Coaching becomes micromanaging when the cadence is fixed but the supply of real problems is not, because the lead then has to manufacture the difference. The line between them is not tone, it is selection and evidence: whether the week’s evidence chose the topic, and whether the agent can open the conversation it came from and read it. A programme that can record no coaching point this week is the one agents end up trusting.
In short
- Micromanagement here is a supply problem, not a personality problem. Fix the cadence and most of it disappears without anyone changing their style.
- The difference is selection and evidence, not tone. A gently delivered weekly development point that nobody needed is still micromanagement.
- Coaching the score is what agents resent. The score is a symptom, and a good share of the time the cause is not theirs to fix.
- Every coaching point should be traceable to a moment the agent can go and read. Feedback that cannot be pointed at is an opinion with a number attached.
- Most weeks the value sits with two or three people and one systemic issue. Spreading it evenly is how a programme turns into a ritual.
- Saying there is nothing to coach this week is the highest-trust move available to a lead, and it has to be a recordable outcome rather than an act of personal courage.
Why does a weekly coaching requirement turn into micromanagement?
This is an arithmetic problem, not a personality problem. Fix the cadence and the supply of real problems does not politely follow it. A lead with twelve agents who is expected to bring a development point to every one to one owes twelve points a week, whether or not the week produced twelve. Ask for three each and the debt is thirty six.
A settled team working from a stable scorecard does not generate thirty six coachable problems in a week. On a normal week it generates about four: one criterion failing across the team, one or two people with a repeated behaviour they can genuinely change, and a handful of things that turn out to be hard tickets handled reasonably.
The gap between what the calendar demands and what the week produced has to be filled with something, and everything in it is invented. The clearest version of this confession is manager-side rather than agent-side: leads describing the weekly trawl through conversations looking for something, anything, worth raising, because the slot exists and the field on the form is mandatory.
Agents work this out inside a month. Once a development point arrives every week regardless of what actually happened, the only consistent explanation left is that someone is looking, and the programme gets read as surveillance rather than support. That reading is not free. Harvard Business Review reports that monitoring employees makes them more likely to break rules, because close observation shifts responsibility for conduct away from the person being observed. A programme experienced as monitoring does not merely fail to improve behaviour, it can move it the wrong way. The broader version of that failure is in how to run a QA program agents do not hate.
So the first move is not a better coaching technique. It is to stop requiring a coaching point.
| The requirement | Points owed per week | What the evidence actually supported |
|---|---|---|
| One development point per agent, twelve agents | 12, every week, indefinitely | About 2 |
| Three points per agent, twelve agents | 36 | About 4, including the one that is not a person |
| A quiet week after a scorecard refresh | Still 12 | Frequently 0 |
| The week after a policy change nobody communicated | Still 12, all aimed at agents | 1 escalation, coachable to nobody |
What actually separates coaching from micromanaging?
Both conversations can be warm, well structured and delivered by someone who genuinely cares. Tone is not the variable, which is why every article telling leads to ask more questions and give fewer instructions leaves the problem exactly where it found it.
The variables are how the topic got chosen and what it is anchored to. Coaching selects: the week produced evidence of a specific, repeated, changeable behaviour, and that is what gets raised. Micromanaging selects nothing, because there was nothing to select from, so it reaches for whatever is visible. Coaching points at a moment. Micromanaging points at an impression, or at a number, which is the same thing wearing better clothes.
| Dimension | Coaching | Micromanaging |
|---|---|---|
| What starts the conversation | Something in the week’s evidence | A slot in the calendar |
| Who chose the topic | The evidence did | The person who needed something to say |
| What it points at | A specific moment the agent can reopen and read | A number, or a general impression, or a habit noticed once |
| What is being controlled | The outcome, and only where the agent owns it | The method, including the parts that do not matter |
| Frequency | Variable, because weeks differ | Fixed, because the calendar does not |
| What good looks like a month later | The agent needs you less | The agent checks with you more |
| How the agent reads it | Someone noticed something real | Someone was looking for something |
Why does coaching the score make agents resent QA?
