QR technology company Flowcode is the offline to online company building direct connections between brands and consumers. More than 10,000 companies use Flowcode to generate advanced QR codes that are beautifully designed, privacy-compliant, and backed by data and analytics.
Each QR code scan provides brands with valuable insights about its customers. To provide customer-facing analytics as part of its enterprise plans, Flowcode ingests and analyzes millions of user interactions with its QR codes. This rich data is then internally used for its data science activities.
Flowcode turned to Snowflake to keep up with rapid growth as brands seek to connect the offline world with the online world.
Supporting Data Scientists Without Overwhelming a Small Data Engineering Team
Previously, Databricks was primarily used to run machine learning (ML) workflows as well as being the analytics source of truth. While a small subset of data scientists were comfortable with the tool, they were spending a lot of time managing and scaling environments to prepare data rather than building models. Outside of this team, accessing the data and running self-serve analytics was not an easy task because most of the organization was only familiar with SQL.
In order to enable data scientists to spend as much of their time building models, a strong partnership developed between the ML team and the newly formed data engineering team. But as a small team, keeping up with the internal demand for data was a challenge. Scaling and maintenance of existing Spark-based pipelines were taking time that prevented timely and reliable access to data. “The environment management complexity was preventing us from moving fast,” Flowcode’s Data Tech Lead, Uttam Kumaran, said.
Boosting Data Engineering’s Productivity While Remaining Lean
Flowcode chose Snowflake on Azure because it simplifies data operations and frees up technical staff to focus on higher-value work. With Snowflake, Flowcode was able to not only provide a single source of truth for both structured and semi-structured data; it also paved the way for self-service analytics both internally and externally. With Snowflake, no manual tuning was required and it seamlessly scaled to meet the demands of thousands of concurrent users consuming Flowcode’s advanced analytics and reporting offered through its premium plans.
“With Snowflake, we can sit our strategic product offering for advanced analytics to external customers on the same data platform our data scientists use for machine learning.”
—Uttam Kumaran, Data Tech Lead, Flowcode
To focus on core business functions as much as possible, Flowcode also uses Fivetran connectors to streamline data loading and reduce the need for custom ETL. And with Snowflake Secure Data Sharing, Flowcode will enable live data sharing with its partners without having to manage any pipelines or move any data sets. Powering Flowcode’s embedded Looker visualizations with Snowflake enables customer-facing analytics for thousands of concurrent users. “The only reason I’m able to focus on actual core business functions is that a lot of traditional DBA activities are taken care of,” Kumaran said.
Increasing Collaboration with Data Scientists
Architecting on Snowflake provides a SQL-based framework for Flowcode’s data scientists to access high-quality data and focus primarily on building models. “They’re able to access high-quality data from a single source and can continue to use their preferred tools for model training,” Kumaran said.
To build a training data set, data scientists access and prepare data using the Snowflake Connector for Spark. With the connector, the Spark operations are translated into a SQL query and then executed in Snowflake to improve performance. Once a model is in production, continuous data loading with Snowpipe automates the ingestion of model data back into Snowflake, which increases collaboration and streamlines the handoff to data engineering. According to Kumaran, “My team is not only tasked with getting the data but also with delivering machine learning driven insights to analysts so they can solve business problems.”
Enhancing Customer-Facing Analytics with ML Model Data
Using Snowflake to deliver richer insights in near real time is a top priority for Flowcode. “Every new data point that we get enhances the information that we can feed into our ML models to provide richer insights in our customer data product,” Kumaran said.
Kumaran sees Snowflake’s Direct Share capability as an opportunity to further democratize Flowcode’s data beyond its partner ecosystem. Account-to-account data sharing will provide customers in the enterprise plan with instant and secure access to data for a fraction of the effort.
Expanding Flowcode’s use of Snowflake will help the data engineering team deliver more value while staying agile. According to Kumaran, “Snowflake is the right partner for Flowcode because it offers a scalable and easy-to-use platform to drive business outcomes while supporting the needs of every data team.”