You are here:
Actionable Variables and Prescriptions
In AI Models (formerly Einstein Studio) Model Builder, which uses data from Data 360, actionable variables are fields in a model created from scratch (AI-created) that you can change to influence a predictive outcome. In constrast to other fields that you can't change, such as a customer's location, gender, and age, actionable variables can be changed. These variables are important because they provide suggested actions (prescriptions) that users can take for more succesful outcomes.
Here are some actionable variables and how you can adjust them.
- Credit Limit: Offer a limit increase
- Interest Rate: Set the rate for different customer groups
- Product Price: Increase the unit price of a product
- Shipping Option: Offer free or faster shipping
After you set actionable variables for your predictive model, you can use these Salesforce integrations to get prescriptions.
Batch Data Transforms
Use a batch data transform to run a model on a large dataset. The model generates predictions for each data point based on the actionable variables that you set in Model Builder. Batch data transforms use actionable variables to process prescriptions. These prescriptions are added to columns next to each prediction in the dataset, and are valuable for decision tools such as dashboards.
[Variable Name]: [New Value] - Discount_Offered: 10%
[Impact Value] - 20 (This value means that if you change the discount from 5% to 10%, the predicted conversion score increases by 20 points).
Flow
Use automated flows to turn predictive insights into automated, personalized, and prescriptive actions such as to update prices or initiate offers. This improves explainability and confidence in decisions. After setting actionable variables:
- The model doesn’t just return a prediction score.
- The model can also return prescriptions that suggest what actions to take to improve an outcome. We recommend using Prompt Builder for a more user-friendly output such as: Extend an offer period by 7 days.
These prescriptions become output fields in an invocable action (cdpMlPrediction) in Flow. You can build screens that show what to change for specific customers, products, or regions. Flows can also:
- Dynamically adjust messaging, discounts, or routing based on prescribed values.
- Segment customers based on what actions are likely to have the most impact. For example, offer a discount or escalate to a service rep.
If your flow has screen elements such as a screen flow that's created for reps, you can:
- Describe the prescription in more detail.
- Provide expainability for users. For example: Offer a 10% discount because it can increase the purchase likelihood by 20%.
Let's look at the flow logic when a new lead is added to a system.
- Trigger: New lead
- Prediction: Low likelihood of conversion
- Prescription: Suggest follow-up within 24 hours
- Flow Result: Assign the lead to top service rep and schedule a call on the same day.
This flow closes the loop: you’re not just scoring a lead, you’re also prescribing and automating an action so that a rep can follow up with the lead.
- Set Actionable Variables
In Model Builder, actionable variables are fields that can you can change to influence a model's predicted outcome. Setting actionable variables enables the model to provide prescriptions (suggested actions that a user can take to improve an outcome).
