Deploy a model by adding it to a table

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

Once you've created a model in the Data Analyzer app, you can deploy it to a table to dynamically calculate a prediction score for each record in your table. 

Deploy a model

1. Open the model from the ML Models table in the Data Analyzer app
2. Navigate to the Model Deployment tab
3. Copy the prediction formula
4. Navigate to original table that contains the source data that was analyzed
5. Navigate to table settings and select +New to create a field
6. Select Formula - numeric as the type, then Add fields
7. Click into the field to edit the settings, and paste the Prediction Formula you copied
8. Save your changes

Now, you’ll see a prediction score appear for each record in your original table.

Change the prediction type

The prediction formula automatically displays the prediction score as a percent. You can change the prediction type to display the result as Yes or No instead. A Yes indicates that record is predicted to impact your selected outcome, while a No means it does not. To do so:

1. Open the model from the ML Models table
2. Select the pencil icon to edit the record
3. Navigate to the Model Deployment tab
4. Change the prediction type from Percentage to Yes/No Prediction and save the changes.
5. In the original table, follow the steps to Deploy a model, using a Formula - text field instead. The prediction formula will now display a Yes/No prediction rather than percent.

Optional: You can also edit the probability threshold to increase or decrease the prediction value needed to display Yes or No. The default value is 50%, meaning that if the prediction formula gives a prediction value for the record of 49%, it would result in No, and 51% would result in Yes.

Re-run your model

The data analyzer formula is based on the machine learning model you created at a static point in time. To ensure the outcomes predicted are as accurate as possible, we recommend creating a new model every 1-2 months to capture new records and fields that may be potential factors for the outcome.