Scoring tickets is easy. Making the score change behaviour is the hard part.
Quality assurance in support usually fails in one of two ways. Either nobody reviews the work and quality is inferred from complaint volume, or a detailed scorecard exists but nothing changes as a result. Both approaches detect problems too late.
Good QA is a feedback system. It samples real work, applies a consistent standard, finds patterns, coaches one behaviour at a time and checks whether that behaviour improved. The score is evidence, not the purpose.
Start with the customer promise
A scorecard should reflect what customers and the business actually need. A hotel guest asking about a booking, for example, cares that the reservation is understood, the answer is accurate, the next step is clear and any change is recorded correctly. Perfect grammar cannot compensate for the wrong booking date.
- Accuracy: was the information and action correct?
- Resolution: was the actual question answered rather than acknowledged?
- Ownership: did the agent explain what happens next and who will do it?
- Tone: did the response fit the customer's situation?
- Process: were records, tags and required checks completed correctly?
- Risk: was sensitive, financial or urgent work escalated when required?
Use critical failures carefully
Some errors should outweigh the rest of the score. Sharing information with the wrong person, making an unauthorised financial promise or changing the wrong booking cannot be averaged away by good spelling. Define a short list of critical failures and explain exactly why each one matters.
Keep that list short. If half the scorecard counts as critical, reviewers stop distinguishing serious risk from ordinary coaching opportunities.
Choose a sample you can defend
A small random sample reviewed properly beats a large sample skimmed. Select conversations before looking at outcomes, otherwise reviewers gravitate towards known complaints and miss the ordinary work that shapes most customer experiences.
Add targeted samples for new starters, changed processes and known risks. Random review tells you how the operation is performing. Targeted review tells you whether a particular intervention worked. Keep the two views separate in reporting.
Calibrate the reviewers
If two reviewers score the same conversation differently, the trend is unreliable. In a calibration session, reviewers grade the same cases independently, compare differences and refine the guidance. The useful part is not agreeing on a number. It is exposing ambiguous standards before they reach the team.
Beacon uses calibration across the Sri Lanka and Madagascar teams so the same behaviour is judged against the same expectation. When a client changes a policy or tone guideline, examples are added to the scorecard rather than left in a meeting note.
Turn findings into coaching
An agent cannot improve from a monthly percentage alone. Coaching should point to a specific moment, explain its impact, show a stronger alternative and agree what to practise next. One clear behaviour for the next week is more useful than a list of ten weaknesses.
- Show the exact sentence or action being discussed.
- Ask the agent what they noticed before giving the answer.
- Connect the issue to the customer or operational impact.
- Practise a better response using a similar scenario.
- Check a fresh sample within days, not at the next quarterly review.
A QA score that never leads to a conversation is just a number with a decimal point.
Separate individual coaching from operational learning
Repeated mistakes are not always an agent problem. If several capable people misunderstand the same rule, the procedure may be unclear. If the answer is hard to find, the knowledge base may be the problem. If a queue encourages rushed replies, staffing may be the problem.
Report what changed
A useful client report shows the quality trend, the main error themes, the actions taken and whether those actions worked. Include examples where helpful, but protect customer information. The final question should always be: what will be different in the operation next week because we reviewed this work?
At Beacon, quality review sits inside day-to-day management. Sampling, feedback, coaching and cross-team calibration are part of running the service, not an audit added after a problem has already reached the customer.
