PCA analytics
PCA analytics answers the questions supervisors ask every week:
- How many calls did we analyze?
- Is average quality going up or down?
- What share is good vs poor?
- Which tags fire most often?
Numbers update as analyses complete (and are easy to read after a nightly intake run). This is trend-level insight — for a single call, use Analyzed records.
Where to open it
Section titled “Where to open it”- Inside a Call Analyzer project → Analytics, or
- Org Analytics with a Call Analyzer project selected
When the selected project is a Call Analyzer project, CXGear shows PCA metrics instead of outreach funnel charts.
src/assets/screenshots/53-pca-analytics.pngWhat you see
Section titled “What you see”Summary metrics
Section titled “Summary metrics”| Metric | Meaning |
|---|---|
| Calls analyzed | How many calls completed analysis in the period |
| Avg QA score | Average numeric score across those calls |
| Good / Fair / Poor | Counts in each grade band |
These cards are the “health check” at a glance.
Daily volume
Section titled “Daily volume”A day-by-day bar view of how many calls were analyzed (often with average score on the day). Use it to spot:
- Missing nights (schedule or runtime issues)
- Spikes after a campaign weekend
- Gradual volume growth as more recordings land
src/assets/screenshots/54-pca-analytics-daily.pngTop tags
Section titled “Top tags”Tags ranked by how often they matched. This is your coaching backlog in chart form:
- High “prohibited phrase” hits → urgent training
- Low “disclosure read” hits → script reinforcement
- Unexpected extract values → prompt tuning on GenAI tags
src/assets/screenshots/55-pca-analytics-tags.pngPeriods
Section titled “Periods”Choose a reporting window:
| Period | Window |
|---|---|
| Last 7 days | Week-over-week coaching |
| Last 30 days | Monthly quality review |
| Last 90 days | Quarterly trend |
Pick a period that matches your operating rhythm. Short windows are noisy when volume is low.
Why analytics is separate from Records
Section titled “Why analytics is separate from Records”| Analytics | Records | |
|---|---|---|
| Grain | Aggregates | One call |
| Job | Trends and mix | Coaching and audit |
| When | Weekly / monthly reviews | Daily desk work |
Analytics is built from daily rollups as calls finish analysis. If Records has calls but Analytics is empty, wait for rollup after successful analysis or confirm the period includes those days.
How data gets here
Section titled “How data gets here”- Intake spinner analyzes a call.
- Score, grade, and tag results are stored on the analysis.
- CXGear updates a daily rollup (call count, average score, grade counts, tag hits).
- Analytics reads those rollups for the selected period.
No separate “export to BI” step is required for the built-in views.
What success looks like
Section titled “What success looks like”- After nightly intake, morning analytics reflects yesterday’s volume.
- Grade mix matches what supervisors see sampling Records.
- Top tags align with known script issues.
- Empty state only appears for brand-new projects (with a clear “upload recordings / wait for overnight run” message).
Common problems
Section titled “Common problems”| Symptom | Likely cause | What to try |
|---|---|---|
| All zeros | No analyses in period | Run intake; widen period |
| Volume flatlined mid-week | Schedule or runtime stopped | Check runtime and spinner schedule |
| Avg score looks wrong | Scorecard or band change | Remember history is snapshot-based; new calls use new card |
| Top tags empty | No matching tags or tags not configured | Configure tags; confirm they match on Records |
| Outreach-style charts shown | Wrong project type selected | Select a Call Analyzer project |
| Permission denied | Role lacks analytics read | Assign Supervisor / Admin or grant permission |
Using analytics in the quality loop
Section titled “Using analytics in the quality loop”A simple weekly rhythm:
- Analytics — grade mix and top tags for the week.
- Records — open a sample of poor-grade calls.
- Tags / QA — adjust keywords, prompts, or questions if the model is wrong.
- Coaching — train agents on real patterns, not anecdotes alone.
- Re-check analytics the following week.
CXGear’s Post Call Analyzer is meant to scale listening — analytics is how you know the system is changing outcomes, not only producing scores.