Analyzed records & scores

After intake runs, each analyzed call appears on the project Records page. This is the supervisor’s day-to-day list: who scored well, who needs coaching, and what the model heard.

Records here are results, not the full import spreadsheet. Source rows always live under Data tables; Records shows outcomes after analysis.

What you see on Records

Typical columns and fields for a Call Analyzer project:

Field	Meaning
Score	Numeric QA score for the call
Grade	Band such as good / fair / poor (from your Setup thresholds)
Tags	Flags and extracted values from the tag master
Summary	Short supervisor-facing recap
Metadata	Contact / call context you marked as metadata fields

Open a row for detail: full tag list, QA breakdown, and a View of the frozen scorecard snapshot.

Screenshot needed

PCA Records list showing score, grade, and summary columns.

Where: Call Analyzer project → Records

Save as: src/assets/screenshots/51-pca-records.png

Why snapshots matter

When a call is scored, CXGear stores a snapshot of QA results on that analysis. If you edit the scorecard next week, old records keep their original scoring.

That keeps audits honest: you can explain why a call got 72 last month even if the scorecard changed.

How to review a call

Open the Call Analyzer project → Records.
Scan grades (good / fair / poor badges) for coaching priorities.
Click a row to open detail.
Read the summary for a quick narrative.
Inspect tags for script and compliance flags.
Open QA View for the question-by-question snapshot.
If something looks wrong, open Pipeline journey on the job record to see which step failed or skipped.

Screenshot needed

Record detail panel with grade, tags, summary, and QA view action.

Where: Call Analyzer project → Records → open a row

Save as: src/assets/screenshots/52-pca-record-detail.png

Scores and grades

On Setup you define:

Max score for the scorecard
Good and poor limits (bands)

The analyzer turns the scorecard total into a grade. Grades are what most supervisors filter on; scores are for finer ranking.

Grade	Typical meaning
Good	At or above your good threshold
Fair	Between poor and good
Poor	At or below your poor threshold

Tune bands after you have a few dozen real calls — do not overfit on three samples.

Tags on the record

Tags appear as the outcomes of tag types (metadata, BoW, GenAI, derived). Use them to:

Filter “all calls missing disclosure”
Validate that GenAI extract values look sane
Feed coaching conversations with concrete evidence

If tags are empty, intake may not have reached apply-tags, or tag config is inactive.

Summaries

The summary is a short recap for humans who will not read the full transcript. It is generated with your LLM provider. Treat it as an assistant, not a legal transcript — always open the transcript or audio when a decision is high stakes.

What success looks like

New analyses show up after each intake job without manual export.
Grades match your coaching intuition on a sample of listened calls.
QA View shows the same questions that existed when the call was scored.
Poor-grade calls have enough tag and summary detail to start a coaching conversation in under a minute.

Common problems

Symptom	Likely cause	What to try
Records empty	Intake not run or failed	Jobs page; run intake; check runtime
Score but no summary	Summarize step failed	Journey errors; LLM provider
Grade missing	Thresholds or score null	Setup max score and bands; confirm QA questions
Tags missing	Tag master empty or step failed	Configure tags; re-run sample
“Wrong” historical score	Scorecard changed later	Expected — snapshots do not rewrite history
Looking for full CSV	Wrong page	Use Data tables for source rows

Records vs Analytics

Page	Best for
Records	One call at a time, coaching, audit
Analytics	Trends: volume, average score, grade mix, top tags

Use Records in the morning huddle for individuals; use Analytics for weekly quality reviews.