Most sales forecasting relies on probability fields updated manually by reps — a figure that reflects confidence rather than evidence. Zia Field Prediction replaces that guesswork with machine learning: a configurable prediction model that analyses your historical CRM data and forecasts the likely outcome of deals, purchases, or any other measurable business metric.

The key distinction from off-the-shelf AI tools is that Zia Field Prediction is trained on your own data. It learns from the patterns in your CRM records — what characterised deals that closed, what distinguished customers who churned, what correlated with a high deal value — and applies those patterns to predict future outcomes with a quantified confidence score.

What You Can Predict

Zia Field Prediction is flexible by design. Common use cases include:

• Win or loss probability for deals in your pipeline
• Likelihood that a lead will convert within a given timeframe
• Probability that a customer will make a repeat purchase
• Expected deal value for new opportunities
• Any quantifiable outcome tied to fields in your standard or custom modules

Supported Field Types for Prediction

Zia can predict values for the following field types: Date and Date/Time (predict when an event will occur, e.g. expected close date), Number, Decimal, and Currency (forecast numeric outcomes such as deal value or order size), Percentage (predict probability-based fields such as win likelihood), Boolean/Checkbox (binary outcome predictions — will this deal close? will this customer churn?), and Single Picklist (predict categorical outcomes such as deal stage or lead quality classification).

Availability

How to Configure a Prediction Rule

1. Navigate to Setup › Zia › Prediction. Click “New Prediction” to open the prediction builder.
2. Name your prediction and choose the source module. The module determines which records Zia will analyse — for example, Deals for a win/loss prediction.
3. Select the field you want Zia to predict. This is the outcome field — the value Zia will forecast for future records.
4. Configure the learning data scope. Choose whether Zia trains on all records in the module or only records matching specific filter criteria. Filtering is useful if your historical data includes record types that are not representative of your current business model.
5. For picklist fields, optionally define negative values. Specifying which values represent an undesirable outcome improves model accuracy for binary-style predictions.
6. Click Save and confirm the custom field creation. Zia creates a new field in the module to display prediction results and confidence scores. This field is auto-populated as the model runs.

Permissions

Two separate profile permissions control access to Zia Field Prediction. Administrators configure these under Setup › Security Control › Profiles › Zia › Prediction:

• Manage Configuration: Allows creating, editing, enabling, disabling, and deleting prediction rules. Typically assigned to CRM administrators or operations leads.
• View Results: Allows viewing prediction outcomes on records without the ability to modify rules. Enabled by default for all profiles.

What the Output Looks Like

Once a prediction model is active, each qualifying record displays the predicted value alongside a confidence score from 0 to 100. For binary predictions (win or loss, churn or retain), the confidence score indicates how strongly the model supports its prediction. Zia also surfaces the key contributing factors — the fields that most strongly influenced the prediction — giving reps and managers a basis for understanding why a particular outcome is forecast and what might be done to improve it.