When a week produces no genuine quality problem for an agent, the correct move is to say so and use the time differently, not to find something. Manufactured feedback is the most reliable way to lose an agent’s belief in the entire quality programme, because they know their own work and they can tell when a development point was reverse-engineered from a calendar invite. Before accepting that there is nothing to coach, rule out the opposite explanation: that the reviewer stopped looking properly, which produces identical symptoms.
In short
- The mechanism is structural. A fixed weekly cadence meets a variable supply of real problems, so the lead has to manufacture the difference.
- Agents can tell. Invented feedback teaches them the programme is theatre, and that conclusion does not reverse when you later have something real.
- Nothing to coach and the reviewer stopped looking have identical symptoms and opposite fixes. Check score distribution and reviewer agreement first.
- A genuinely clean week is a finding worth recording, not an empty slot to fill.
- Five legitimate uses of the slot: reinforce what worked, escalate a process problem, let the agent pick, prepare for something upcoming, or give the time back.
- Detach the cadence of review from the cadence of coaching. Review can be continuous; coaching should be triggered.
Why does a quiet week turn into invented feedback?
Because the cadence is fixed and the supply of real problems is not. Somebody committed to weekly one-to-ones with a development point in each. Some weeks an agent hands you three genuine things. Some weeks they hand you none. The commitment does not flex, so the lead closes the gap the only way available, which is by finding something.
What gets found in those weeks is recognisable to anyone who has been on the receiving end. A slightly different phrasing that would have been marginally warmer. A macro that was technically correct but not the one the reviewer would have picked. An observation about a conversation that went perfectly well.
The agent knows. This is the part that gets underestimated. They were there, they remember the conversation, and they can tell the difference between feedback that came from their work and feedback that came from a calendar. The pattern reported by support agents discussing this is consistent and blunt: when reviewers could not find a real fault they invented one, and the agent responded by disregarding the reviewer entirely from that point on.
Note the cost. It is not that one piece of feedback was weak. It is that the agent has now reclassified the whole programme as theatre, and every future piece of feedback, including the accurate ones, arrives inside that classification. You cannot spend your way out of it afterwards, which is why a programme agents trust is easier to protect than to rebuild.
One invented development point costs more than a skipped session. The skipped session says this week was fine. The invented point says the process will always find something, so the process is not about your work.
Is there really nothing to coach, or did the reviewer stop looking?
These two situations look identical from the lead’s chair and need opposite responses, so rule out the second before you accept the first. Four checks, in order of speed.
- Look at the score distribution. If almost everything is passing across the whole team, the scorecard has stopped discriminating rather than the team having become uniformly excellent. A criterion passing more than about 95% of the time for two quarters running is no longer carrying information, which is a signal to revisit the criteria rather than to celebrate.
- Check reviewer agreement. Have two reviewers score the same five conversations blind. If one finds issues the other did not, the quiet week may be a reviewer effect, and a calibration session is the fix.
- Check what was reviewed. At a typical 3% sample, an agent’s week might be represented by two conversations, and if selection favoured short or simple ones, a clean result says almost nothing. This is the most common cause by a wide margin.
- Look outside the scorecard. Repeat contacts, escalations and reopens on that agent’s tickets. Behaviour catches what a small sample misses, and a clean scorecard sitting next to three reopens is not a clean week.
If all four come back clean, that is a real finding and it should be recorded as one. A quiet week is data. Written down over a quarter it is the evidence for reducing this agent’s review frequency, which frees reviewer time for people who need it and is a more meaningful signal to the agent than any compliment.
A clean week at a 3% sample is not evidence of a clean week. It is evidence that the two conversations reviewed were fine. Say which of those two things you mean when you tell the agent, because overclaiming here is how a programme loses credibility in the other direction.
What are the five legitimate uses of the slot?
The lead needs somewhere to go that is not invention. All five of these are honest, and all five are better than a manufactured development point.
| Use the slot for | What it sounds like | When this is the right choice |
|---|---|---|
| Reinforce something specific that worked | Name the conversation and the exact move, not general praise | Almost always available, and the evidence says it outperforms correction |
| Escalate a process problem the agent hit | Ask what got in their way, then own taking it somewhere | When the agent is fine but the tooling, policy or knowledge base is not |
| Let the agent choose the conversation | Ask them to bring one they were unsure about | Strongest option for experienced agents. Their uncertainty is better targeted than your sample |
| Prepare for something coming | A product change, a peak period, a new channel | When the useful work is forward-looking rather than retrospective |
| Give the time back, and say why | This week was clean, here is the time | When the other four are genuinely not there. Costs nothing and buys credibility |
Why is reinforcing what worked not just a consolation prize?
It gets treated as the soft option, and the evidence does not support that reading. The Harvard Business Review argument in The Feedback Fallacy is that corrective feedback is far weaker than management practice assumes, that people learn least in the moment they are being told what is wrong, and that directing attention to what someone did well is the more reliable route to improved performance.
That does not mean stop correcting. It means the week with no correction available is not a wasted week, and the reinforcement option is not a fallback. It is arguably the primary tool that a problem-hunting cadence has been crowding out.
