> ## Documentation Index
> Fetch the complete documentation index at: https://docs.charle.agency/llms.txt
> Use this file to discover all available pages before exploring further.
# QA Scoring
> Understanding QA scores and how they improve scheduling accuracy
QA Scoring tracks how many QA cycles a client's work typically requires before passing. This helps the system suggest more accurate time allocations for QA-related subtasks.
## What is QA Scoring?
QA Scoring provides data-driven insights into each client's QA performance:
* **First-Pass Rate**: Percentage of tasks that pass QA on the first review
* **Average Cycles**: How many QA review cycles a client typically needs
* **QA Score**: A multiplier used to adjust QA time estimates
These metrics help PMs schedule realistic time for QA activities.
## Understanding QA Scores
### Score Values
A QA score of **1.0** means tasks typically pass on the first QA review:
| Score | First-Pass Rate | Meaning |
| ------- | --------------- | -------------------------------- |
| **1.0** | \~100% | Work passes first time (ideal) |
| **1.5** | \~60% | Often needs one round of fixes |
| **2.0** | \~30% | Usually needs multiple QA cycles |
### How It's Calculated
The QA score is based on:
1. **First-Pass Rate**: What percentage of tasks pass QA on the first review
2. **Average Cycles**: The mean number of review cycles needed before passing
3. **Sample Size**: More completed QA reviews = higher confidence
Only tasks with completed QA reviews contribute to the score. In-progress tasks don't affect calculations until QA is complete.
### Confidence Levels
Like deliverability scores, QA scores have confidence levels:
| Confidence | Reviews | Reliability |
| ---------- | ------------ | ------------------------------- |
| **Low** | Fewer than 5 | Limited data - use with caution |
| **Medium** | 5-15 | Reasonable confidence |
| **High** | 15+ | Reliable data for scheduling |
## How QA Scores Affect Scheduling
### T-Shirt Size Suggestions
When Alan suggests a t-shirt size for a development task, it considers the client's QA score:
* High QA scores (>1.5) may trigger a recommendation to bump up the size
* The reasoning explains if QA overhead influenced the suggestion
### Subtask Time Estimates
When a quote is approved and converted to a task:
1. Subtasks are created with pre-calculated AI suggestions
2. QA review subtasks get adjusted estimates based on QA score
3. Higher QA scores = more time suggested for QA activities
The schedule dialog shows "Alan suggests: Xh Ym" with an explanation when the QA score influenced the estimate.
### Example
For a client with QA score of 1.8:
* Standard QA allocation: 30 minutes
* Adjusted suggestion: \~54 minutes (30 x 1.8, rounded to 15 min)
* Reasoning: "Adjusted for high QA cycles (1.8x avg)"
## Viewing QA Scores
QA scores are displayed in the Client Intelligence section of each client's detail page.
### What You'll See
* **QA Score**: The multiplier (e.g., 1.5)
* **First-Pass Rate**: Percentage (e.g., 60%)
* **Average Cycles**: Mean review cycles (e.g., 1.5)
* **Confidence**: Based on sample size
* **Sample Size**: Number of completed reviews
### Interpreting the Data
Score near 1.0, first-pass rate above 80%
**What it means:**
* This client's work typically passes QA on first review
* Requirements are usually clear and well-understood
* Developer work is consistently high quality
**Action:**
* Standard QA time allocations are appropriate
* Consider this client's positive pattern
Score 1.3-1.7, first-pass rate 50-70%
**What it means:**
* This client often needs one round of QA fixes
* Could indicate evolving requirements or complex work
**Action:**
* Allow slightly more time for QA activities
* Review if requirement clarity can be improved
Score above 1.8, first-pass rate below 50%
**What it means:**
* This client's work typically requires multiple QA cycles
* May indicate unclear requirements, scope creep, or complex work
**Action:**
* Schedule significantly more time for QA
* Consider t-shirt size bumps for development work
* Review root causes - are requirements clear enough?
## QA Event Tracking
### What's Recorded
Each QA subtask completion records:
* **Task and Requirement**: What was reviewed
* **Cycle Number**: Which review round (1st, 2nd, etc.)
* **Pass/Fail**: Whether QA passed or failed
* **Developer**: Who did the development work
* **Reviewer**: Who performed the QA review
* **BugHerd Points**: Number of issues found (if applicable)
### Automatic Score Updates
After each QA completion:
1. The QA event is recorded
2. Client's QA score is recalculated
3. Score history is updated (if change > 0.1)
## Resetting QA Scores
Only admins can reset QA scores. This should be done rarely and with good reason.
### When to Reset
Consider resetting when:
* The client's development processes have fundamentally changed
* New team members have significantly improved quality
* Historical data is no longer representative
### How to Reset
Use the API endpoint or admin interface to reset:
* Score returns to 1.0 (neutral)
* Sample size resets to 0
* Reset is logged in score history
## Impact on AI Suggestions
QA scores influence AI in two ways:
### 1. T-Shirt Size Suggestions
When generating size suggestions for quotes:
* QA metrics are included in the AI prompt
* High QA scores may trigger recommendations to size up
* Reasoning explains if QA overhead was a factor
### 2. Subtask Time Suggestions
When tasks are created from approved quotes:
* Each subtask gets a pre-calculated AI suggestion
* QA subtasks are adjusted by the QA score
* Suggestions appear in the schedule dialog
## Related Documentation
Overview of all client scoring systems
Understanding subtask types and scheduling
How time estimates work
QA phase in the task lifecycle