A QA score is not coaching. Handing an agent a number and a call recording is the beginning of the work, and most programmes stop there — which is why scores improve slowly and agents treat reviews as something done to them.
The difference between coaching that lands and coaching that gets nodded through is usually specificity.
Why most QA feedback fails
Three recurring reasons, all fixable:
- It is anecdotal. One call is arguable. "You were abrupt on this call" invites a defence about that particular customer, and the defence is often fair.
- It is late. Feedback three weeks after the call is about a conversation nobody remembers.
- It is not actionable. "Improve your listening" is not an instruction. There is nothing to do differently tomorrow.
Specific beats emphatic
| Weak | Strong |
|---|---|
| You need to work on your tone | You interrupted the customer on 34 of 210 calls; the team average is 9 |
| Improve your call openings | You skipped self-identification on 40% of calls this month, up from 12% |
| Be more thorough | You offered further assistance before closing on 55% of calls; target is 90% |
| Your scores are down | Your probing score fell 14 points in June; every other category held |
The right column is only available with coverage. At a five-call sample you cannot say "34 of 210" — you can only say "on this call", which is the weakest form of the argument.
Coach the pattern, not the call
A single bad call is usually noise. Everyone has a bad Tuesday. What deserves a conversation is a pattern — a parameter consistently missed, a category trending down, a gap against the team on one specific behaviour.
Then use one call as the illustration, after establishing the pattern. That order matters: pattern first makes the call an example, and call first makes the pattern look like justification.
Distinguish the three causes
Before coaching, work out which of these it is, because they need different responses:
- The agent does not know. A training gap. Show them.
- The agent knows but cannot. A process, tooling or time-pressure problem. Coaching them harder will not help and will damage trust.
- The agent knows and chooses not to. A conduct conversation, not a QA one.
Most QA programmes treat all three as the first. If a parameter is failed across the whole floor, it is almost never the first — look at parameter pass rates across the team before booking individual sessions.
Make it routine, not an event
Coaching that happens when someone remembers does not change behaviour. What works is a rhythm: a short monthly report per agent, delivered the same way every month, with the same structure — where you are, what moved, what to focus on next.
Xperia generates that as a per-agent monthly report with strengths, focus areas, month-over-month movement and a written coaching plan, emailed automatically. The point is not the document; it is that the conversation happens every month without anyone scheduling it.
Let agents see their own data
Agents who can see their own scores between reviews correct themselves without a session. Those who only see scores when a supervisor produces them experience QA as surveillance.
Self-scoped access — an agent sees their own calls and nobody else's — is the cheapest trust-building change available in most programmes.
Coaching works when it is specific, patterned, timely and routine. All four depend on having enough coverage to describe a pattern rather than an anecdote.
And separate the three causes before the conversation. Coaching someone for a problem the process created is the fastest way to lose the floor.