Marketing Analytics Jumpstart consists of an easy, extensible and automated implementation of an end-to-end solution that enables Marketing Technology teams to store, transform, enrich with 1PD and analyze marketing data, and programmatically send predictive events to Google Analytics 4 to support conversion optimization and remarketing campaigns.
APACHE-2.0 License
Marketing Analytics Jumpstart is a terraform automated, quick-to-deploy, customizable end-to-end marketing solution on Google Cloud Platform (GCP). This solution aims at helping customer better understand and better use their digital advertising budget.
Customers are looking to drive revenue and increase media efficiency be identifying, predicting and targeting valuable users through the use of machine learning. However, marketers first have to solve the challenge of having a number of disparate data sources that prevent them from having a holistic view of customers. Marketers also often don't have the expertise and/or resources in their marketing departments to train, run, and activate ML models on paid channels. Without this solution that enables innovation through predictive analytics, marketers are missing opportunities to advance their marketing program and accelerate key goals and objectives (e.g. acquire new customers, improve customer retention, etc).
After installing the solution users will get:
This solution is intended for Marketing Technologist teams using GA4 and GAds products. It facilitates efforts to store, transform, analyze marketing data, and programmatically creates audiences segments in Google Ads to support conversion optimization and remarketing campaigns.
Role | User Journeys | Skillset | Can Deploy? |
---|---|---|---|
Marketing Scientist | Using an isolated and secure sandbox infrastructure to perform and monitor explorations with sensitive data. Using automated machine learning to accelerate time-to-value on building use cases solutions. Faster learning curve to quickly and easily access and analyze data from the marketing data store. Ability to collaborate with other teams by reusing similar components. | Vertex AI, Python, SQL, Data Science | No |
Marketing Analyst | Simplifying the operation of the marketing data store (data assertions), machine learning pipelines (model training, prediction, explanation) and the activation application. Monitoring Ads Campaigns Performance, Web Traffic and Predictive Insights Reports. Interpreting the insights provided to plan and activate Ads campaigns. Defining audience segments using predictive metrics. | BigQuery, Looker Studio, Google Analytics 4, Google Ads | Yes |
Digital Marketing Manager | Gaining insights into customer behavior to improve marketing campaigns. Identifying and targeting new customers. Measuring the effectiveness of marketing campaigns. | Looker Studio, Google Analytics 4, Google Ads | No |
IT/Data Engineer | Building and maintaining marketing data store transformation jobs. Developing and deploying custom marketing use cases reusing a consistent infrastructure. Integrating 1st party data and Google 3rd party data by extending the marketing data store. | Python, SQL, Google Cloud Platform, Data Engineering | Yes |
This solution enables customer to plan and take action on their marketing campaigns by interpreting the insights provided by four common predictive use cases (purchase propensity, customer lifetime value, audience segmentation and aggregated value based bidding) and an operation dashboard that monitors Campaigns, Traffic, User Behavior and Models Performance, using the best of Google Cloud Data and AI products and practices.
These insights are used to serve as a basis to optimize paid media efforts and investments by:
Use Case | Data Sources | Model | Looker Report Name | Activation Event | Google Ads Campaign Optimization |
---|---|---|---|---|---|
Audience Segmentation | Google Analytics 4 | BQML Kmeans | Demographic based Audience Segmentation | maj_audience_segmentation_15 | Custom Data Segments |
Auto Audience Segmentation | Google Analytics 4 | BQML Kmeans | Interest based Audience Segmentation | maj_auto_audience_segmentation_15 | Custom Data Segments |
Customer Lifetime Value | Google Analytics 4 | Vertex AI Tabular Wokflows AutoML | Customer Lifetime Value | maj_cltv_180_30 | Custom Data Segments |
Purchase Propensity | Google Analytics 4 | Vertex AI Tabular Wokflows AutoML | Propensity to Purchase | maj_purchase_propensity_30_15 | Custom Data Segments |
Churn Propensity | Google Analytics 4 | Vertex AI Tabular Wokflows AutoML | Propensity to Churn | maj_churn_propensity_30_15 | Custom Data Segments |
Aggregated Value Based Bidding | Google Analytics 4 | Vertex AI Tabular Wokflows AutoML | High Value Action | - | Static Conversion Values Bid Adjustment |
The solution's source code is written in Terraform, Python, SQL, YAML and JSON; and it is organized into five main folders:
config/
: This folder contains the configuration file for the solution. This file define the parameters and settings used by the various components of the solution.infrastructure/terraform/
: This folder contains the Terraform modules, variables and the installation guide to deploy the solution's infrastructure on GCP.
infrastructure/terraform/modules/
: This folder contains the Terraform modules and their corresponding Terraform resources. These modules corresponds to the architectural components broken down in the next section.python/
: This folder contains most of the Python code. This code implements the activation application, which sends model predictions to Google Analytics 4; and the custom Vertex AI pipelines, its components and the base component docker image used for feature engineering, training, prediction, and explanation pipelines. It also implements the cloud function that triggers the activation application, and the Google Analytics Admin SDK code that creates the custom dimensions on the GA4 property.sql/
: This folder contains the SQL code and table schemas specified in JSON files. This code implements the stored procedures used to transform and enrich the marketing data, as well as the queries used to invoke the stored procedures and retrieve the data for analysis.templates/
: This folder contains the templates for generating the Google Analytics 4 Measurement Protocol API payloads used to send model predictions to Google Analytics 4.In addition to that, there is a tasks.py
file which implements python invoke tests who hydrate values to the JINJA template files with the .sqlx
extension located in the sql/
folder that defines the DDL and DML statements for the BigQuery datasets, tables, procedures and queries.
The provided architecture diagram depicts the high-level architecture of the Marketing Analytics Jumpstart solution. Let's break down the components:
This high-level architecture demonstrates how Marketing Analytics Jumpstart integrates various Google Cloud services to provide a comprehensive solution for analyzing and activating your marketing data.
Note: Google Ads Customer Matching currently only works with Google Analytics 4 Properties linked to Google Ads Accounts, it won't work for subproperties or Rollup properties.
To facilitate the installation, use this Step by Step Installation Video.
Please follow this Installation Guide to accompany the video.
Alternatively, follow the step by step installation guide with Google Cloud Shell.
Note: If you are working from a forked repository, be sure to update the cloudshell_git_repo
parameter to the URL of your forked repository for the button link above.
We welcome all feedback and contributions! Please read CONTRIBUTING.md for more information on how to publish your contributions.
This project is licensed under the Apache License, Version 2.0.
This a list of public websites you can use to learn more about the Google Analytics 4, Google Ads, Google Cloud Products we used to build this solution.
This is not an officially supported Google product. This solution in a work in progress and currently in the preview stage.