Decision Tree Builder

What is a Decision Tree Builder?

LendAPI’s Decision Tree Builder is a powerful visual authoring tool that brings variables from the Variable Library together with Actions and Outcomes. These Actions and Outcomes lead to workflow orchestration processes and end-customer communications. Decision trees are triggered applications designed by our Product Builder as well as a suite of API calls.

In this section, we will describe the process of building a decision tree using LendAPI’s Decision Tree Builder. We will attempt to run through an example of how to build a decision tree and all of the necessary elements that go into creating a decision tree.

Most importantly, we will talk about the consequences and the outcomes of a successful decision tree process. We hope that this section will tie together some of the most critical concepts in LendAPI for you to build complex business rules yourself.

What do I need to start building decision trees in the LendAPI Decision Tree Builder?

There are several major ingredients that you need to consider before building a decision tree in LendAPI’s decision tree builder.

1. Your third party data providers - If your bank wants to launch a new bank account opening process, there are at least a small handful of third party integration partners you need to use to complete the origination process. First, you will need an identity bureau to verify your user’s identity. There are several commercially viable vendors, some of them are in LendAPI’s Marketplace. If you want to offer a credit product alongside the DDA (demand deposit account) opening process, you might consider using a credit bureau to check into your applicant’s credit worthiness. If you are a commercial lender, you might want to consider working with LexisNexis or CLEAR to verify the identity of the business owners as part of the application process. If you are a point of sale finance lender, there might be a day where you need to verify your merchants as well as the applicants they send to you through their merchant portal. With the ever popular bank account verification as well as bank transaction data providers, you might want to line them up before laying down credit decision rules.

2. Actions and Outcomes - There are two concepts in LendAPI’s Decision Tree Builder that we want you to get familiar with. Action such as Approved, Declined, Review or Request Documentation are the result of each branch of the decision tree. For example, if you write a rule that states “If Credit Score is greater and equal to 600 then Approve”, that “Approve” decision from that single rule is called an action. Conversely to finish the aforementioned rule, we need to have a decline node as well. The Decline node is also an action. After an Action is assigned, we will connect this action to an outcome. For example, if the decision from the three mentioned before is an Approved action, then the approved action could be mapped to an Outcome. An Outcome could be a Bank Account offer or a credit card offer. On the other hand, the Declined action will need an outcome to be mapped. We can map a Credit Declined outcome to it and that outcome will generate an email with a declination letter which you can program as an email template.

3. Outcomes and Offer Builder - If the decision node approves an application, it will trigger a mapped outcome which is mapped to a product offer. These offers are further defined in the Offer Table section of the LendAPI Workflow suite of tools. 

4. Templates and Documentation - Other important settings you need to worry about are email templates that can be triggered by an Action linked to an outcome. Offer documents need to be filled out with our Document Builder as well. Although you can run your rules without a fully flushed out email or document template, we suggest that you program in with what you have to be more illustrative during demo tests.

5. Product Builder - Once you have the rules all set and actions/outcomes linked. It is time to build your online application pages with LendAPI’s Product Builder. You can build each page and elements in each page through our graphical interface. Each page can then link to a decision tree. Our product builder can work on top of multiple decision paths and bring the customers to a separate experience based on how your decision tree’s branches.

These are some of the minimal set of elements you can set up before building your decision tree. Of course, the staff here at LendAPI are happy to train you to set up the fundamentals and help you to build your first decision tree.

Where do I find the Decision Tree Builder?

Head on over to the left navigation bar on LendAPI’s administrative portal. Then goto “Workflow” section of the platform and click on Decision Trees.

A. Decision Tree Main Panel: To locate the Decision Tree Builder is easy. Log into the LendAPI administrative portal. On the left hand side, there is a navigation menu. Click onto “Workflow” and then click onto “Decision Trees”.

B. Decision Tree List: All of the decision trees you’ve built so far are listed in the tree listing view. You can search for a particular tree as well with keywords. In this example, we searched for the keyword “underwriting” and all of the trees with the word underwriting appear in the decision tree listing view.

C. Add New Decision Tree: The “Add New” button will let you begin to build a new decision tree.

LendAPI Decision Tree Builder overview

LendAPI Decision Tree Builder utilizes a graphical user interface to display the decision trees as they are being built. We believe that this is the best way to display your work and provide the best visual feedback for the people working on these trees.

