Skip to main content
CUSTOMER STORY

Driving meaningful connections between customers and brands

32%

Lower TCO compared to their multicloud data warehouse

15%

Faster queries compared to their multicloud data warehouse

20%

Improved performance of supplier reporting

CLOUD: Azure

When it comes to advertising, consumers want to see content that’s relevant and personalized — especially if it’s appearing on their mobile devices, where time is a valuable resource — so it’s important to instantly capture the consumer’s attention and drive engagement. InMobi does this by using real-time customer data to deliver targeted mobile advertising and lock screen experiences. But as data processing requirements increased to 20+ terabytes per hour, the cost of running their multicloud data warehouse skyrocketed, while its proprietary nature created silos that hindered collaboration and data sharing. InMobi migrated from their multicloud data warehouse to Databricks to unify their various workloads (data warehousing, AI and analytics), streamline operations and free up their engineers to work on more valuable tasks, helping them to achieve better operational agility and efficiency. Since moving to the lakehouse, the company has not only significantly reduced their total cost of ownership (TCO) compared to when they used a multicloud data warehouse — they have also improved productivity across the organization, resulting in faster time-to-market of new products.

Complex legacy infrastructure and multicloud data warehouse unwieldy to manage

InMobi is all about targeted advertising and helping brands reach and engage consumers in a meaningful and cost-effective way. But to accurately deliver relevant ads, you need data — a lot of it. Over time, they extended their on-premises Hadoop system by adding multiple cloud data warehouses to address various problems. However, as the amount of data the company needed to process increased exponentially (20TB of data per hour), InMobi continued to build on top of their legacy system, resulting in a multicloud data warehouse that had a host of challenges: It was overly complex, had outages, was extremely costly as they scaled, and led to data silos that limited data sharing and collaboration. The team at InMobi realized that if they continued with this system in place, it would slow down their ability to innovate and keep precious engineering resources locked into maintenance mode.

“The data infrastructure that we built worked, but it created significant complexity and overhead that pulled our focus away from our own core products,” said Madhan Sundaram, Senior Director of Platform Engineering at InMobi. “We needed our talented engineers to create more value for our customers, and to do that, we needed a simpler and more unified system.”

What InMobi wanted was a single system that could solve multiple problems. To accomplish this goal, the company needed to consolidate their disjointed systems into a single platform to free up engineers to focus on higher-value tasks such as developing ML and large language models. That’s why the team looked to the Databricks Data Intelligence Platform to unify their data warehousing and AI workloads on a single platform.

Migration to lakehouse results in unified data, analytics and AI

Although InMobi has a team of fully capable engineers, they learned a valuable lesson in productivity and operational agility. “We found that we were spending more time maintaining our environment, which was hurting our ability to support and collaborate with the business,” explained Sundaram. “We wanted to be more strategic and identify ways to operationalize data more efficiently.” After careful evaluation of their current multicloud data warehouse and the prospects of building in-house, they determined that the Databricks Data Intelligence Platform clearly aligned best with their goals — to improve developer productivity through reduced infrastructure complexity while achieving the best possible price/performance.

Once they made their selection, the team partnered with Databricks to start the migration planning process. With over a decade of customizations built on top of existing systems and over 1 petabyte of data, InMobi knew the migration was going to be complex. On the ETL side alone, there were 150 pipelines and eight teams involved with migrating from open source Apache Spark™. They also needed to migrate approximately 300 reporting dashboards to ensure there was no disruption of information delivery to suppliers and customers. To help navigate this process, Databricks worked closely with InMobi — which had allocated two in-house engineers per team to provide support — and Databricks’ implementation partner Celebal Technologies on the optimizations needed for a seamless migration.

As a result, Databricks was able to help InMobi migrate from their multicloud data warehouse to the lakehouse with ease. “Databricks gives us the edge in terms of being able to utilize industry-first features that help us establish technical differentiation,” said Sundaram. “Their expertise came through during the migration as they helped us optimize our compute resources for cost and performance. The ownership that Databricks took to drive the optimizations and how transparent and educational they were was commendable.”

Bringing more personalized mobile ads to customers everywhere

Now, with a unified, more streamlined lakehouse architecture, InMobi is able to take full advantage of their robust customer data to deliver smarter, more personalized mobile advertising. Various teams leverage Databricks notebooks for ad hoc analysis, Power BI for visualizations on top of Databricks SQL — the serverless data warehouse on the lakehouse — and MLflow to build their next-generation AI platform. They’ve also found great success with Delta Live Tables for anomaly detection, with a 50% improvement in SLAs and an 80% reduction in costs. Data silos and data discoverability are no longer an issue either, thanks to Unity Catalog. With Unity Catalog, they can now govern access at the table and column levels, while ensuring complete data lineage is captured to ensure they have visibility into where data comes from and whether it’s stale or not. With a platform designed to meet their analytics and AI needs, the InMobi team is diving into new technologies, including large language models (LLMs), to help deliver insights to their customers more efficiently. “We’ve started to look at implementing LLMs to make it easier for our end users to ask a question and find the information they need. Lakehouse architecture will make this effort easier, as jobs will run automatically under the hood. This will give our teams the ability to simply ask a question without any expertise and get the contextual answers they want at their fingertips,” explained Sundaram.

In terms of business impact, there’ve been a number of measurable improvements across the board after the migration. Not only are infrastructure costs 34% lower than before, but there is also a 15% boost in query speeds and 20% fewer job failures compared to their previous data environment, all contributing to 20% better performance of reporting and insights delivery to end users. The TCO is 32% lower compared to when they used a multicloud data warehouse, and there’s been a 24% cost reduction in running their ETL pipelines as well. More qualitatively, the team is seeing overall better reliability — with systems more stable than ever — and experiencing a positive reputation boost with their customers.

“Our rate of experimentation has improved tremendously,” said Mohit Saxena, Co-founder and Group CTO at InMobi. “Databricks has simplified collaboration, and the unified approach that the lakehouse offers allows us to be more efficient, productive and compliant when delivering new features and products.”

By moving to a unified platform on the Databricks Data Intelligence Platform, InMobi can now focus on innovating in the mobile advertising space to deliver real-time personalization that drives value for both InMobi’s customers as well as their internal end users.