Two of those rows carry the argument. If you cannot say what evidence chose the topic, and you cannot open the thing you are talking about, you are not coaching, however kindly you put it. The definitional groundwork is in what agent coaching is and the full method in the customer service coaching guide. Here is how it breaks specifically in a QA programme.
Coaching the score is the most common failure and the one agents name first. An agent comes back at 78 on resolution accuracy and the conversation opens with the number. Your accuracy is 78, let us get it to 85.
That number is a symptom with several possible causes behind it, and only some of them belong to the agent. They may be quoting a policy that changed without reaching the floor. The correct answer may be three clicks deep in a tool nobody can search mid-conversation. The criterion may mean one thing to the reviewer and another to everyone else. Or the ticket mix may simply be harder, which is what happens in every programme that marks agents down for things they cannot control.
Coaching the score asks an agent to move a number. Coaching the cause asks them to change one behaviour, and it can only be asked once you know which behaviour. The sorting question takes about ten seconds: what would have to be true for this to be the agent’s to fix? If the honest answer involves a policy, a tool, another team or a staffing decision, it is an escalation with a coaching point stapled to it. Telling those two apart reliably is its own skill, set out in agent error or process error.
There is a deeper reason score-led feedback underperforms. Buckingham and Goodall’s The Feedback Fallacy in Harvard Business Review argues that people are unreliable raters of other people, and that telling someone where they fall short is a poor mechanism for making them better at anything. That does not make the score useless. It makes it a pointer to a conversation rather than the content of one, which is what an internal quality score is built to be. A weekly quota turns it into the content instead.
What makes a coaching point evidence rather than an opinion?
Every coaching point should be traceable to a moment in a conversation the agent can go and read. That is the test, and it is unforgiving in a useful way.
Done by hand it costs about five minutes a point. Open the conversation the score came from. Find the exchange that produced the deduction. Copy the two or three lines around it. Bring that into the room, on screen, before you say what you think about it. If you cannot find the moment, you do not have a coaching point, you have an impression, and the honest thing is to drop it.
Three things change once this is the rule rather than the aspiration.
- Invented points die on contact. You cannot manufacture a development area if you have to open the evidence for it first. The rule enforces itself in a way no amount of coaching training does.
- Disagreement becomes cheap and specific. An agent who can see the exchange can tell you the customer had already been transferred twice, or that the account was flagged. That is not defensiveness, it is the fastest route to the real cause.
- The score stops being the argument. Once the conversation is about a moment, nobody has to defend a percentage.
This is also where automated scoring earns its place or fails to. A deduction nobody can explain is worse than no deduction, because the agent now holds a number with no moment behind it and no route to challenge it. What it looks like when this works is set out in QA score traceability, and the phrasing that makes a traced point land rather than sting is in these QA feedback examples.
Who actually needs coaching this week?
Most weeks the value sits with two or three people and one systemic issue. Not with everyone equally, and the instinct to spread it evenly is what converts a QA programme into a calendar obligation.
Even coverage looks like fairness. Twelve agents, twelve slots, nobody singled out. What it produces is twelve shallow conversations, most of them recapping numbers the agent has already seen, paid for out of the eight or so hours of real coaching capacity a lead has left once queues, escalations and their own meetings are covered.
Sort the week’s evidence into what it will actually support.
| What the week produced | Roughly how many | What it actually deserves |
|---|---|---|
| One criterion failing across most of the team | 1 finding | A rubric or process fix. Never twelve one to ones about the same thing |
| A repeated, changeable behaviour in one person | 1 or 2 people | A real coaching conversation, with the conversation open on screen |
| Something moved once, cause unclear | 2 or 3 people | Nothing this week. Write the name down and look again |
| A cause sitting with policy, tooling or staffing | Whatever the week threw up | An escalation to the owner, and tell the agents you raised it |
| Ordinary variation, and hard tickets handled reasonably | Everyone else | Nothing at all |
How do you stop triage drifting back to everyone?
