I Tested Google’s TimesFM 3 in BigQuery Against a Traditional Model (leans into the head-to-head)

TimesFM3 in BigQuery: AI Forecasting Without Training Data (Tested Against Real Results) Google's TimesFM3 foundation model is now available in BigQuery — and it can forecast your time series data with zero training. I tested it against a real week of NYC bike share data and compared it head-to-head with a traditional ARIMA model. Here's what happened. In this video, I walk through the entire process: setting up BigQuery, cleaning the data, running the AI.FORECAST command, and checking the forecast against what actually happened. I also show how to use BigQuery's Conversational Analytics to get the same forecast with plain English instead of SQL. ⏱️ Chapters: 0:00 Introduction & Demo Preview 1:30 What is TimesFM3? 2:23 Setting Up BigQuery 2:49 Understanding the Data 3:10 Cleaning the Data 3:53 Setting Up the Forecast Week 4:07 Running the AI Forecast 4:48 Comparing with Traditional Model 5:51 Evaluating the Results 6:23 Conversational Analytics 6:50 Who Should Use This? 7:14 Wrap-Up 🔗 Links & Resources: • TimesFM3 on Google Research: https://research.google/blog/timesfm/ • BigQuery public datasets: https://cloud.google.com/bigquery/public-data • NYC Citi Bike trip records: https://data.cityofnewyork.us/transportation/ 📌 About this channel: I'm Surya — 22 years in enterprise tech, taking the cutting edge of AI and making it make sense. New videos every week on practical AI for analysts and data teams. All opinions are my own and do not represent my employer. If you found this useful, hit the like button and subscribe for more! In this video, we put Google's TimesFM 3 to the test in BigQuery, comparing its performance against a traditional model in a head-to-head matchup. The TimesFM3 is a powerful tool, but how does it stack up against established methods? We dive into the details of our experiment, exploring the capabilities and limitations of TimesFM 3 in BigQuery and examining the results of our comparison. If you're interested in learning more about the potential of TimesFM 3 and how it can be utilized in BigQuery, this video is for you. We'll take a closer look at the features and functionality of TimesFM 3 and discuss the implications of our findings for those looking to leverage this technology. By the end of this video, you'll have a better understanding of the strengths and weaknesses of TimesFM 3 in BigQuery and how it compares to traditional models.