Skip to main content
CUSTOMER STORY

Satisfying customers through better help desk support

500+ TB

And 40+ data sources and tools migrated, across 750+ users

75%

Reduction in data platform maintenance costs

4x

Faster predictions for triaging help desk tickets

freshworks-header-image
CLOUD: AWS

“The move from Hadoop to Databricks Data Intelligence Platform was the right choice — not only are we more collaborative and productive, but we have unlocked the ability to use machine learning to greatly improve our ability to innovate for our customers.”

— Prasad Ramakrishnan, SVP of IT and Chief Information Officer, Freshworks

Whether you’re in customer service, sales or marketing, your end goal is the same: delight your customers with outstanding products and service. Freshworks makes this possible with a wide range of CRM and customer experience solutions. As its customer base continued to grow, so did the pressures to improve the performance of its customer support organization. Saddled by a legacy Hadoop infrastructure built on Cloudera and an assortment of data tools, the company quickly realized it needed to modernize its infrastructure, which would help it use machine learning to introduce new ways to boost help desk productivity and customer satisfaction. After migrating to the Databricks Data Intelligence Platform, it was able to predictively triage help desk tickets in an automated manner, greatly increasing ticket resolution speeds, which directly impacts customer satisfaction. Looking ahead, Freshworks is confident in the lakehouse as its underlying architecture to deliver ongoing innovations that improve operations and delight customers.

Legacy Hadoop infrastructure stifled innovation

Freshworks’ mission is to help its clients delight their employees by providing solutions for CRM and customer experience (CX) across a wide range of job functions, including IT, customer service, sales, marketing and HR. But with over 60,000 enterprise customers and multiple product lines, the volume and level of support required to maintain exceptional customer satisfaction are both high, which directly impacts its ability to triage support tickets and resolve them in a timely manner. A manual approach to managing help desk tickets was just not providing the level required to keep up with the demand. With so many customers engaging with Freshworks’ portfolio of products, the company saw the opportunity to leverage copious amounts of data being generated by its customers to help it optimize processes and accelerate time to resolution.

However, its internal enterprise data platform — powered by Hadoop (self-managed CDH distribution of Cloudera) — was made of multiple data and analytics tools, which incurred massive IT overhead to manage upgrades and monitor for performance. This environment created performance bottlenecks as data volumes increased and the downstream impact on the customer support team’s ability to efficiently service customers was slowed.

“Our IT and data teams were more focused on maintaining our infrastructure than creating solutions and capabilities that could help the internal customers of our data platform extract value from the data,” said Jeganathan Velu, Director of Business Analytics at Freshworks. “It was clear that we needed to rethink our technology stack to something better.”

To improve its ability to deliver great solutions to its customers, Freshworks looked for a platform that could effectively tackle all the challenges it was facing — and it ultimately selected the Databricks Data Intelligence Platform over native tools from Azure and AWS for its flexibility, multicloud support, and most importantly, its unified approach to data, analytics and AI.

Migrating to the lakehouse brings unification and a path to machine learning

With 500+ TB of data and 40+ data sources and tools across multiple clouds to move, Freshworks knew the migration wasn’t going to be easy, but the whole process was achieved within seven months. When the migration was kicked off, Freshworks first worked with Databricks to identify any dependencies and patterns, allowing it to map out a migration plan across data sources, integrations and endpoints with confidence. This resulted in a faster migration during months two to four of the process, as the team could focus on designing the end framework.

From there, it was an iterative process to bring all of its data pipelines over to the lakehouse. Then, the team had to validate the data and get the final sign-off before the full stack went live. The smooth process was key in helping Freshworks realize its mission, as it could now move its data engineers and data scientists onto a modern lakehouse architecture, thus simplifying data engineering and democratizing data for analytics and machine learning using tools like Unity Catalog without sacrificing compliance requirements around PII data.

“Our internal team was about six members. But we received excellent technical and account support from the Databricks team,” said Velu. “This really helped us make the migration painless and has made our overall data platform maintenance much easier.”

With the Databricks Data Intelligence Platform replacing its Hadoop-based platform, Freshworks was able to not only democratize data access but open up operational analytics through Databricks SQL for various teams to improve decision-making. Delta Lake helps ensure data reliability and consistency across all layers of data as the company builds pipelines to support analytical and machine learning workloads. And with MLflow, its data science team has been able to streamline the machine learning lifecycle from training and experimentation to versioning and deployment.

Improving customer experiences and opening doors to new possibilities

After migrating to the Databricks Data Intelligence Platform, Freshworks saw immediate results. From a cost management standpoint, the simplified approach of the lakehouse architecture reduced overall platform maintenance costs by 75%. With less time spent on IT/DevOps and efficiency features like collaborative notebooks and support for multiple programming languages, its data teams saw a boost in productivity of over 60%.

Additionally, the data processing speeds were 4–5x faster, resulting in efficiency gains within the data science team, as they were able to train their ML models 4x faster. Faster training of models means better predictive insights for the customer support team to rely on when triaging help desk tickets, which has led to higher customer satisfaction.

“Databricks greatly bolstered our engineers’ productivity and made things much easier for us,” said Mohamed Feisal, Senior Manager of Big Data Engineering at Freshworks. “Instead of doing manual upkeep, we can focus on actually creating excellent products for our customers.”

With a modern and unified data lakehouse in place, Freshworks can set its sights on additional ways other products and teams can extract value from its data for operational analytics and AI — all to continue to provide an enhanced experience for its customers.