How Relata built completely native embedded analytics that runs 85% faster.

The Background

Relata is a B2B SaaS startup based in London. As a revolutionary AI-powered relationship intelligence and management platform, it helps businesses unlock the full potential of their relationships—and the ones they want to build.

The value proposition of their proprietary technology—at the intersection of next-generation NLP, data analytics, and machine learning techniques—is that it can consume a tremendous amount of network and relationship information and translate it into actionable insights for salespeople and managers.

Hence, of course, the clever portmanteau, ‘Relata’—or rather, ‘relationship data.’

With Relata, salespeople get a comprehensive understanding of real-time client engagement (everything from email opens and response times to outreach optimization), relationship health, and even answers to questions such as, “Is anyone in my network connected to this prospect?”

Managers, on the other hand, access information about the entire sales organization, with dashboarding about success rates of individual salespeople, network bottlenecks, relationship health trends, upselling opportunities, and so on.

Relata manages to synthesize all of the powerful relationship intelligence data a company has into a potent, single source of truth: actionable, accessible information containing opportunities to both mitigate risks and unlock significant value.

To paraphrase the adage, “With great data comes great presentation layer responsibility.” Of course, it’s relevant here because Relata wanted to optimize the usability of its platform so that customers could get the most out of it. And with Cube, they really did—to tremendous customer feedback.

The Challenge

Before moving to Cube, Relata’s team used a big-name, traditional BI platform as the basis for their customer-facing dashboards. In it, they would create star schemas and data models; the platform would orchestrate measures, aggregations, and chart generation. Relata then embedded its iframes into the web application to display dashboards and charts to end-users.

However, the team wanted flexibility and fast performance—which aren’t qualities traditional BI supports in embedded analytics. Embedding iframes from their previous solution also became difficult to troubleshoot and not ideal from a best practices and scalability perspective.

Relata was looking for granular control over data modeling, versioning, performance, and, most importantly, a way to create a native, streamlined data experience for their users. Upon realizing their previous solution couldn’t provide any of those things, they evaluated several other traditional BI platforms—only to conclude they all shared the same limitations.

Fortunately, Cube’s semantic layer does not.

The Requirements

Relata was looking for a solution that could serve as the basis of semantic layer-powered embedded analytics while also offering:

• Universal compatibility with front-end frameworks
• Ability to build a fully native UX/UI
• Flexible data modeling
• Improved embedded dashboard speeds
• Granular performance management, including of refresh rates and pre-aggregations
• Proper dashboard versioning via integrations with tools such as GitHub
• Flexibility to support and power features beyond the web application
• Streamlined scaling capabilities

“We realized—pretty much instantaneously—that Cube was the way for us to go. The problems we were experiencing with our previous traditional BI solution were the worst—and they were simply non-issues with Cube’s semantic layer.” — Francesco Mancusi, Development Team Lead at Relata

The Solution

Relata’s team needed a BI solution that would give them more flexibility and the ability to build a customized, scalable and high-performing analytics dashboard for their customers.

Through a friend’s referral, Francesco Mancusi, Development Team Lead, found one; here’s how he and his team implemented Cube:

They first built a star schema in Postgres for their raw data stored in AWS Redshift. An ETL process extracts data from sources like email and calendars and transforms it into atomic elements of interactions. This data is then reprocessed in their ETL to create more dedicated tables tailored to specific views in Cube, which are then uploaded into Cube to power the analytics portion of Relata’s application based on React, Python, and Django.

To automatically deploy updates to Cube, the team uses Terraform, CircleCI, and Argo CD for continuous integration and delivery. They also configured VPC peering and transit gateways to allow Cube to access multi-tenant data in Redshift. And to improve application performance, they took advantage of pre-aggregations and granular control over refresh rates in Cube.

After onboarding, it took two engineers about a month to build an initial prototype—and two and a half months later, they went live with Cube, with a full implementation including pre-aggregations optimization and stress testing.

And as for results?

First and foremost, Relata found major value in its ability to serve as a universal API layer. The team was able to build completely native interfaces—in the standard languages, they already knew.

Moreover, on top of gaining the ability to fully customize the UX/UI of the front end, they now also have granular control over data modeling, which enabled them to build a more “interconnected” user flow. And the team noted significant performance improvement with pre-aggregations and granular control of refresh rates in Cube.

All in all, Francesco and the Relata team built and delivered a tailor-made, streamlined UI that, on average, performed 55% to 85% faster than before—to an “extremely positive” user reception.

The Future

Looking forward, Relata is considering supporting more features with Cube, such as new methods of delivering reports to their users through real-time insight feeds and other notification methods.

And for advice to others considering Cube Cloud, Francesco says: “Embrace the philosophy of Cube...It's a different way to build things with respect to what I was used to, so we lost some time trying to make things work the same way they used to work in regular business intelligence tools. Cube is more like an engine than a dashboard—so that it can be much more tailored to the use case.”

We can’t wait to see Relata continue to grow—and we’re thrilled we can help.

Special thanks to Francesco for taking the time to speak with us—and check out Relata to see how they’re revolutionizing relationship intelligence.

---

Have an exciting Cube case study and want to be featured? Drop us a line: hello@cube.dev

Interested in managed Cube Cloud for your data modeling, access control, and caching needs? Request a 1:1 here, or find us on Slack and GitHub.