The requirement is specificity. Good work is fine as a sentence and worthless as feedback, because it names nothing the agent can repeat deliberately. Reinforcement has to carry the same evidence a criticism would:
- The conversation. Which one, so they can go and look at it.
- The moment. The specific exchange, not the whole ticket.
- The mechanism. What that move achieved, so it becomes a repeatable technique rather than a compliment.
That is the same standard as a legitimate coaching point, and it is why a score that traces back to the conversation matters as much for reinforcement as for correction. A lead who cannot find the moment cannot reinforce it specifically, and generic praise reads as filler in exactly the way an invented criticism does.
How do you stop the problem recurring?
Separate the cadence of review from the cadence of coaching. This is the structural fix and everything else is a workaround. Conversations can be reviewed continuously. Coaching conversations should be triggered by something found, not by a date.
Three changes make that work in practice:
- Keep the one-to-one, drop the requirement that it contain a development point. The regular contact is valuable on its own. The obligation to produce a finding is what causes the damage, and it is a separate thing that can be removed without touching the meeting.
- Define a threshold for a coaching conversation. Write down what qualifies: a repeated pattern, a single serious failure, a criterion missed twice in a fortnight. Below the threshold, nothing happens. This protects the lead from having to justify an empty week, which is what makes it a structural fix rather than a request for restraint.
- Report coaching volume as a distribution, not a rate. A report saying every agent was coached every week is a report of a process being followed, and it creates exactly the pressure that produces invention. What is worth reporting is where coaching went and what changed after it, which is the argument in turning QA data into coaching.
The measurement worth watching is where reviewer and lead time actually goes. Kaizo’s published customer result here is a 75% decrease in time spent coaching at SteelSeries, and the mechanism behind a number like that is not less coaching, it is less time spent looking for something to coach. Kaizo’s Auto QA scores the full population against your own rubric, so a lead arrives at a one-to-one already knowing whether this week contains a real pattern, which is what makes an honest empty week possible: the lead can say nothing came up and mean it, because the answer did not depend on which two conversations a reviewer happened to sample. The wider version of that argument is in coaching from QA without micromanaging.
What should you never do in a quiet week?
Four specific things, each of which is common and each of which does more damage than the empty slot it was avoiding.
- Do not lower the bar to generate a finding. Marking down a conversation you would have passed last month is not coaching, it is quota filling, and it corrupts the data as well as the relationship. It also means the agent’s score for that week is not comparable with any other week.
- Do not coach the score. Telling an agent they are at 91% and should aim for 95% is not a coaching point, because it names no behaviour. The score is a summary of things that happened; the things that happened are what can be discussed.
- Do not raise something outside their control. Handle time inflated by a slow internal system, a policy they cannot bend, a handoff from another team. Being scored for things outside your control is the single most reliable source of resentment in a quality programme, and a quiet week is when leads reach for it. If it comes up, it is a process problem to escalate, not a development point to record.
- Do not save up small things for a quiet week. A note kept for three weeks then delivered in a batch reads as a file being maintained on the agent. Anything worth saying was worth saying at the time, and anything not worth saying then is not worth saying now.
The general principle underneath all four: the quality programme should be able to report that nothing happened. A system that structurally cannot produce that answer will produce a false one instead, and the team will work out which it is well before leadership does.
Frequently asked questions
What do you do if an agent has no QA issues to coach?
Say so, record it as a finding, and use the time for one of four things: reinforcing a specific move that worked, escalating a process problem the agent hit, letting them bring a conversation they were unsure about, or preparing for something upcoming. If none of those apply, give the time back and explain why. What you must not do is find something, because manufactured feedback costs more than a skipped session.
Is it bad to skip a coaching session?
Skipping the development point is fine and often correct. Skipping the contact is a separate decision and usually worse, since the regular conversation has value independent of whether it contains a finding. The better structural answer is to keep the one-to-one and remove the requirement that it produce a coaching point, which is what causes leads to invent one.
How do you know if there is genuinely nothing to coach?
Check four things before accepting it. Whether the score distribution shows the scorecard still discriminates, whether two reviewers agree when scoring blind, how many of that agent’s conversations were actually reviewed, and whether behaviour outside the scorecard is clean, meaning reopens, repeat contacts and escalations. At a 3% sample, a clean week often just means the two conversations reviewed were fine.
How often should you coach support agents?
Detach coaching frequency from review frequency. Review can be continuous, but a coaching conversation should be triggered by something found rather than by a date, with a written threshold for what qualifies: a repeated pattern, a single serious failure, or a criterion missed twice in a fortnight. A fixed weekly requirement guarantees that some sessions have nothing real in them.
Why do agents resent QA feedback?
The most reliable causes are being scored for things outside their control and receiving feedback they can tell was manufactured to fill a slot. Both teach the agent that the process is not about their work, and that conclusion carries over to accurate feedback afterwards. Being able to say that a week was clean, and mean it, is one of the cheapest ways to keep the programme credible.
Related terms
Walk into the one-to-one already knowing if there is something real
Bring a month of conversations and the scorecard you use today. We will show you where the genuine patterns are per agent, and which weeks honestly had nothing in them.
