Skip to content

QA Scorecard(QA)

Also called: Quality assurance scorecard · Conversation review rubric · QA rubric

A rubric reviewers use to grade a sample of support conversations on accuracy, tone, compliance and process, giving the team's view of quality.

Reviewed

What is a QA scorecard?

A QA scorecard (quality assurance scorecard) is the rubric a support team uses to review a sample of conversations and grade them against agreed standards. Each criterion — correct answer, verified identity, tone, policy compliance, correct use of tools, clear next step — has a definition, a scoring scale and a weight. The total gives a quality score per conversation, per agent and per team.

For WhatsApp support, a good scorecard includes channel-specific items: did the agent reply inside the 24-hour service window or use an approved template correctly, did they avoid asking for sensitive data in chat, did they tag the conversation, and did they close with a clear next step. Some items are marked critical, so a single failure — for example, asking a customer to type their OTP or card number — fails the whole review regardless of the other scores.

The scorecard is not a replacement for CSAT. CSAT asks the customer how satisfied they were; the scorecard asks whether the agent did the job correctly. A friendly agent who promised an unauthorised refund may get high CSAT and a failing QA score, while a correct but curt answer can be the reverse. Teams use both, plus operational metrics such as average handle time, to see the full picture.

Key takeaways

  • A QA scorecard is a fixed rubric reviewers apply to a sample of conversations to grade accuracy, tone, compliance and process.
  • Mark a few critical items as auto-fail — such as asking for an OTP or sharing someone else's data — so they cannot be averaged away.
  • Score a random sample, not only complaints, and calibrate reviewers regularly so the same chat gets the same score.
  • CSAT is the customer's view of one interaction; the QA scorecard is the team's view against its own standards.

See InfiQ pricing

Interactive stop

Score one WhatsApp conversation

8/8

ticked

Required
3 / 3
Optional
5 / 5

Review passes the scorecard

Every item ticked.

A sample eight-item scorecard for a WhatsApp support chat. Tick each item the agent met; required items are the auto-fail checks that must pass for the review to count.

Review passes the scorecard. 8 of 8 ticked.

How does a QA scorecard work?

  1. Define 6–10 criteria

    Group them into accuracy, compliance, communication and process. Write each as an observable behaviour ("Confirmed the order number before sharing status"), not a feeling ("Was helpful").

  2. Weight them and pick auto-fails

    Give accuracy and compliance the most weight. Mark two or three items as critical so a failure there sets the score to zero, such as requesting OTPs or passwords, or sharing another customer's details.

  3. Sample conversations fairly

    Review a fixed number per agent per week, chosen at random across channels, queues and shifts, and add targeted samples (low CSAT, reopened, escalated) as a separate stream.

  4. Calibrate reviewers

    Have reviewers score the same conversations independently each month and discuss differences until their scores agree within a few points.

  5. Coach and track

    Share the scored chat with the agent, with quotes from the transcript, and track trends per criterion so training targets the item that is actually failing.

A QA scorecard: key facts and limits

Typical criteria
6–10 behaviours, weighted
Critical items
2–3 auto-fail behaviours
Sample
Random per agent, plus targeted reviews
Calibration
Same chats scored by all reviewers, monthly
Customer view
Captured separately by CSAT

Common mistakes with a QA scorecard

  • Reviewing only complaints

    If reviewers only see low-CSAT chats, scores look worse than reality and good habits are never reinforced. Keep a random sample as the main stream.

  • Vague criteria

    "Showed empathy" means different things to different reviewers. Define each item with examples of pass and fail so two reviewers agree.

  • Letting critical failures average out

    A chat that leaked personal data should not score 85% because the greeting was good. Use auto-fail items for compliance-critical behaviours.

  • Scoring without coaching

    Scores that agents never discuss change nothing. Pair every review with a short feedback note and a specific example from the transcript.

Why does a QA scorecard matter?

Metrics like FRT and resolution time say how fast a team works, not whether it works correctly. A QA scorecard catches wrong answers, compliance risks and habits that hurt customers before they become complaints or regulatory problems. For regulated sectors such as lending, insurance and healthcare in India, documented conversation reviews are also evidence that staff follow scripts and data-handling rules. And because every agent is scored against the same rubric, coaching becomes specific rather than opinion.

Example: A Chennai lending app tightens compliance

A Chennai consumer-loan app reviews five random WhatsApp chats per agent each week on an eight-item scorecard. Two items are critical: never ask for an OTP or full bank details in chat, and never confirm loan approval before the system shows it. In the first month, 4% of reviewed chats auto-fail, almost all on premature approval messages from one shift. A targeted coaching session and a canned response for "under review" bring that to under 1% by the next month, while CSAT stays flat.

— A Chennai lending app tightens compliance

Best practices for a QA scorecard

  • Include internal notes in the review — they show whether decisions were documented.
  • Score bot-handled conversations too; the bot is part of the customer's experience.
  • Let agents self-score a few chats; the gap between self and reviewer scores is a coaching topic.
  • Revisit the rubric when policies, templates or products change.

QA FAQs

What should a customer support QA scorecard include?+
Six to ten observable behaviours covering accuracy, compliance, communication and process, each with a weight, plus two or three auto-fail items for critical compliance failures.
How many conversations should QA review?+
Enough for a fair picture per agent — commonly a handful of random conversations per agent per week, plus targeted reviews of escalations and low-CSAT chats as a separate stream.
What is the difference between a QA score and CSAT?+
CSAT is the customer's rating of one interaction; a QA score is a reviewer's grade against internal standards. They can disagree, which is why teams track both.
What is QA calibration?+
It is a session where reviewers score the same conversations independently and then compare. Repeated regularly, it keeps scores consistent across reviewers and over time.

Talk to InfiQ

From definition to your first message

InfiQ is an official Meta Business Partner. Start a 7-day free trial — no card needed — and see this running on your own number.

Start free trialBrowse all 550 terms