Marketing Science with BigQuery Canvas: A No-Code Approach to Churn Prediction (Part 1)

Earlier this year, Google Cloud introduced BigQuery Data Canvas, a Gemini feature within BigQuery. This innovative data analytics tool streamlines the entire data analysis process – from data discovery and preparation to analysis, visualization, and collaboration – all within the BigQuery platform. Leveraging natural language processing, BigQuery Data Canvas enables users to ask questions about their data using plain English or various other languages.

I've been thinking about using this for a really interesting use case in Marketing Science. I want to leverage some of the user's app and web behavior data to generate some really interesting insights.

BigQuery Data Canvas eliminates the need for complex SQL queries, exactly what I've been looking for to help abstract technical details from a business user's perspective.

I'll break this down into two articles. In part 1, we'll explore how Generative AI and visual analysis make it easy to access and analyze the data. In part 2, I'll leverage BigQuery Machine Learning (BQML) to craft the perfect machine learning model for some visually stunning predictions.

Here's a business problem that perfectly illustrates my point:

As a marketing data scientist for an online retailer, your mission is to identify customers at risk of churning. To prevent this and boost retention, you plan to engage them proactively with personalized campaigns. Google Analytics 4 (GA4) is already integrated into the website, along with a GA4 to BigQuery export. However, navigating the vast expanse of GA4 data feels like an overwhelming task.

Using… BigQuery Data Canvas

BigQuery Canvas is a great solution. It's a visual tool within BigQuery that lets you explore, analyze, and visualize your GA4 data without getting bogged down in complex SQL queries. And with the added magic of Gen AI, you can even ask questions in plain English!

Here is how my overall Canvas looks like with all the different steps for an end to end workflow...

Lets look at it step by step:

Step 1: A Canvas to solve your Marketing mystery!

1. Open the BigQuery console and navigate to your GA4 dataset.
2. Create a fresh Canvas—a blank slate for your data exploration journey.
3. Add your GA4 dataset to the Canvas, ready to unlock its hidden gems.

Step 2: Asking Questions with Gen AI

Question 1 - Show me the number of purchases made by users from each marketing campaign.

Instead of writing complex SQL queries, you use the Gen AI tab in Canvas to ask questions in natural language. For example, to understand user engagement, you ask:

"Show me the number of sessions in the last month, the total number of sessions, and the number of days since the last visit for each user."

In this case, i asked a question about “Show me the number of purchases made by users from each marketing campaign.”

Then, i can ask another interesting question which will be in the area of Data Engineering, because i am asking to create new variables. Question would be 

Question 2 - Calculate the number of days since the last visit, the number of sessions in the last month, and the total number of sessions for each user.

Another question to create a new variable would be

Question 3 - Calculate the total number of purchases, total revenue, and average order value for each user

Step 3: Visualizing the Data

Canvas's visualization tools let you create histograms that reveal the distribution of generated features. This helps you make sense of user behavior and flag any outliers or data quality issues.

In the next article, we would join the different tables and then create a Churn Prediction Model with BQML

This approach empowers marketing analysts to leverage the power of machine learning and Google Analytics data without needing extensive coding expertise. It enables data-driven decision-making and proactive customer engagement to reduce churn and drive business growth.