Skip to content
Sign in Get started
Documentation

9bx user guide

Team performance

Quality (QA)

Score calls and chats with scorecards, by people and by AI, then coach, calibrate and report.

Contents

What it is for

Quality shows what "good" looks like (scorecards), scores real interactions against it, lets agents acknowledge or dispute their scores, and reports where coaching is needed.

Who can use it

Needs the PBX & Call Center module.

Permission Lets you
Manage Scorecards, sampling rules, settings, resolve disputes, reports and export.
Evaluate Pick interactions, score them, work the review queue, review AI scores.
Calibrate Calibration sessions and their variance.
View own An agent's own evaluations: acknowledge or dispute.

Where to find it

PBX & Call Center → Team performance → Quality checks: Review queue, Evaluations, Scorecards, Score calibration, QA reports. My evaluations is under Home → My work. Settings: PBX & Call Center settings → QA settings.

The pages explained

Scorecards

The questions, their weights and the pass mark. Status Draft, Active, Archived. Buttons: New scorecard, Duplicate, Activate, Archive.

Editor field Meaning
Name, Channel (Call or Chat), Description Basics.
Pass mark A score at or above this passes.
Applies to Queues, chat teams or teams (departments). Empty means everywhere. The most specific card wins: queue, then team, then the default.
Sections and criteria Each section has a Weight. Each criterion has an Answer type (Yes / no or Scale with Lowest and Highest), Help text, Guidance for the AI, Can be N/A, and Critical (auto-fail): a failing answer fails the whole evaluation.
AI auto-QA Score with AI, Share to score (below 100% scores a steady sample), Shortest call, Fewest messages, Flag scores below, Show AI scores to the agent, Extra instructions for the AI.

When a scorecard already has evaluations, changing questions, weights or pass mark saves a new version; finished evaluations keep their version.

Evaluations

Every scored call and chat. Tabs: Evaluations, Find a call or chat, My drafts. Filters: channel, source (Human or AI), status, result, flag (Low score, Critical fail, Compliance miss, Unhappy customer), agent, Only disputed. Export CSV.

Find a call or chat: choose channel, scorecard, agent, direction, dates, Longer than (seconds), and Not scored by a person yet, then Search or Random pick. Only answered calls with a recording and chats an agent resolved are listed. Press Evaluate to open the workspace.

The evaluation workspace

The call or chat transcript with the recording on one side, the scorecard on the other. Answer each criterion (Yes, No, N/A) with notes. Buttons: Save draft, Submit, Discard draft, comments with timestamps, Coaching notes and Mark coaching done / Reopen. For AI-scored items: Confirm AI score, Override answers, Save review. For disputes: Adjust answers and Resolve dispute (Uphold keeps the score, Adjust recalculates).

My evaluations (agents)

Your Average score, Pass rate, evaluations, items Waiting for you and a weekly trend. Open one to read it, then Acknowledge or Dispute (say what you disagree with; a manager reviews the call; you can dispute within the number of days set in QA settings).

Review queue

The calls and chats picked for you to score this week by sampling rules. Start to score, Skip to remove an item, Show everyone's for managers.

Calibration

Several evaluators score the same call or chat blind so you can see how far apart they are. New session: Name, Call or chat ID (from Call history or the chat inbox), at least two Evaluators (not the agent of the interaction), Due. Score it to take part; Close session shows the Mean score, Range, Spread, each evaluator against the mean or a Reference, and the Criteria where evaluators disagree most.

QA reports

Tabs: Overview (average, pass rate, evaluations, AI against human, critical fails, open disputes, a trend, and breakdown by agent, team and queue), Failure heatmap (pick one scorecard; share of answers that failed each criterion), AI vs human (how far the AI can be trusted: answers left unchanged, mean score difference, independent comparisons) and Coaching (open coaching items by age).

QA settings

Quality assurance is on (off stops AI scoring and sampling for the whole organisation), Dispute window in days, AI evaluations a month / a day caps, and Sampling rules: Add rule to pick random calls or chats per agent per week, optionally for chosen agents, with Run now. AI scoring uses your own AI keys first; the platform engine is used only while your plan includes AI and minutes remain.

Tasks

Create and activate a call scorecard

  1. Open Scorecards and press New scorecard.
  2. Name it "Support calls", channel Call, pass mark 80.
  3. Add a section "Opening" (weight 20) with the criterion "Greeted with company name" (Yes/no).
  4. Add a section "Resolution" (weight 50) with "Solved the issue" (Yes/no) and mark "Verified the customer's identity" as Critical.
  5. Save and press Activate.

Score a call

  1. Open Review queue, press Start on an item.
  2. Listen, answer each criterion, add an overall comment and coaching notes.
  3. Press Submit.

Dispute a score (agent)

  1. Open My evaluations, open the evaluation.
  2. Press Dispute, explain, submit.

Set up weekly sampling

  1. Open QA settings → Add rule, choose 3 per agent per week.
  2. Press Run now to test; items appear in each evaluator's Review queue.

Example scenario

Padma Net Support sets a "Support calls" scorecard, pass mark 80, with AI scoring on for 30% of calls. Every week 3 calls per agent land in the review queue. Imran scores 72 on one call because he skipped identity verification (a critical fail). He reads the coaching notes, acknowledges them, and Nusrat marks coaching done after a short chat.

Tips

  • Keep scorecards short (8 to 12 questions).
  • Run a calibration session each month so all evaluators score alike.
  • Review a sample of AI scores in AI vs human before trusting AI-only scoring.
  • Use Critical sparingly, only for compliance items.

Troubleshooting

Problem Fix
"Create and activate a scorecard first." Find a call needs an active scorecard.
Find a call shows nothing Only answered, recorded calls and resolved chats are listed.
"This evaluation was not found" It may have been discarded, or you lack access.
AI does not score AI is off in QA settings or on the scorecard, caps were reached, or no transcript exists.
Cannot dispute The dispute window has passed.
"There is no recording you may play for this call." You lack permission for that recording.