How to build and use a rule-based model

Context 

Rev.Up allows you to build an unlimited number of rule-based models. This article covers what a rule-based model is and some example uses for it. 

What is a rule-based model? 

Rule-based modeling allows you to create your own custom scoring model. In the rule-based model wizard, you are able to define rules that partition a chosen segment into sub-segments and apply a rating to each sub-segment. Similar to other types of models, once the rule-based model is published, the model ratings can be used elsewhere in Rev.Up to build audiences to be included in campaigns. 

What is a good use case for a rule-based model? 

Rule-based models are useful for multiple reasons. Some examples of when you would want to use a rule-based model:

1. When you have an idea for how you want to segment your audience in a consistent way and do not want to duplicate the segmentation logic again and again. This is most useful when you are experimenting with engagement models. 
   • Use case example 1: 
     • Bombora intent topics in Rev.Up are given a High, Medium or Low value. A customer wanted to consistently use these values in many segments without creating the logic over and over again in each segment. In order to achieve this, they built a rule-based model with the logic of: 
       • Accounts with an intent topic that is High are given a score of A 
       • Accounts with an intent topic that is Medium are given a score of B 
       • Accounts with an intent topic that is Low are given a score of C
     • Result: The model evaluates the criteria every day, and accounts that have a changing intent will be given a new rating. This allows for the ratings to be used in multiple segments and always have the most up to date audience membership. 
   • Use case example 2: 
     • An example of a rule-based model being used to score accounts based on engagement with your company is using opportunity in the model. One customer built a model that segmented accounts based on the number and stage of opportunities. The criteria of the model were: 
       • Accounts with no opportunities or an opportunity in a late stage were rated as A. This allowed the customer to pay special attention to prospects with no opportunities. 
       • Accounts with an opportunity in a later stage were given a rating of B.
       • Accounts with an opportunity in an early or middle stage were given a rating of C. 
   • Use case example 3: 
     • Another example of a rule-based model being used to score accounts based on engagement with your company is using web visit data and intent data in the model. One customer built a model that includes visits to specific product web pages, and intent for specific topics. This can be used to create rule-based models that are product-specific. 
       • Accounts with intent or visits to the web page that relate to a specific product or solution were given a rating of A
       • All other accounts were given a rating of F 
     • Result: The model evaluates the criteria every day, and accounts that have a changing intent will be given a new rating. This allows for the ratings to be used in multiple segments and always have the most up to date audience membership. 
2. When you do not have enough data to be able to build a AI machine learning model. 

• • Sometimes there is not enough data to build an AI machine learning model. This happens when the product or service you are trying to sell is new and there are not enough success events to be able to build a model. Then defining your own rules in a rule-based model is a suitable alternative. 

Considerations for using rule-based models

• The output of a rule-based model is a rating from A to F. 
• Rule-based models can only be used to score accounts, not contacts.
• An account will always be given the highest rating that it qualifies for. Therefore, an account cannot have two or more ratings. 
• If the account does not qualify for any rules that are defined in the rule-based model, such account receives the default rating from the model. You can choose any rating A to F as the default. 
• Any attribute that can be used to build a segment can also be used to build a rule-based model. 
• Rule-based models are not predictive because they do not use machine learning to score and apply ratings to records. A rule-based model is a fixed collection of rules that you define. 
• Once a rule-based model is published, the model ratings can be used as an attribute in any other segment you build.