Python Models

Build advanced data transformations, ML predictions, and statistical analysis with Python templates.

Last reviewed September 15, 2026

Python models run custom Python code against your BigQuery data for advanced logic. Examples are customer segmentation, predictive scoring, and recommendation engines. Start from a pre-built template or write your own.

Creating a Python transformation

Open Studio.
Click New Data Model.
Select Python.
In Input data, select the connected sources that the model reads.
Set the output table name and the Write mode.
Customize the Python code and parameters.
Save the model.

Template Categories

Category	Templates	Use Case
Recommendations	Product affinity, cross-sell models	Suggest products based on purchase history
Segmentation	RFM analysis, behavioral clustering	Group customers by behavior or value
Prediction	Churn prediction, LTV forecasting	Predict future customer behavior
Analytics	Customer Acquisition Cost, Cohort Retention Analysis, Forecasting	Measure acquisition cost, retention and trends

Input Tables

Map one or more BigQuery tables as inputs to your model. Each input table is available in your code as a pandas DataFrame, with the alias of the input as the variable name.

Output Configuration

Setting	Description
Output Table	Name of the BigQuery table to write results to
Write Mode	`replace` (overwrite table each run) or `append` (add new rows)

Writing Python Code

Your code runs in a secure environment with these libraries:

Library	Version	Purpose
pandas	2.x	Data manipulation
numpy	1.x	Numerical computing
datetime, math, json, re, collections, itertools	Python standard library	Dates, math, JSON, regular expressions and iteration helpers

Code Structure

Your code receives the mapped input tables as DataFrames. It must set the output or result variable to a DataFrame:


import pandas as pd
 
# RFM segmentation example
rfm = orders_df.groupby('customer_id').agg({
    'created_at': 'max',        # Recency
    'id': 'count',              # Frequency
    'total_price': 'sum'        # Monetary
}).reset_index()
 
rfm.columns = ['customer_id', 'last_order', 'frequency', 'monetary']
output = rfm

Note: Vendo disables external network access for security. All data must come from the configured input tables.

Scheduling and Triggers

Python models can run in these ways:

Trigger	Description
Manual	Run on demand from the model detail page
Source trigger	Run automatically when a source app finishes syncing
Model trigger	Run after an upstream model completes

Chaining Models

Set a source model trigger to make a pipeline of models. For example:

Source sync imports Shopify orders
SQL model computes daily revenue (triggered by source sync)
Python model runs RFM segmentation (triggered by the SQL model)

Vendo detects circular dependencies and prevents chains longer than 10 models.

Best Practices

Keep models focused: one transformation for each model.
Use descriptive output table names: for example customer_segments or product_recommendations.
Handle edge cases: check for empty DataFrames and null values.
Start from templates: customize a pre-built template instead of writing from scratch.
Use replace mode for idempotent outputs, and append for time-series accumulation.

Data Studio
SQL Models: for query-based transformations
Audiences: for user lists that you sync to ad platforms
Agent → Building Models: create Python models with AI assistance