進行状況の追跡を始めよう
Trailhead のホーム
Trailhead のホーム

Create a Next Best Action Strategy

What Is a Strategy?

Think of your next best action strategy as kind of a flow chart. By the end of the flowchart, you’ve assigned a next best action to each of your customers.

You can also think of a strategy as a filter that sorts through recommendations and assigns them to customers based on predictions made for those customers.

Recommendations filtered through a strategy

Strategies can have a lot of different types of elements and can be quite complex. For this project, you use just two elements.

  • A Load Recommendation component Load icon, for adding a recommendation to your strategy
  • A Branch Selector component Branch Selector icon, for adding branches (choices) to your strategy’s workflow

When you apply a strategy to a set of objects, like your VeloMateo customers, Einstein Next Best Action assigns recommendations to them based on their predicted behavior. These recommendations are the next best action for each customer.

Create a Strategy

  1. If you’re not already on the Setup page, click Setup icon and choose Setup from the drop-down menu.
  2. From Setup, enter Next Best Action in the Quick Find box and select Next Best Action.
  3. Click New Strategy.
  4. Name your new strategy VeloStars NBA Strategy.
  5. Under Objects Where Recommendations Display, choose Contact. Then click Done.
    The Strategy Builder window comes up.
  6. Load a recommendation into the strategy by dragging the Load icon Load icon (on the left side) into the main window. The Load window appears. Fill it out as follows.

    Load window

    • Label: Current Member: 20% off bars
    • Field: Description
    • Operator: Contains
    • Value: 20
  7. Click Done.
    You set the Value to 20 because it matches the “20” in the description you gave one of your recommendations (“Offer 20% discount on Marshvelo Bars”). It’s important to use a unique identifier. A single Load node can in fact load multiple recommendations; if you had used “Offer” instead of “20,” that would have matched more than one recommendation.
  8. One by one, drag three more Load icons onto the main window and fill them out as follows. (Don’t worry yet about where to place them.)
    • High likelihood
      • Label: High Likelihood: Join VeloStars
      • Field: Description
      • Operator: Contains
      • Value: VeloStars
    • Medium likelihood:
      • Label: Medium Likelihood: 10% off bars
      • Field: Description
      • Operator: Contains
      • Value: 10
    • Low likelihood:
      • Label: Low Likelihood: 15% off purchase
      • Field: Description
      • Operator: Contains
      • Value: 15
  9. Drag the Branch Selector icon Branch Selector icon into the main window and call it Is a Member? and then click Done.
  10. Drag a second Branch Selector icon into the main window and call it Likelihood of Joining, and click Done.
  11. Arrange the icons to look like this.

    VeloMateo Next Best Action strategy layout

    It’s important that you arrange the icons and connectors exactly as shown in the image, or the following steps won’t work. (It might take you a little while to get the hang of moving the icons around.)

    Einstein processes the strategy from left to right, but to read it as a flowchart, read from right to left, starting with Output.

    Now you populate the forks in your tree. You start by determining whether or not someone is already a VeloStars member.

  12. Double-click the Is a Member? Branch Selector icon. You’ll see that it contains two branches. Branches are ordered from top to bottom in a strategy diagram, with Branch 1 at the top. If a branch is displayed with the Advanced tab selected, and there is text (such as “true”) in the Filter Expression box, delete that text to switch the view to the Standard tab.

    Populating the "Is a Member?" Branch Selector

    1. For Branch 1:
      1. Click the Standard tab.
      2. Click Add Condition.
      3. Click in the Resource box. From there, choose Contact ($Record) > VeloStars Member (the inserted value will be $Record.VeloStars_Member__c.).
      4. Operator: Under BOOLEAN, choose Equals.
      5. Set Value to True.
    2. Repeat the steps for Branch 2, but set Value to False.
    3. Click Done.
  13. Double-click the Likelihood of Joining Branch Selector icon.
    1. Again, for all branches, if a branch is displayed with the Advanced tab selected, and there is text (such as “true”) in the Filter Expression box, delete that text.
    2. For Branch 1, add a Condition in the Standard tab:
      • Resource: Contact ($Record) > Membership Likelihood (the value will be $Record.Membership_Likelihood__c)
      • Operator: Greater Than Or Equal To
      • Value: 30
    3. For Branch 2, add a Condition in the Standard tab:

      Populating the "Likelihood of Joining" Branch Selector

      • Resource: Contact ($Record) > Membership Likelihood
      • Operator: Greater Than Or Equal To
      • Value: 20
      • You want to make the second next best action for folks whose predictions fall between 20 and 30. Add a second condition by clicking Add Condition, and fill it out.
        • Resource: Contact ($Record) > Membership Likelihood
        • Operator: Less Than
        • Value: 30
    4. For Branch 3, add a Condition in the Standard tab:
      • Resource: Contact ($Record) > Membership Likelihood
      • Operator: Less Than
      • Value: 20
  14. Click Done, then click Save to save your strategy.
    Try testing some results by clicking the Test button. (Here’s where those four or more contacts you viewed earlier come into play.) Do you get a variety of recommendations for the different contacts? Do the recommendations match the predictions you see when you look at your Contacts list? (For example, does a customer with a high likelihood of joining VeloStars get a recommendation to join?) If not, take another look at your strategy diagram. Make sure every element is in the right column and that every column is properly connected to the column to its right. Check your Branch Selectors, too—did you set them up with the correct prediction values?