TimesFM 3.0 in BigQuery (Preview)

Google's forecasting model in BigQuery can now predict several related numbers at once and factor in known drivers like holidays or promotions, for BigQuery users.

Details

Earlier versions forecast one number at a time with no outside inputs. Version 3.0 forecasts multiple related numbers together and takes in extra factors
It accepts two kinds of factors: ones you only know from the past, like foot traffic, and ones you know ahead of time, like holidays or planned promotions
No training needed. You call it from SQL with AI.FORECAST and set the model to TimesFM 3.0, and existing forecasts switch over by changing that one setting
Each forecast comes back with a predicted value plus upper and lower bounds
You can also ask for a forecast in plain language through BigQuery's Conversational Analytics, which writes and runs the query and returns charts
In preview. The announcement does not list pricing or regions

Best for

Forecasting metrics that move together, using outside factors you already track.

When to use it

Your forecasts need to account for things like scheduled promotions, weekends, or holidays, and you don't want to train a custom model.

Who benefits

Data analysts and analytics teams who work in BigQuery.

Staying informed is great. Staying ahead is better.

Pick a topic and leave with something you created.

Start for Free Bring to Your Team

Get new module drops and weekly AI strategy.

Details

Best for

When to use it

Who benefits

More from Google

Staying informed is great. Staying ahead is better.