A yellow arrow pointing to the right.
Fran Britschgi
Fran Britschgi
Solution Architect, AI & Data Science
No items found.
April 23, 2024

Develop a Predictive Model using Snowflake and Sigma

April 23, 2024
Develop a Predictive Model using Snowflake and Sigma


Last year, Sigma shared the potential of predictive ML models explaining how to connect the frayed ends of the ML development pipeline to the line-of-business teams that could stand to benefit from those models, greatly increasing cross-team collaboration and workflows. 

Now, leveraging Snowflake's new machine learning capabilities and Sigma's user-friendly interface, entire organizations can analyze and interpret live data together in real-time, democratizing data science and bridging the gap between the business and the data. The integration makes it easier than ever for ML to be part of business user’s every day work, so organizations can harness the full potential of their data.

Why use Sigma for ML?

Sigma empowers countless customers to think differently about model ops, embedding their ML endpoints directly into their business intelligence workspace for a unified experience, rather than in custom-made, brittle, and single-use siloed data applications. Processes like automated forecasting, generative AI text summarization, and governed classification models can all be deployed by non-technical users in Sigma, while keeping data where it belongs–in the warehouse.

A screenshot of a computer screen displaying a table of data.

Streamlined data analysis

Data producers—like data scientists and engineers—tend to work in Python and SQL. But the individuals who need to put data to work and make critical business decisions are often more comfortable working in spreadsheets. In Sigma, data producers and consumers can work together, in a platform that supports their data analysis method of choice, enhancing productivity and accelerating collaborative insight generation.

Accessible model creation and deployment

Traditional processes of building and applying predictive models are complex and often only practical for a technical audience familiar with coding languages. This process is simplified in Sigma so data producers can now do all of their work in a platform that's not just available, but easily accessible to the broader business to analyze and take action. 

Empowered decision making

With this data at their fingertips, organizations can take predictions and turn them into actionable insights to make informed decisions that drive efficiency, innovation, and competitive advantage.

Ways to Use Sigma for ML

Now, with additional ML capabilities built in Sigma, organizations can expand their applications of predictive modeling to drive efficiency and optimization, deliver personalized recommendations, and enhance decision making.

A screenshot of a website showing a map of the US with a blue bar at the bottom.

Supply chain teams in the retail space can predict future demand to optimize stock levels, reduce waste, and maximize profits, providing a better, more efficient way to forecast sales.

E-commerce companies can perform customer behavior analysis to anticipate customer needs by clustering them with ML methods, improving satisfaction and brand loyalty.

Operations teams in service industries can identify drivers of peak periods, allocating resources more efficiently to improve customer service and reduce costs.

How to use Sigma for ML

Step 1: Data preparation and exploration in Sigma

The initial exploration phase begins with data manipulation and visualization performed in Sigma directly on top of Snowflake without needing to write complex SQL queries or Python code. It enables analysts to format data, aggregate it on different levels (i.e. daily or monthly), and visually explore patterns, and then write back the finalized dataset directly to Snowflake in a secure, governed environment.

Step 2: Model development in Snowflake using Snowpark

The next phase leverages Snowflake's powerful compute capabilities and the Snowpark library to execute data transformations, feature engineering, and model training. In this step, Snowflake handles heavy data processing tasks and trains models directly on its platform using a few simple lines of code—all utilizing the data prepared and explored in Sigma.

Step 3: Use the model in Sigma

This final step showcases the integration between Snowflake and Sigma, where Snowflake's machine learning models are directly called from Sigma workbooks. Utilizing Sigma's custom functions and the ability to call Snowflake model functions results in a seamless workflow from model development to application, all within a business context.

Where is Sigma headed?

Sigma’s commitment to enhancing the integration between Snowflake and Sigma is rooted in the transformative power of accessible predictive analytics. Sigma is continuously exploring ways to refine and expand capabilities, ensuring businesses have the advanced tools they need to thrive in a data-centric world.

Ready to explore the potential of predictive analytics for your business? The comprehensive quickstart guide provides a step-by-step walkthrough of developing a predictive model using Snowflake and Sigma.

Summary

The deepened integration between Snowflake and Sigma heralds a new era in predictive modeling, one where the power of data science becomes a part of everyday business decisions. By simplifying the process and making it accessible to a wider audience, Sigma is not just predicting the future with ML, but helping to shape it.

Dive deeper into the world of predictive analytics with the tutorial and discover how to harness the full potential of your data with Snowflake and Sigma.

WATCH THE PRODUCT LAUNCH