Two conversations and one fix. That is what a normal week on a settled team earns, and a programme demanding twelve is not asking for more coaching, it is asking for ten pieces of fiction.
Triage drifts back to even coverage for one predictable reason: concentrating attention looks like favouritism unless you say out loud what you are doing. Three habits hold it in place.
- Tell the team the rule before you apply it. Coaching goes where the week’s evidence points, most weeks that means two or three people, and this week it is not you is a normal outcome rather than a verdict.
- Rotate what you reinforce. If the same two names come up every week and nobody else is ever mentioned, those two are not being coached, and the rest of the floor will say so.
- Treat the sorting as weekly, not permanent. Someone in look again four weeks running has stopped being noise. Someone coached on the same thing for two months is a training or role conversation instead.
The mechanics of the weekly pass, in the order that surfaces the systemic issue before the individuals, are in what a team lead actually does with QA scores each week, and the route from a chosen cause to a conversation that changes something is in how to turn QA data into coaching.
What do you do in the week when there is genuinely nothing to coach?
Some weeks produce nothing worth an agent’s time, and saying so out loud is the single highest-trust move a lead can make. It is also the move most programmes have made structurally impossible.
It carries that weight because it is the only unfakeable proof that the feedback is a signal rather than a schedule. An agent who has heard you say nothing came up this week knows that when you do raise something, it was chosen. Every future coaching point borrows credibility from the ones you did not invent, and it costs ninety seconds.
The failure is structural rather than personal. The template has a field labelled development area and it is mandatory. The lead’s own manager reviews completion. Leaving it blank looks like not doing the job, so somebody gets told their greetings could be warmer.
Fix the structure, not the lead. Three changes, all administrative.
- Make no coaching point this week a recordable outcome. A legitimate option in the template with a value of its own, not an empty field someone has to defend. If the form will not accept it, the form is manufacturing the feedback and no amount of training will stop it.
- Report on the review, not on the points raised. Measure whether the lead did the weekly pass. Counting coaching points sets a quota by accident, and quotas get met.
- Give the slot back, or spend it differently. Cancel it, or ask what is getting in their way and which ticket type they dread. There was a systemic issue on the team-level pass that nobody is coaching because it is not a person. Go and fix that.
One caution, because this fails in the other direction too. Nothing to coach is a conclusion you reach after doing the weekly review, never a reason to skip it, and agents can tell the difference immediately. When there is something to say, the structure that keeps the conversation from drifting into a numbers recap is in what a coaching framework is.
And when the honest answer is that the week contained nothing, there are five legitimate uses of the slot that are not coaching, set out in what to do in the week there is nothing to coach.
Where does the reviewer’s job end and the lead’s begin?
Micromanagement also arrives through the back door, when the lead quietly starts doing the reviewer’s job as well as their own. The symptom is unmistakable: the week goes on re-scoring and re-litigating conversations instead of doing anything with them.
From the agent’s seat this is worse than it sounds. Their work is now inspected twice, by two people, against a standard that is apparently negotiable, and the second inspection has no dispute route because officially it is not happening.
Split the two jobs by question rather than by job title.
- Was this scored correctly, and would a second reviewer agree? The reviewer’s call, settled in the review itself and in calibration, never in a one to one.
- Is the rubric measuring the right thing? Also the reviewer’s, with the lead as a loud input at the scorecard review.
- Why does this keep happening, and what changes next week? The lead’s, and the only one of the three that coaching can answer.
The working rule: dispute a score once, through the route, on the record, then act on the corrected picture. A lead who reopens scores inside a one to one teaches the agent that the number is negotiable, and spends the hour defending someone else’s work instead of doing their own.
What changes when scoring is automated?
Under sampled review the scarce resource was evidence: a few percent of conversations, a handful per agent per month, and half a day hunting for an example concrete enough to make a point stick. The standing complaint was that the sample missed the thing you already knew was happening.
