Customizable embedded dashboards and natural language AI queries drive customer analytics

The Cube x Quantatec user story.

Quantatec logo
Customizable embedded dashboards and natural language AI queries drive customer analytics
HQSãO Paulo, Brazil
StackPostgres, Angular, Azure Kubernetes Service, Azure OpenAI, Langchain
Use Case Embedded Analytics

Making fleet management data-driven

With more than 32 years of experience in automotive electronics and more than 23 years of leadership and innovation in the automotive tracking segment, Brazilian company Quantatec specializes in tracking, logistics control, fleet management, accident, and cost reduction for fleets of vehicles of any type and size.

Quantatec provides fleet managers access to their tracking device data in Movias, their exclusive fleet management platform. Mauricio Cirelli manages the Movias platform. The embedded analytics solution, available on the web and mobile, presents account-specific data to hundreds of customers in South America. Six million tracker messages are received and processed daily, amounting to about 1TB of data annually.

Supporting the growing demand for data

Before Cube, Quantatec’s multi-tenant embedded analytics solution delivered static, hard-coded reports with basic filtering. Fleet managers soon wanted customizations, such as adding a column for the driver’s name or the fleet name to the report. Each request was slightly different. The team would either create a new report or modify an existing one in a way that would not impact others.

Demand was high. A growing number of unique SQL queries for the current user were sent to the database to return account-specific data. The increased volume of direct database queries combined with real-time data processing from over 10,000 devices resulted in unpredictable database performance and degraded customer experience.

Quanatec Pre Cube Architecture

The traditional three-tier application architecture struggled to support the demand for data because of the volume direct database queries and real-time data processing.

When customers began asking for their own interactive dashboards and the ability to customize them, Mauricio knew that more custom queries would only add to the existing performance issues. He began searching for a solution. It turned out to be just the beginning of a series of major innovations for Quantatec’s customers.

Improving query response time

Mauricio began researching possible solutions for the performance issues. He found BI platforms like Power BI and Tableau to be expensive and proprietary. He experimented with Redis as a stand-alone cache, but it didn’t meet requirements. In testing, Cube exceeded expectations and allowed full control of the customer experience.

They moved forward with Cube Cloud. Mauricio shared, “Before Cube, we had no knowledge about the power of a universal semantic layer. We were only looking for a cache and dashboard API in our research. Cube has added a lot of value to our software and saved a lot of time in development due to the nature of our platform.”

By implementing Cube’s pre-aggregation and caching capabilities, query performance improved from tens of seconds to milliseconds, often bypassing database interactions altogether. Ensuring data security and privacy, Cube seamlessly filtered data based on each customer's fleet with its query rewrite capabilities for row-level security. This laid the foundation for rapid data analysis and customizable dashboards.

Quanatec Architecture with Cube

With the addition of Cube into the application architecture, direct database queries are often avoided altogether.

Delivering on self-service customizations

Because fleet managers lacked software and programming skills, the team developed a user-friendly dashboard creation tool. Mauricio favored simplicity to enable self-service customizations. Two months after implementing Cube Cloud, customers were empowered to design their own dashboards without intricate knowledge of database structures. In just a few clicks, fleet managers could easily select fields, generate visualizations, and save their changes.

Quanatec Embedded Dashboard

Dashboards are easily customizable by fleet managers with the user-friendly dashboard creation tool.

Making data conversational

Movias quickly evolved to the next level with the addition of a chatbot with natural language processing capabilities. Inspired by Patterson Consulting’s demo to showcase an LLM chatbot's potential, Mauricio imagined a user-friendly, chat interface over Quantatec’s fleet data. Mauricio partnered with Josh Patterson, CEO of Patterson Consulting, to realize his vision.

Using the Cube’s universal semantic layer, OpenAI API, and LangChain framework, Mauricio and Josh introduced Movias AI, a chatbot interface, capable of interpreting natural language queries and fetching relevant data from Cube's cache or database as needed. Development took only about six weeks from POC to production. With chatbot integration, Quantatec provided a truly unique experience to deliver data to fleet managers. Questions can be written in any language supported by the LLM. In particular, English, Spanish and Brazilian Portuguese were evaluated.

Mauricio shared, “Ultimately, fleet managers have questions and want them to be answered. We can do that by providing them a report or a dashboard, but it is much more efficient for them to just ask the system and get a clean response to their questions. And, thanks to Cube, most of these queries are not even hitting the database.”

By augmenting user queries and reasoning over a well-defined data model, Cube’s semantic layer powered accurate and actionable insights effortlessly from the LLM. This paradigm shift marked a game-changing capability, where fleet managers could simply articulate questions such as "Who's my worst driver?" or "What is a vehicle's fuel consumption history?" and receive instant answers.

Josh explains, “Cube's semantic layer played a pivotal role in giving the SQL Agent the context it needs to reason about which tables can answer the questions posed.” With a unified API, access controls, and well-defined data model, Cube facilitated better query construction and interpretation by the LLM. Moreover, the integration of prompt analysis agents mitigated hallucinations, ensuring accurate query interpretation and response generation.

LLM Architecture with Cube

With Cube’s semantic layer, LLMs return accurate answers because the data model provides context for the LLM to reason over.

Driving better business outcomes

Solving the initial database performance issues served as a springboard to fully leverage Quantatec’s investment in Cube Cloud. Combining the power of the universal semantic layer’s centralized data modeling, access controls, pre-aggregation caching, and native APIs with natural language processing, Quantatec revolutionized fleet management operations, empowering fleet managers to extract actionable insights with ease.

Mauricio reports, “We implemented pretty much everything Cube has to offer: multi-tenancy, data modeling, query-rewrite, pre-aggregations, time zones, and semantic improvements on the data layer. We even found many ways to optimize our database schema.”

Ready to upgrade your data stack?

Related Use Cases

Check out Cube’s other solutions

Related Blog Posts

Stay up-to-date with the latest from Cube