Replacing Looker with Cube Cloud for Customer-Facing Embedded Analytics

About the Company

The data team from a car dashcam technology company realized that while their dash cams are an important and visible product, delivering data and analytics from those dash cams is an even more important component of their mission. The company had been collecting data about how enterprise customers were using the company's APIs and it was clear that sharing the analytics of the API usage with customers would be incredibly valuable.

Evaluating Looker: Complexity and Cost

As a first step, the Infrastructure Team Lead needed to expose that usage data into their developer portal at which point they could then design a simple front-end application that would share the analytics with customers. The company already used Looker with Databricks, but it was clear that it would not work for this embedded analytics use case. Looker was too complex to set up: it required deploying a separate instance of Looker dedicated to that particular use case - and that was a “bit of a hassle” says the Team Lead. And, more importantly, Looker came with a hefty price tag.

“Implementing Looker is a pain, especially compared to the lean model of Cube where you don't have to serve a full dashboard experience.”

Building from Scratch: A “Bunch of Trouble”

Some of the engineers on the team were working on building this connection from scratch. But the Team Lead realized that not only would they be taking on all the hassles of typical software engineering, all of these hard coded SQL queries in a back end application would be difficult to maintain.

“I kind of had my eye on Cube Cloud for a while. When we realized we needed to expose API metrics to the developer portal, I thought to myself, this is a perfect use case for Cube Cloud.” It would save the team a “bunch of trouble” to avoid building it themselves, especially because they wouldn’t need to maintain it and Cube Cloud was so easy to provision.

When the Team Lead shared Cube Cloud with the engineering team, “they were kind of incredulous at first”. The team didn't know such a thing existed. Once the Team Lead explained Cube Cloud, the team jokingly wondered “how did we not know about this before?”

Cube Cloud: Speed without “Blowing Up” Your Warehouse Costs

Not only was Cube Cloud quick to deploy and easy to maintain, with Cube’s pre-aggregations in the caching layer, they were not subject to the latency constraints of the data warehouse which, the Team Lead commented, “is very, very powerful if you want to serve an API at a consistent latency SLA without blowing up your data warehouse costs.”

Security Context in Cube Cloud Saves Time

The Team Lead loved that in Cube Cloud everything is done around the definitions and is done declaratively, “you just basically describe your schema and the rest is done. The rest is processed by Cube itself.” The only exception to this is how you handle the security context where you can put in some imperative code. And they definitely appreciated the ability to enforce security at the semantic layer instead of in the front end or in the data warehouse.

“Security Context is incredible. I love how it works. You can actually enforce things like row level security so easily but at the API layer.”

Cube Cloud: 2 Engineers Deployed in a Month

With just two engineers: one backend engineer and one front end engineer, the Company spent about two to three weeks to get a working prototype. Then it was only a few weeks later and they were ready to deliver a production level product and could go live.

“Using Cube Cloud is very cost effective. It saves you a bunch of time and trouble. It’s not trivial to deploy and maintain and operate another app in the cloud.”

Expanding Use Cases to Other Apps

The team also deployed a geospatial analytics app on top of Cube Cloud to accelerate the queries and also expose those queries as APIs. “We were happy we didn’t have to develop our own back end application just to translate REST in JSON to SQL and back and forth.” As far as impact, this app turned out to be very important to the sales team. At one point, they were really on edge about potentially losing access, because it was an important tool for them to help close deals.

“I see Cube as a key technology in our data stack because it serves a very real purpose in adding semantics, speed, and security administration at the API layer. It's something that you don't want to build yourself. I can see us using Cube much more in the future.”