Analytics | Predictive Evaluations (Legacy) dashboard

The Analytics | Predictive Evaluations (Legacy) dashboard compares AI-generated (predictive) evaluation scores against manual human evaluation scores. It enables quality managers to assess model accuracy, understand score distributions for both methods, and identify contacts where predictive and manual scores diverge significantly — which may represent calibration opportunities or contacts worth reviewing in detail.

This dashboard is designed for Quality Managers who want to understand the alignment between AI-generated quality scores and human evaluations, and for Operations Managers who want to monitor quality coverage across a larger contact population than manual evaluation alone provides.

NOTE Predictive Evaluations are AI-generated quality scores applied automatically to transcribed contacts. They extend quality coverage to a higher percentage of contacts than manual evaluation alone. Manual scores are completed by human evaluators in the standard QM workflow.

Questions this dashboard helps answer:

How does the average predictive evaluation score compare to the average manual score?
How are manual and predictive scores distributed?
How are both score types trending over time?
Which agent groups have the highest and lowest predictive scores?
Which specific contacts have predictive and manual scores available for comparison?

Prerequisites

You have the Insights Reader license.
You must have access to Insights dashboards with permission to view Analytics data.
You must have access to the Contacts and Evaluations datasets.
Predictive Evaluations require transcription and the Predictive Evaluation model to be enabled and configured.

Page location

Insights > Dashboards > Analytics | Predictive Evaluations (Legacy)

Dashboard layout

The dashboard contains the following sheets:

Sheet	Description
Predictive Evaluations	The primary analysis sheet. Contains KPIs and gauge charts for average manual and predictive scores, distribution histograms, trend lines for both scores, and group-level comparison bar charts.
Manual vs Predictive Drill Detail	A contact-level table showing individual contacts with both manual and predictive scores for direct comparison. Includes contact URL links.

Controls and filters

NOTE Filters are hierarchical. Selecting a Group limits the available Teams and Agents.

From / To (date pickers) — both sheets

Sets the date range for all metrics.
Group(s) / Team(s) / Agent(s) (drop-downs) — both sheets

Standard hierarchical scope filters.
Form(s) (drop-down) — both sheets

Filters to a specific evaluation form. Apply when comparing predictive and manual scores for a single form type.
Evaluation State (drop-down) — both sheets

Filters to a specific evaluation state (e.g. Scored, Pending). Use to focus on completed evaluations only.
Language — both sheets

Filters to a specific language.

Procedures

Assess predictive model alignment

Set the date range and apply the Form filter to a single evaluation form.
Compare the Avg Manual Score and Avg Predictive Score gauges.
Compare the two distribution histograms for shape similarity.
Review the trend line chart to assess whether the two measures track together over time.
Review the by-Group bar chart to identify groups with the largest divergence.

Use this to:

Validate the predictive model against manual scoring for each evaluation form.
Identify whether the model needs recalibration for specific groups or contact types.

Review specific contacts with divergent scores

Navigate to the Manual vs Predictive Drill Detail sheet.
Review the contact list to identify contacts where manual and predictive scores differ significantly.
Click Contact URL links to open the contact record and review the manual evaluation alongside the predictive score.

Key metrics

Metric	Definition	How to use it
Sheet: Predictive Evaluations
Avg Manual Score (gauge)	The average manual (human) evaluation score for the selected filters, displayed as a gauge against the form’s total points maximum.	The human quality benchmark. Use as the reference point when assessing whether the predictive model is scoring in a similar range.
Avg Predictive Score (gauge)	The average AI-generated predictive evaluation score, displayed as a gauge against the predictive quality score maximum.	Compare directly to Avg Manual Score. Significant divergence may indicate model drift or that the predictive model is weighting different factors than human evaluators.
Total Contacts / Contacts Evaluated	KPIs showing total contacts in scope and the count with manual evaluations.	Assess manual evaluation coverage. The gap between total contacts and contacts evaluated represents the population covered only by predictive scoring.
Manual Scores Distribution (histogram)	Distribution of manual evaluation total scores across the selected population.	Understand the shape of human scoring. A heavily skewed distribution may indicate evaluator leniency or harshness.
Distribution of Predictive Score (histogram)	Distribution of predictive evaluation scores.	Compare to the manual score distribution. Similar shapes indicate model alignment; divergent shapes warrant investigation.
Score Trending charts	Line charts showing Avg Manual Score and Avg Predictive Score trended over time (Contact Start Time).	Track whether both scores move in the same direction over time. Diverging trends may indicate the model needs recalibration.
Avg Scores by Group (bar charts)	Manual and predictive average scores displayed side by side by agent group.	Identify groups where predictive and manual scores diverge most significantly. These are priority candidates for manual evaluation review.

Data notes

Data is sourced from the Contacts and Evaluations datasets.
Avg Predictive Score, Avg Manual Eval Score, PES (Predictive Evaluation Score), and related fields are calculated fields.
Contacts with neither a manual evaluation nor a predictive score will not appear in the scoring metrics.