A. Rules Editor: This is where you can program your rule or branch. You can select any variable from the variable library through a third party integration or a variable you have customized yourself with LendAPI’s Variable Builder. The rules editor also has all the logical or mathematical operators made available through a dropdown. The third section of the rules editor is where you can enter a value which you want to compare against the variable. You can write compound rules with and/or buttons within the rules editor. Each branch could have its own mini-tree. Keep in mind that if you want a particular rule or branch to have a distinct outcome, then you should keep all the branches by themselves. The compounding rules will only have one action mapped to it.   

B. Decision Tree Visualizer: The entire tree and its branches as well as the branch outcome (Action/Outcome) are displayed visually for the rules engineers to detect any anomalies or logical issues.

C. Decision Tree Controls: Decision trees are complex and delicate. We’ve built multiple decision tree controls for you. First, we have a version concept where you can save multiple versions of the same tree. If your business evolves, you can build a new version of the tree and keep all prior versions of the decision tree for compliance purposes. We also have a Tree Tester built in. We will show more of the testing method later. You can also change the name of the tree as well as publish the tree into production so the next application that comes through your website can take advantage of the new version of the tree live.

LendAPI Decision Tree Builder - Decision Node Editor

The final node on each branch of the tree is as important as the branches themselves. The end node of the branch signifies that a decision has been made. An end node could be an approved decision or a declined decision. It could be a review decision or automates a particular action that the applicant must take. In this section, we explore the Node Editor and how these end nodes are linked to Actions and Outcomes.

A. Decision Node Editor - The decision node editor displays the node that’s in question. We highlight the node in question in the decision tree visualizer so it is clear which node you are editing. The node type can be an Action or a Rule. If the Node Type is an Action, then we can map an action, then outcome. If the node type is a rule, then the branch continues to grow. Each Decision Node can also chain to another Decision Tree. You can chain as many trees together through the tree branch to reduce the overall complexity of the entire decision tree.

LendAPI Decision Tree Builder - Approved Node (Action -> Outcome)

To further illustrate the decision tree path and how the branches resolve into an end decision node, we will take a look at a decision node which has an approved decision and how it maps to an outcome.

A. When a Decision Node is an Approved decision. You can map the Approved decision to an approved action. The approved action drives an outcome that will trigger a product offering. If you want to see how Actions and Outcomes are set up, please visit those actions within the launch guide. In this case, you can see that the approved action drives outcomes such as “Prime, Super Prime etc…”. These outcomes are all mapped to a financial product.

LendAPI Decision Tree Builder - Declined Node (Action -> Outcome)

On the flip side, there are many decision nodes that result in a decline decision. These decline decision nodes are also mapped to a decline action and decline types of outcomes. In this section, we flipped the decision node to Declined and you can see that the Outcome drop down automatically switches to all of the decline types of outcome. 

A. When a Decision Node is a Declined decision. You can map the decline type of outcomes here. These outcomes often drive a workflow arrangement or it will drive an automated email that gets sent to the application. And of course you can program and edit these declined emails in the email template section of our platform.

LendAPI Decision Tree Builder - Tester

One of the most popular features of our Decision Tree builder is our Decision Tree Tester feature. Once the tree has been created, before publishing these trees into production mode, the risk manager can test the tree’s logic with this testing application.

You can enter all of the acceptable and unacceptable values to drive different actions then outcomes. Here, we will illustrate both a positive and a negative testing result.

LendAPI Decision Tree Builder - Tester (Positive Result)

A. Here we entered various values that triggered each branch of the decision tree. The result of the decision in the Output section says “Action: approve”. This means that each of the tree branches rendered an approval path and therefore the entire tree ends with an “approve” decision.

LendAPI Decision Tree Builder - Tester (Negative Result)

A. We again enter values that we believe will render a declined decision. And indeed the rule tester tells us that the rule has rendered a declined action.

LendAPI Decision Tree Builder Summary:

We had a lot of fun building this decision tree builder for our banks and financial institutions. You won’t believe some of the complex rules they’ve built without our guidance. We hope you can sign up and start building your own trees. We are always here to help, please contact us or email us at info@lendapi.com