Unlocking the potential of data with instructor-led training

Lacking lakehouse expertise hampers downstream analytics and AI needs

Data drives new innovations across every aspect of the sales process, from prospecting to closing deals. At Outreach, their sales execution platform is powered by data and AI. The engineering team continuously harnesses massive datasets for analytics and AI use cases, from in-product recommendations to delivering various insights to internal teams tasked with making their products better and more engaging. “These 24-hour-cycle reports look at the performance of teams, the success of sequences and overall how we can convert sales effort into revenue more efficiently,” explains Nick Ambrose, Director of Engineering at Outreach.

Internal Outreach teams need to make data-driven decisions from methods such as email sentiment analysis. “This is the main area of AI for my team,” adds Ambrose. “We analyze emails to determine the mood behind what is written.” The engineering team also extracts job titles from emails to see how important one meeting is over another. Engineering delivers this data to the BI team, who leverage these insights alongside their own sales data to provide business insights about the Outreach customer base and their usage of Outreach. This generates terabytes of data daily.

“Overnight processing time was taking longer than it needed to. My team is fairly new to the Databricks Platform and Spark,” says Ambrose. “So we looked into the training courses offered by Databricks to speed up onboarding and improve the reliability of our build processes.”

In addition to getting their engineers more proficient on Databricks, Outreach needed to shorten onboarding times for new engineers so they could be more productive with the data faster. “It’s critical for our on-call engineers to be able to research a customer problem. And as we onboarded more people to my team, we had to bring them up to speed with all the tools,” explains Ambrose. “We needed to get them trained on our core technologies, including Databricks, so that everyone starts with the same basic foundation.”

Democratizing data with Databricks training

The primary value Databricks instructor-led training has provided the Outreach team is the democratization of data and insights. Outreach bolstered the expertise and capabilities of their engineering team, which put the business teams they supported in positions to succeed with data. They also eliminated previous barriers that kept other teams across the organization from reaching their full potential with the Databricks Data Intelligence Platform.

For Outreach, what set Databricks training apart from typical courses was the hands-on nature of the workshops. “The best thing for me was that a good 60–70% of the Databricks training was hands-on. We were running things, actually typing in the notebooks, and solving problems with an instructor present to help as needed. That way when the time comes in real life to work within the platform, you can do it with confidence,” says Ambrose.

Since completing the Databricks training, his entire engineering team has learned valuable ways to use Databricks more cost-effectively and how to optimize the speed of nightly builds. The course that added the most value for Outreach’s power users was the Optimizing Apache Spark™ course. “The optimizing Spark course shed light on what we needed to do to fine-tune our environment for peak performance. It was a game changer,” says Ambrose.

Building a foundation for complete data empowerment

Databricks training has helped to more quickly instill a data-first culture at Outreach. “Engineering onboarding is now about 50% faster than before training, which means our engineers and data users are becoming more productive with their data faster,” explains Ambrose. “This has a downstream effect on our ability to deliver value to the business faster as well.” Outreach also learned a lot about how Databricks stores their Delta files, which allowed them to safely reduce the amount of data stored and ultimately lower storage costs. In fact, they cut 200 terabytes of Amazon S3 storage, saving them roughly $50,000 per year in AWS costs.

Right now, the power users — the core engineers — are focusing on more deep-dive training; this is about 10% of the engineering department. Others benefiting from training are the machine learning teams and data platform engineers, as well as some nontechnical business users like data analysts and data scientists who want to be more self-sufficient on Databricks.

Eventually, Outreach wants all data users to take at least the basic training so they have a baseline understanding of how to be productive in Databricks. This will help remove bottlenecks for support engineers — the more people who can self-serve within Databricks, the fewer requests they have to field. When builds are both faster and more reliant, Outreach can focus on what it does best: provide better insights to sales teams looking to close deals, faster.

“It has served as a launching pad for data democratization across the organization. With Databricks training, we hope to eventually give all our data users the foundation to be productive on Databricks and our power users the specialized skills to go above and beyond,” concludes Ambrose.