Call QA

Call center quality control

Score 100% of calls against a checklist instead of spot-checking. The model finds violations, computes scores and sends alerts.

100% of calls instead of spot-checking
A 0–100 score with quotes and timestamps
Webhook alerts on violations

Get an API key Open documentation

83.3 Call evaluation · rec_8f3c

Checklist

Greeting and introduction Passed
Clarifying the request Passed
Forbidden phrases Violation

Evidence

“Hello, my name is Anna, company…” · 00:04

“…look, just cut it out already” · 12:47

Checklist scoring

Required and forbidden steps, per-item weights and a final 0–100 score.

Evidence and timestamps

Every verdict comes with a quote and a timestamp, so supervisors review in seconds.

Analytics and alerts

Day-by-day trends, agent weak spots and webhook alerts on violations.

How it works

Three steps to your first evaluation

Upload the recording

The call lands in your recording library — via the API, Telegram bot or dashboard.

Define a checklist

Required script steps and forbidden phrases — each item has its own weight.

Get a score and alerts

Verdicts with quotes and timestamps, day-by-day trends and webhook alerts.

API

One request — call evaluation

Evaluate a recording against a checklist and get a per-item verdict with evidence and a weighted score.

curl -X POST https://ttsapi.ru/v1/qa/evaluate \
  -H "X-Api-Key: rtt_…" \
  -H "Content-Type: application/json" \
  -d '{
    "recording_id": "rec_…",
    "checklist": [
      { "id": "greeting", "kind": "required", "weight": 2 },
      { "id": "profanity", "kind": "forbidden" }
    ]
  }'

from voicekit import VoiceKitClient

client = VoiceKitClient(api_key="rtt_…")

result = client.qa_evaluate(
    "rec_…",
    [
        {"id": "greeting", "kind": "required", "weight": 2},
        {"id": "profanity", "kind": "forbidden"},
    ],
)
print(result["score"])  # 83.3

import { VoiceKitClient } from "voicekit-client";

const client = new VoiceKitClient({ apiKey: "rtt_…" });

const result = await client.qaEvaluate("rec_…", [
  { id: "greeting", kind: "required", weight: 2 },
  { id: "profanity", kind: "forbidden" },
]);
console.log(result.score, result.passed, result.failed);

using VoiceKit;

var client = new VoiceKitClient("rtt_…");

var qa = await client.QaEvaluateAsync(
    "rec_…",
    new object[]
    {
        new { id = "greeting", kind = "required", weight = 2 },
        new { id = "profanity", kind = "forbidden" },
    });
Console.WriteLine(qa?["score"]);

{
  "recording_id": "rec_…",
  "score": 83.3,
  "total_checks": 3,
  "passed": 2,
  "failed": 1,
  "violation": true,
  "items": [
    { "id": "greeting", "verdict": "passed",
      "reason": "Оператор поздоровался и представился",
      "quote": "Здравствуйте, меня зовут Анна…", "start": 0.0, "end": 2.1, "weight": 2.0 },
    { "id": "profanity", "verdict": "failed",
      "reason": "Найдена запрещённая формулировка",
      "quote": "…", "start": 12.4, "end": 13.0, "weight": 1.0 }
  ]
}

Manual review vs automated QA

Manually

5–10% of calls are checked
Hours listening to each conversation
Scoring depends on the individual supervisor
Violations are found after the fact

With Voice kit API

100% of calls against one checklist
Seconds per call, not hours
A single 0–100 score with evidence
A webhook alert right after a violation

FAQ

Frequently asked questions

Which plans include call QA?

QA is available on Pro and Business: evaluation is built from a library recording and requires a language model.

What does the checklist check?

Required script steps and forbidden phrases. Each item has its own weight; the result is a 0–100 score with evidence.

Can I evaluate calls without the recording library?

No, evaluation is built from a library recording — upload audio via the API, bot or dashboard.

How fast is the result?

Usually a few seconds, depending on recording length and queue. The result can be delivered via webhook.

Start controlling call quality

Available on the Pro and Business plans — evaluate QA on your own calls.

Get an API key Pricing