Managing QA Parameters
You can adjust your QA parameters at any time. Changes take effect for new evaluations immediately. Existing evaluations are not retroactively updated.
9.1 What You Can Change
| Change | Where | Notes |
|---|---|---|
| Edit parameter instructions | Configuration > QA Parameters | Updates pass/fail and N/A instructions; for Boolean parameters only |
| Change rule type | Configuration > QA Parameters | Boolean (pass/fail), Categorical (closed option set), or Descriptive (free-text answer up to a word limit). Unset parameters are treated as Boolean for backward compatibility. Type-specific fields appear in the edit form depending on the selected type -- see 9.4 Rule Types |
| Change weight | Configuration > QA Parameters | Must be > 0 |
| Change criticality | Configuration > QA Parameters | Critical = auto-fail on failure |
| Change mechanism | Configuration > QA Parameters | Transcript Only, Knowledge Verified, or Action Based |
| Add parameters | Configuration > QA Parameters > Add, or via Auto-Learning | Maximum 50 per request. Adding via Auto-Learning learns only the new parameters -- see 2.4 Configure QA Parameters with Auto-Learning |
| Delete parameters | Configuration > QA Parameters | Soft delete -- parameter is hidden but data is preserved. Select multiple (or all) via checkboxes to bulk delete, including down to an empty set |
| Add/delete contact reasons | Workflow Builder | Changes the categorization structure |
| Edit contact reason instructions | Configuration page | Updates behavior without changing structure |
| Configure Manual Review | Configuration > Manual Review | Set confidence level per rule type and the flag-on-insufficient-data toggle -- see 9.5 Manual Review Configuration |
9.2 When to Re-Run Auto-Learning
Consider re-running when:
- You have made significant changes to multiple parameters
- Substantial feedback has been accumulated and applied through Continuous Learning
- You want updated accuracy metrics measured against your golden data
- Your evaluation standards have meaningfully shifted
Re-running is not required for every small change. Individual edits and Continuous Learning improvements take effect immediately.
9.3 Training Material Generation
Personalized training documents for individual agents are generated as Coaching Reports, available from the Insights tab rather than from QA Parameters -- see 7. Using the Insights Dashboard for the full walkthrough.
9.4 Rule Types
Each QA parameter has a rule type that determines what kind of answer the evaluator produces for that question. The type is set when creating or editing a parameter and controls which fields appear in the edit form.
| Type | Answer format | Type-specific configuration |
|---|---|---|
| Boolean | Pass, Fail, or N/A | Pass/fail instruction and N/A instruction fields. This is the default behavior -- parameters with no type set are treated as Boolean. |
| Categorical | One option from a closed set | An option set (CategoryOption) where each option has a name, description, and examples; plus a default option used when the evaluator cannot choose from the set. |
| Descriptive | Free-text answer up to a word limit | Maximum word count (max_words) and answer guidelines (answer_guidelines) fields. |
Parameters created before rule types were introduced have no type set. They are evaluated as Boolean and continue to work without any changes.
The parameter edit form shows only the fields relevant to the selected type. Switching types resets the type-specific configuration fields.
9.5 Manual Review Configuration
Manual Review lets you hold low-confidence rule evaluations for a human to check before they affect the QA Score. You configure it in Configuration > Manual Review. Only managers and admins can change these settings.
9.5.1 Confidence Levels
You set a confidence level independently for each rule type. The level controls the threshold below which a result is flagged for review.
| Level | Behavior |
|---|---|
| Off | No results flagged for this rule type, regardless of confidence. |
| Low | Flags results where the evaluator's confidence is below 0.50. |
| Balanced | Flags results where confidence is below 0.70. |
| High | Flags results where confidence is below 0.85. |
The default policy when no configuration has been saved is: Balanced for Boolean rules, Low for Categorical rules, and Low for Descriptive rules. Manual Review is off until you save a policy.
Setting every rule type to Off disables the Save button. At least one rule type must be set to Low, Balanced, or High before you can save.
9.5.2 Flag on Insufficient Data
When the Flag on insufficient data toggle is on, results are also flagged when the evaluator did not have enough transcript content to make a confident determination, regardless of the confidence level setting. This toggle is off by default.
9.5.3 Projected Review Load
The configuration panel shows a projected review load line that updates as you change the levels and toggle. The projection replays the current on-screen policy against the persona's completed interactions from the last 7 days, so you can see roughly how many items would enter the review queue before you save.
If the persona has fewer than 10 recent interactions with confidence data, the projected load is not shown. Score more interactions first to get a useful projection.
9.5.4 Saving the Policy
The Manual Review policy saves as a whole: clicking Save writes all rule-type levels and the toggle together in one action. If you have made changes that have not been saved yet, the panel shows an unsaved-changes indicator.
To go back to the recommended defaults (Balanced for Boolean, Low for Categorical, Low for Descriptive, flag-on-insufficient-data off), click Reset to recommended. This restores the default values in the form but does not save until you click Save.