Every piece of feedback that flows into Enterpret is automatically classified by ML models into predictions — full taxonomy paths consisting of a keyword (L1 → L2 → L3) paired with a theme and sub-theme.

Prediction Editing gives you direct control to fix these classifications at the record level when the model output doesn't fully reflect the customer signal. Corrections take effect immediately for downstream use in reporting, Quantify, Dashboards, Wisdom, and exports — and are retained to inform future model improvements.

Note: Prediction Editing is record-level only. It does not modify your taxonomy structure. To rename, merge, or archive themes and keywords across your taxonomy, use Taxonomy Management.

Note: Edits do not yet retrain the model in real time. Corrections are saved and will feed future model updates as part of Enterpret's alignment strategy.

Prediction Editing is available inline wherever feedback records are surfaced in Enterpret, including:

When the model has missed a taxonomy path that should apply to a record, you can add it using one of two entry points:

Start by searching for the keyword, then Enterpret filters the valid theme paths for that keyword.

Start by navigating to the theme, then Enterpret shows only the keywords that belong to that path.

Note: Both entry points add the same result — a complete taxonomy path (L1 keyword → L2 keyword → L3 keyword + Theme + Sub-theme). Use whichever entry point is more intuitive for how you think about the feedback.

When the model has applied a taxonomy path that doesn't accurately describe a record, you can remove it.

The prediction is removed immediately and the record is updated across Quantify, Dashboards, Wisdom, and exports.

Only Admins and Editors can add or remove predictions on feedback records. Members have read-only access.

You can check and manage user roles in Settings → Users. Hover over the Understand user roles tooltip for a breakdown of what each role can do.