Google Upgrades BigQuery With Smarter Business Forecasting

Google's BigQuery now offers powerful forecasting that analyzes multiple related variables, all through a simple SQL command without custom model training.

Google is making sophisticated business forecasting more accessible within its BigQuery data platform. According to a recent announcement on the Google Cloud Blog, the company has upgraded its forecasting capabilities with a new model called TimesFM 3.0. This update introduces multivariate time-series forecasting, a significant enhancement over the previous versions which could only handle single-variable predictions. The new functionality is available directly within BigQuery through the existing `AI.FORECAST` SQL function. This means that data analysts and developers can now generate more complex and accurate predictions about future trends without ever leaving their familiar data warehouse environment, simplifying a process that traditionally required specialized tools and expertise. The "zero-shot" nature of the model means users do not need to train or manage custom machine learning models to get started.

The key technical shift is the move from univariate to multivariate analysis. Previously, a user could forecast a variable, like future product sales, based only on its own historical data. While useful, this approach ignores other factors that influence outcomes in the real world. With TimesFM 3.0, users can now incorporate multiple related variables into their forecasts. For example, a company can now predict future sales by analyzing not just past sales figures, but also marketing campaign expenditures, website traffic, and seasonal trends simultaneously. TimesFM is a large, pre-trained foundation model, which means Google has already done the heavy lifting of training it on a massive set of time-series data. By packaging this power into a simple SQL function, Google abstracts away the immense complexity of model selection and deployment.

This development is part of a broader industry trend of embedding artificial intelligence and machine learning capabilities directly into core data infrastructure. Platforms like Snowflake with its Cortex functions and Databricks with its SQL AI functions are pursuing a similar strategy: bringing the analytics to the data, not the other way around. The goal is to democratize data science. For years, tasks like multivariate forecasting were the exclusive domain of data scientists building complex models in Python or R. By integrating these tools into the data warehouse layer and exposing them through SQL, vendors are empowering a much larger group of data analysts, business intelligence professionals, and application developers to leverage predictive analytics in their daily work. This shift reduces friction and allows organizations to derive value from their data more quickly.

For technical and business teams, the practical takeaway is a significant increase in both the speed and sophistication of their analytical capabilities. Instead of a multi-step process involving data extraction, modeling in a separate environment, and loading results back, teams can now run complex forecasting scenarios with a single query. This directly impacts business operations, enabling more accurate inventory management, better resource allocation, and more reliable financial planning. Looking ahead, we can expect Google to continue this pattern, likely introducing more pre-trained models for other common business tasks directly into BigQuery. It will also be crucial to watch how competitors respond, as the race to offer the most powerful and user-friendly "in-database" AI functions is quickly becoming a key battleground for cloud data platforms.

This update embeds a sophisticated multivariate time-series model directly into the data warehouse layer. For developers and data engineers, it eliminates the need to build, train, and maintain separate forecasting models, drastically reducing the complexity and infrastructure overhead for generating business-critical predictions.

Integrating multivariate forecasting into BigQuery allows businesses to generate more accurate predictions for inventory, sales, and resource planning. This can lead to significant cost savings and better strategic decisions, as companies can now easily model complex relationships between variables without specialized teams.