Create a model

Beta This is a beta feature, which means we’re still developing it. Some functionality might change.

Models are machine learning algorithms that uncover patterns in your data and deliver insights to help app admins understand key factors behind specific business results. After you install the Data Analyzer app, you can create models to predict specific outcomes. For example, you may want to create a model to predict:

• Which projects are more likely to be over budget
• What predicts low satisfaction ratings
• What customers that churn have in common

Data requirements

You can build models to create predictions for up to 10 unique values in your realm.

The table you analyze must have:

• A minimum of 20 records
• A maximum of 2 million records
• At least 5 fields (excluding the field you want to evaluate)
• Up to 1GB of data

Create a model

Before you begin, identify a specific outcome you'd like to predict in your data. This outcome should be a field and value pair you want to generate predictions for and will serve as the foundation of your model.

To create the model:

1. On the Data Analyzer template app, click the New model button.
2. Click Start the wizard.
3. Fill out the wizard with information on:
   • The app that contains your table—Select the app you’d like to analyze.
   • The specific table to analyze—Select a table. Data analyzer scans all records in the selected table.
   • The field to evaluate—Select the field of the outcome you want to predict.
   • The value to predict—Select the value within the field that you want to generate insights for. The value must be:
     • An exact value, meaning the field type is Text, Multiple Choice, Checkbox, FormulaText, or FormulaCheckbox
     • Case-sensitive
     • Present in 2% to 80% of the records
   • Data source based on—Optionally select a specific report to analyze. Data Analyzer only analyzes data within this report. Otherwise, all data in the selected table is analyzed.
   • Date field used for time period—Optionally select a date field to influence how Data Analyzer organizes time-based insights. For example, some factors may become more or less significant over time. Date created is used by default.
   • Model name—Enter a unique name to identify the model.
4. Select Analyze and wait for your model to be created. It may take up to 15 minutes.

Read Understand model results for information on how to interpret the results, or learn how to Deploy a model in a table.