Automated scoring inverts that, and the consequence is not the one most teams expect. The bottleneck moves from finding problems to choosing which ones to act on, which is a harder judgement and one nobody has been trained for. Leads who get full coverage often report feeling less in control for the first month rather than more.
It also makes the manufacturing problem worse before it makes it better, which is the part no vendor page prints. Under a 3% sample, a lead forced to invent a coaching point at least ran out of material. With every conversation scored there is always something to point at, so a weekly quota can be met indefinitely with feedback that is technically true and completely pointless. Full coverage removes the excuse for inventing coaching points and removes the friction that used to limit them. If the cadence is not fixed first, automation industrialises the exact behaviour this page is about.
Three things to change when the coverage arrives.
- Raise the bar for what counts as a pattern. With 100% coverage revealing trends that 3% sampling never could, one poor conversation stops being worth a meeting. Three instances of the same failure in a month is, and now you can tell the two apart.
- Make traceability a requirement, not a feature. If a deduction cannot be opened and read, it cannot be coached, and agents are right to push back. Some of them prefer machine scoring precisely because it is consistent and inspectable, which is the argument in why some agents prefer being scored by AI.
- Spend the reclaimed preparation time on selection, not volume. The hours saved are hunting hours. They should buy better decisions about who to coach, not more sessions to fill.
Kaizo’s AI coaching is built around that last point: patterns surface with the underlying conversations still attached, which is where EverHelp reported a 90% reduction in coaching prep time. Note what that does and does not remove. Preparation is the part that automates. Deciding which two people are worth the week, and having the conversation, are still yours.
Frequently asked questions
What is the difference between coaching and micromanaging?
Selection and evidence, not tone. Coaching raises something because the week produced evidence of a specific, repeated behaviour the person can change, and it points at a moment they can go and read. Micromanaging raises something because the calendar said a conversation was due, and it points at a number or a general impression. Both can be delivered warmly. A gently worded development point that nobody needed is still micromanagement, and agents grade it that way.
How do you know if QA coaching has become micromanagement?
Two tests. First, count how often a one to one produces no development point. If the answer is never, the cadence is generating the feedback rather than the work, because no real team fails at something new every single week. Second, pick last week’s coaching points at random and try to open the conversation each one came from. Anything you cannot point at was an opinion. A programme that fails both tests is being experienced as surveillance whatever it is called internally.
Should every agent get a coaching point every week?
No, and requiring it is the most reliable way to destroy a QA programme. A settled team on a stable scorecard produces roughly two individual coaching points and one systemic issue in a normal week, not one point per person. Fixing the cadence while the supply of real problems varies forces leads to invent the difference, and agents work that out within a month. Coach where the evidence points, tell the team that is the rule, and let most weeks be quiet for most people.
What do you do when there is nothing to coach?
Say so, name one thing that looked good, and give the rest of the slot back. Then fix the structure that made it hard: add no coaching point this week as a recordable outcome in the template, and report on whether the weekly review happened rather than on how many points it produced. The one caveat is that nothing to coach must be a conclusion reached after doing the review, never a substitute for doing it, and agents can tell the difference immediately.
Why do managers avoid coaching?
Usually because the version they have been asked to run is not coaching. Producing a development point for every agent every week means trawling conversations for something raiseable, delivering feedback the lead does not believe in, and absorbing the resentment it generates. That is unpleasant work with no visible payoff, so it gets deferred, and then done badly at speed. Leads who are allowed to coach two or three people on real findings and skip the rest generally do not avoid it.
What is the 70/30 rule in coaching?
The rough guide that the person being coached should be talking about seventy percent of the time and the coach thirty. It is a useful check on a session that has turned into a lecture, but it treats a symptom. If a lead is doing most of the talking it is often because they arrived without evidence and are filling the silence with generalities. Open the conversation the point came from, ask what was happening in it, and the ratio tends to correct itself.
Related terms
Find the two coaching conversations your week actually earned
Bring a month of your own scored conversations. We will run the same triage against your data and show you the one systemic issue, the two people the evidence supports, and the weeks it says to leave alone.
