The BlueConic AI Workbench combines the power of machine learning with the rich profile data in the BlueConic customer data platform. Marketing teams can use machine learning to analyze their data, gain new insights, and further enrich individual profiles.
Using built-in Jupyter notebooks, data scientists can build and train machine learning models to analyze BlueConic data and return new, richer data to user profiles that can be used for segmenting and other CDP use cases.
AI Workbench use cases
With AI Workbench, you can apply the power of machine learning to your BlueConic data. You can use AI Workbench to run machine learning models with data from BlueConic profiles, connections, timeline events, and listeners.
Calculate predictive customer scores based on combinations of profile properties: customer lifetime value (CLV); recency, frequency, monetary value (RFM); or propensity to buy, click, or churn.
Use clustering to discover new customer segments.
Compare different machine learning models to find the optimal algorithm for examining or enriching customer data. For example, you can use the Advanced A/B testing notebook to find the best-performing dialogue for your marketing goals.
Customer timeline events, which can be imported via BlueConic connections, can be used to train a machine learning model for calculating new scores for other profiles.
Use probabilistic or fuzzy matching to find matching profiles in your profile database.
Create new AI-based data visualizations in Python.
Learn more: Visit BlueConic University
Watch Introduction to AI Workbench in BlueConic University.
Setting AI Workbench notebook parameters
In the BlueConic AI Workbench, non-technical users can update the inputs or parameters to an AI model and run the model without writing any code.
Set input parameters for AI Workbench models without coding
In AI Workbench applications, models typically call for customer profile or event data as input. Marketers can use dropdown menus or text boxes in the Parameters tab of AI Workbench to supply the input parameters for running AI models against customer profiles, without writing any code.
How to create notebook parameters
BlueConic provides prebuilt AI notebooks that nontechnical teams can use to run AI models. For custom models, data scientists and developers with the Notebook editor permission can create models in Python code in the built-in Jupyter notebook environment. For these custom models, developers specify which parameters are used in the code, and they provide parameters name. For example, the model might use a segment, and marketing teams supply the values they want to use, for example, "Blog Subscribers." The developer would add the following code to the notebook to create a parameter "Customer segment" for the marketing teams to edit:
# Specify parameters
segment_id = bc.get_blueconic_parameter_value('Customer segment', 'segment')
customer_segment = bc.get_segment(segment_id)
Calling the function get_blueconic_parameter_value() will create the parameter for the marketing team to edit.
Parameter types
Valid parameter types include these types of BlueConic data: channel, connection, date, datetime, dialogue, external_tracker, int (for integer), listener, objective, notebook, profile_property, profile_property_unique, segment, str (for string), and text. For the full list, see the Notebook Parameters API.
Editing notebook parameters
To edit parameters for an AI Workbench application:
In AI Workbench, open the Jupyter notebook that contains your application.
Select the Parameters tab and choose values to supply data to the application.
For example, in the Customer segment field, the application calls for a BlueConic segment. We chose Blog Subscribers for this value.Continue selecting values as called for in your application.
Save your settings. The next time this notebook runs, it will use the values you supplied.
AI Workbench architecture
AI Workbench is an integral part of BlueConic and uses the Jupyter notebook UI for building and training machine learning models on BlueConic data, which can be used for making predictions on other profiles. See the AI Workbench API page for reference documentation.
Resource usage
Your instance of AI Workbench uses a shared resource pool with other BlueConic customers. Note that BlueConic does not share data across customers; you only share resources such as memory and computing power.
Existing BlueConic customers can use AI Workbench free of charge as long as your total data operations remains smaller than the amount that is allotted to your subscription. If you intend to use this feature, contact your Customer Success Manager at [email protected] to ensure your subscription can accommodate your intended usage. We'll discuss your requirements and upgrade your subscription as necessary.
AI Workbench insights
You can view the results of your AI Workbench notebooks on BlueConic Dashboards. Visualize the results of your machine learning algorithms using the AI Workbench Notebook insight to see the results of all notebook cells or the Notebook insight (single cell) to visualize only one cell's results.
Privacy management
Notebooks can be added to BlueConic privacy and consent Objectives, allowing for privacy management of the information that is being picked up. When a notebook is related to an objective, only profiles that have given consent to at least one of the related objectives will be returned to the notebook. As a result, a notebook will only process profiles that have consented to at least one of the objectives that the notebook is linked to.
FAQs
For best practices for specifying Python package versions in AI Workbench, see the FAQ: AI Workbench package dependencies.
Q: What is the Data Scientist role in BlueConic?
A: BlueConic uses roles-based permissions, and you need to have the BlueConic role of Data Scientist to create and run AI Workbench notebooks. To view and edit notebook code, you will need the Notebook editor permission.
Q: Where can I find the AI Workbench API reference?
A: Review the BlueConic Python API documentation site for reference documentation. (This site includes a Changelog section that lists any change or upgrade to the Python package—e.g., the Python version, list of version dependencies.)
Q: I'm a developer working in a Jupyter notebook, and I don't see the changes that another user has applied to my notebook code.
A: While your notebook (kernel) is running, you and your colleague can each work on your own version of the same notebook. If your colleague opens your notebook, makes changes, and saves the notebook, you would need to refresh your browser window in order to see the updated version of the notebook
If you leave a notebook without saving, then the next time you reopen the notebook, the last saved version of the notebook is loaded. If your kernel is terminated (manually or automatically) the latest saved version will be shown when you open the notebook again.
Q: How does AI Workbench handle data for records that are not in the database yet?
A: AI Workbench runs on top of the full BlueConic database, including both anonymous and known data. In AI Workbench, you can also use data that’s not available in the profile, such as weather data or transactions from Shopify or Magento.
Q: Can I use Git in AI Workbench?
A: Yes, the Git command-line tool is included in the container that runs AI Workbench notebooks, allowing you to reference code directly from your Git repository. This functionality helps boost productivity in several ways:
Removes the need to manually copy and paste code from one notebook to the other.
Allows you to efficiently reuse code across different notebooks.
Ensures you are using the most recent and secure version of the code.
Here are examples of how to set this up:
To install a package from a Git repository and specify a branch, use pip with the following syntax:
%pip install git+https://github.com/user/repo.git@branch#egg=package-name
Replace user, repo, branch, and package-name with the appropriate values for the Git repository and package you want to install.
If you don't need to specify a particular branch or package, use pip as such:
%pip install git+https://github.com/user/repo.git
Replace the URL with whatever Git host you want to use, along with the appropriate values for user and repo.
Q: Why am I encountering a notebook validation error?
A: If you are seeing an AI Workbench 404 error, this may be occurring because of the version of the module jupyter-contrib-nbextensions. To remedy the issue, we recommend downgrading the version of this module once installed (or avoid using them). For example, if the version jupyter-contrib-nbextensions==0.7.0 causes an error, try using a version like jupyter-contrib-nbextensions==0.5.1