Stellar

Call QA

Automatic call scoring + replay regression testing

Every completed call is graded 0–100 across five dimensions. Pro and Scale customers can also replay scripted test suites against the agent's current config to catch regressions before they hit a real caller.

Recent calls
Sarah Chen
2m 48s
Booked demo
Marcus Webb
3m 21s
Qualified
Priya Shah
1m 12s
Not interested
David Kim
0m 54s
Callback

Every call, every dimension, every day

When a call ends, Stellar runs structured scoring on the transcript. You get an overall score, five dimension sub-scores, flagged issues, and positive highlights — within seconds of the call hanging up. No manual listening required.

Goal completion

Did the agent accomplish the template's goal — qualify, confirm, or book?

Script adherence

Did it stay on your configured questions and handoff paths?

Persona consistency

Did the tone, style, and voice stay in character the whole call?

Objection handling

Did the agent recover when the caller pushed back?

Technical quality

No dead air, no inappropriate interruptions, clean turn-taking?

See the worst calls first

Open any agent and jump straight to the bottom 10% of scored calls. The Flagged Calls list surfaces exactly the transcripts worth listening to — not the 500 good ones you'd skip anyway.

Pro & Scale

Replay regression testing

Voice agents drift. A helpful prompt change can quietly break the objection-handling you spent weeks tuning. Replay testing catches that before real callers do.

01

Define a suite

Write a handful of canned scenarios — the caller's opening line plus a one-sentence pass criterion (e.g. "books the furnace tune-up after confirming zip code").

02

Run before shipping

Change the persona, script, or KB — then replay the suite against the new config. The run completes in about one minute per case.

03

Diff vs. baseline

The most recent completed run is the baseline. New runs show exactly which cases regressed and which newly started passing.

04

Ship with confidence

When the suite stays green, the change is safe to roll out. When it doesn't, the transcript + judge reasoning tell you why.

Stop guessing whether your agent works.

Scoring is on for every plan. Replay testing unlocks with Pro.