurization often required heavy data engineering support. But with disjointed teams and different programming languages, productivity was low and adversely impacted model training.
MakeMyTrip turned to Databricks to unify their data and gain a centralized, consistent view of their customer segments at scale, so that they could better forecast and cater to customers’ needs.
Unifying data and ML enriches customer experiences at scale
With Databricks’ help, MakeMyTrip developed Feature Store, an internal unified data analytics platform that helps them build reliable data pipelines, simplify featurization and accelerate model training. This enabled MakeMyTrip to enjoy actionable insights into what their customers want, at scale, so they can drive richer, personalized online experiences.
Delta Lake brought reliability to MakeMyTrip’s data lakes and facilitated the accurate analysis of data. Tapping on Apache SparkTM, they are now able to unify streaming and batch data with ease. MakeMyTrip can now easily ingest, extract and transform millions of user-generated signals into a unified data repository that’s a single source of truth for featurization and model training.
“Delta Lake allows us to create a common data model, no matter the data source,” explained Piyush. “This allows us to more easily stitch data together for downstream pipelines.”
Partnering with Databricks enabled MakeMyTrip to expedite the deployment and orchestration of its data lake. Furthermore, the segregation of computing, storing and running streaming jobs became more efficient within the new Databricks Spark Runtime environment.
MakeMyTrip’s teams also enjoy greater cross-team collaboration through synchronized access to the same customer data streams via interactive notebooks. This, in turn, enables MakeMyTrip to deliver more innovations quicker to enhance customer experiences.
More importantly, the Databricks Data Intelligence Platform prepared MakeMyTrip for growth.
“Databricks consolidates all user data so that our teams can have a single, unified view for richer insights. The beauty is that it’s scalable, without limitations, and highly capable of handling a quantum amount of data,” said Piyush.
With a complete view into their customers, they are able to execute more impactful marketing campaigns to improve customer lifetime value by delivering the right message to the right customer through the right channels.
Valuable data insights enable curated experiences
The benefits of the new Databricks Data Intelligence Platform are multifold and have created new opportunities across the funnel. Data collection, tracking, maintenance and refresh are now more efficient and synchronized. Data analysis has also more accurately segmented customer data in a much shorter time, uncovering valuable insights that allow MakeMyTrip to better curate and personalize experiences.
“The innovations in the data platform with Feature Store are helping marketing by leveraging ML models to intelligently target relevant customers per the customer lifecycle journey and drive micro-segmented communications to improve the relevance of campaigns,” said Somil Agrawal, director of marketing, head of customer lifecycle management and loyalty program, at MakeMyTrip.
The ability to deliver more defined and granular customer segmentation has also led to new forms of data-driven customized communication, such as fare alerts and hotel room availability countdowns, generating enthusiasm and a sense of urgency. Over the longer term, MakeMyTrip has observed that consistently providing relevant recommendations translates into user trust, which accelerates the rate of returning customers. As a result of deploying Databricks, the click-through rates of proactive campaigns have increased 20%.
“Databricks has made our centralized Feature Store a reality,” concluded Piyush. “Not only does this help simplify operations and streamline our ML lifecycle, it serves as a foundation for new data products designed to enhance customer experiences.”
