Step 3: Defining your model objective

How to define your model objective

Overview

Setting an objective allows Auxia to train a machine learning model on your data. There are a few core components of an objective:

• Objective name: The name of the objective

• Objective description: An objective's description

• Objective category: This is the type of goal you want to optimize for (e.g. conversion, retention, etc)

• Cohort event: This is the anchor event that defines the cohort in which a user belongs to. This event allows you to cohort any conversion metric to build a cohorted report.

• Conversion event(s): This is the outcome event that you want your users to take. This is also the outcome label that is used to train the model.

• Conversion event weight: Each conversion event can have a weight configured against it. These weights enable the Auxia models to optimize towards multiple high value actions concurrently.

Auxia works with customers during the onboarding process to define the most effective and appropriate modeling framework for their use case based on the diversity of user behavior, volume of events, and number of high value actions to optimize towards.

Set up

In the “Manage” section of the console, navigate to the “Objectives” tab. From there, you can add a new objective or edit an existing objective that you have previously set.

• Step 1 - Add an Objective name and description

• Step 2 - Select your Objective category: This is the type of goal you want to optimize for

• Step 3 - Define your Conversion event(s): Select the event that you want the model to optimize towards. This can be a single event, like purchase_completed, or a combination of events if you want the model to drive multiple high value actions concurrently.

After you have added your conversion event(s), click save and you will see the new objective on “Objectives” tab. Please note that if you edit the objective and toggle “Active” on, this will begin the model training process. You can create as many objectives as you would like, but only one objective can be enabled “Active” at a time.

Model experimentation

Auxia's platform has the ability to set up and test multiple different modeling approaches concurrently, splitting traffic to your team's needs across each of the models. If your team would like to take advantage of this capability, please contact us at [e-mail].