Reserved predictive model column names
This reference covers the reserved column names for the predictive models available as segmentation criteria in Blueprints and journeys in the UI. For the full model catalog, see Machine learning models overview. For configuration steps, see Using predictive models in campaigns and journeys.
This reference lists the reserved column names, data types, and value ranges for the seven predictive models exposed as segmentation criteria in Blueprints and journeys.
Setup dependency​
Predictive models must be enabled in Admin > System Config > System Defaults before these reserved columns are recognized as mapped segmentation criteria in the UI.
Each score column must exist in your own data warehouse tables and be surfaced in the audience query or the Journey User Profile under the exact reserved column name listed above. A column holding equivalent data under any other name is not recognized by the system.
Column name matching is case-sensitive. Data warehouses vary in how they treat identifier casing, for example some fold unquoted identifiers to uppercase or lowercase by default. Confirm the column resolves to the exact casing shown above before relying on it in a Blueprint or journey.
Reserved columns​
| Model | Reserved column name | Type | Values | Description |
|---|---|---|---|---|
| Propensity to Churn | MLChurnScore | INT | 1-100 | Normalized propensity-to-churn score. Higher values indicate greater likelihood of churn. |
| Purchase propensity | MLPurchasePropensityScore | INT | 1-100 | Normalized likelihood that a customer makes a purchase within the model's prediction window. |
| Second purchase propensity | MLNextPurchaseScore | INT | 1-100 | Normalized likelihood that a customer makes a second, or next incremental, purchase. |
| Customer lifetime value (LTV) | MLLTVScore | INT | 1-100 | Ranked scale of predicted customer value, where 100 represents the most valuable customers. |
| Engagement index | MLEngagedScore | INT | 1-100 | Normalized score representing a customer's predicted level of engagement with the brand. |
| Optimal channel | MLOptChan | VARCHAR | email, push, sms | The channel a customer is most likely to engage with. |
| Optimal hour | MLOptHour | INT | 0-23 | The hour of day a customer is most likely to engage. Represents hour of day only; does not include day of week. |
All seven columns are keyed on the recipient identifier used in the audience query or journey user profile.
Optimal channel in channel selection​
In addition to its use as a segmentation criterion, MLOptChan is also usable as a channel-activation option in multi-channel campaigns and journey tiles, when two or more channels are selected. Both usages reference the same underlying reserved column.