How to have an underperformance conversation

The underperformance conversation is the one new managers delay longest, and the delay is what makes it hard. A conversation about last week's work is a ten-minute discussion. The same conversation held six months later is about six months of silence, and the person's first reaction will be about the silence rather than the work.

Is this actually a performance problem?

Four things look like underperformance and are not. Rule them out before you book anything.

Unclear expectations. If you cannot write down what good looks like in one sentence, the person could not have known what to aim at. This is the most common cause and the one managers are least willing to see, because the fix is on their side.

Missing tools or access. Someone waiting three days for approvals is not slow. They are blocked.

Workload. If the volume is genuinely undeliverable, quality is the first thing to go, and telling someone to improve quality without changing the volume is asking them to do something impossible.

Something outside work. Illness, caring responsibilities, a bereavement they have not mentioned. You do not need the details, but you do need to know whether this is a capability question or a temporary one. If the signals point that way, talking to an employee who is burning out is the conversation to have instead.

If none apply, it is a performance conversation.

How do you open it?

With the message. Not with a question.

The standard advice is to open by asking how they think things are going. It is well-intentioned and it usually backfires. Either they say "fine" — and now you have to contradict them, which is a worse start than beginning honestly — or they realise it is a trap, answer defensively, and you have taught them not to trust an open question from you again.

Say it directly:

"I want to talk about the monthly client reports. They're not at the standard we need, and I want to be specific about what I mean. Then I want to hear your side of it."

Fifteen seconds. It names the topic, states your position, and signals that their response is part of this rather than an obstacle to it.

What counts as a specific example?

An example is specific enough when the person could use it to do something differently next time.

"The reports aren't detailed enough" is not usable. "In the June and July reports, the variance section gave the number but not the cause, so the client asked us both times and we had to go back and find it" is. It names the artefact, the recurrence, and the consequence.

Two examples are the right number. One can be dismissed as a bad day, and reasonably so. Two establishes a pattern without turning the meeting into a prosecution. Five tells the person you have been keeping a file, and their reaction will be to that rather than to the work.

Where you have to give a judgement rather than a fact — "this reads as rushed" — say that you are giving an impression and give the detail that produced it.

What do you do when they push back?

Separate the facts from the meaning, and settle the facts first.

Most pushback in these conversations is an argument about significance disguised as an argument about facts. The person is not really disputing that the variance section was thin; they are disputing that it matters enough to warrant this meeting. Trying to settle both at once produces a circular twenty minutes.

"There are two things here. Whether the reports had those gaps, and whether it's a big enough deal to be having this conversation. Let's do the first one — do you disagree that the June and July reports had that gap?"

Once the facts are agreed, the significance argument is short, because you are the one who gets to set the standard. If the pushback is heated rather than reasoned, the specific moves are in what to do when someone gets defensive.

What should you agree at the end?

One change, one date, one offer of support.

New managers routinely agree to five improvements, which guarantees none of them happens and makes the review meeting impossible to run. Pick the change that matters most and let the others wait.

The date is what turns a conversation into a process. "Let's look at this again on the 20th, after the next report goes out" gives both of you a fixed point. Without it, you will either raise it again too early, which reads as monitoring, or not at all.

The support should be concrete and small. "I'll review the next one before it goes out" is support. "Let me know if you need anything" is not, because it puts the burden of asking on the person who has just been told they are not good enough.

Why send it in writing?

Because two people leave every difficult conversation with different accounts of it, and that is normal rather than dishonest.

Send it the same day, short and neutral. What we discussed, what we agreed, when we will look at it again. No new information, nothing that was not said out loud, no softening and no hardening.

This is not building a file. It is making sure that in three weeks you are both working from the same version. If it does later become a formal process, having contemporaneous notes is what makes that process fair — but that is a by-product, not the purpose. The same written record is what makes the eventual performance review conversation straightforward rather than a surprise.

How do you keep it from feeling like an ambush?

By making sure the meeting contains no information the person is hearing for the first time.

That sounds impossible for a first formal conversation, and it mostly is — which is the argument for having small, low-key versions of it much earlier. A sentence at the end of a one-to-one — "the variance section was thin again this month, can we tighten that next time" — costs nothing and means the eventual structured conversation is a summary rather than a revelation.

Where you have not done that, say so at the start. "I should have raised this in July and I didn't, so some of this will be new. That's on me." It costs one sentence, it is true, and it removes the objection the person would otherwise spend the first ten minutes making. It also makes the standard you are about to set feel like a starting line rather than a verdict on work already done.

What if nothing changes?

Then the next conversation is shorter, and it is explicit about consequences.

"We talked on the 5th, we agreed the variance section would include causes, and the September report didn't. I need to be clear about what happens if this doesn't change." Then say what happens. Actually say it — this is the point at which vagueness stops being kind.

Before you get there, check one thing honestly: did you review the next report, as you said you would? Managers frequently agree to support, provide none, and then treat the lack of improvement as confirmation. If you did not do your half, do it now and reset the date once.

What does this conversation look like when it goes well?

It is short, slightly uncomfortable, and boring afterwards.

The person is not surprised, because you have been giving feedback along the way. They disagree with some of it, which is healthy. You leave with one written commitment and a date. Nobody cries and nobody wins.

If your underperformance conversations are dramatic, the drama is usually a symptom of delay rather than of the message. The structure that keeps them small is in the complete guide to difficult conversations for new managers, and the single biggest lever is having the first one weeks earlier than feels necessary.