Underperformance is the conversation most managers delay the longest and handle the worst. The delay isn't laziness; it's because the situation looks ambiguous from the outside and the cost of getting it wrong feels personal. But the ambiguity usually dissolves the moment you run a structured diagnosis instead of staring at the symptoms. This unit gives you that diagnosis, the conversation that follows it, and the plan that turns the conversation into something both of you would defend as fair.

Before you say a word to the underperformer, you need to know what kind of problem you're actually looking at. The cleanest cut is the Skill vs. Will Diagnosis: skill issues point to capability, knowledge, or experience gaps, while will issues point to motivation, commitment, or engagement barriers. They look identical from the outside (missed deadlines, lower-quality work, withdrawal in meetings) and they require completely different responses. Treat a skill problem as a will problem and you'll demoralize someone who's drowning. Treat a will problem as a skill problem and you'll buy them a course they don't need while the real issue rots underneath.

The structured root-cause check is a small set of questions you run before the conversation. Has this person done this kind of work successfully before? Has something in the work itself changed (new scope, new tools, new dependencies)? What does the trend look like over the last few weeks, not the last few days? Are the misses concentrated in one area or spread across everything? When you've coached or trained on the gap, did the work move? And critically: what's changed in their life or in the team around them that you can name?

A skill diagnosis usually surfaces with a stable trend, a specific cluster of work where the gap shows up, and visible improvement when support is added. A will diagnosis tends to look like a recent shift, broad rather than narrow withdrawal, and no real change when you offer help. Many real cases are a blend, and that's fine: the diagnosis isn't a label, it's a hypothesis you walk into the conversation ready to test.

• Jake: I've got someone slipping. Three missed deadlines in two weeks, quiet in design reviews. I'm thinking it's a skill thing, the new architecture is genuinely hard.
• Nova: Was she clearing this kind of work six months ago?
• Jake: Yeah, actually. Crushing it.
• Nova: So what changed about the work, or about her?
• Jake: ...The work didn't really change. She got passed over in the last cycle.
• Nova: That's not a skill problem yet. That's a will problem you're about to mishandle by sending her to a course.

Notice the move: same symptoms, very different intervention, decided before the conversation rather than during it.

Once you know what you're looking at, the conversation has two failure modes and you have to design against both. Lead with so much empathy that the firm line disappears, and your report leaves thinking the situation is fine. Lead with so much firmness that the room never opens up, and you'll never learn what's actually underneath the pattern. The goal is empathy and firmness in the same conversation, not one or the other.

Open by naming the purpose directly. Something like "I want to talk about a pattern I've been seeing, and I want to understand what's going on" works. Then deliver the specifics using the SBI structure you already know: the situations, the observable behaviors, the impact. Hold the firm line on what has to change, then deliberately stop talking. The space you create after the firm line is where the real information lives. Most managers fill that silence with reassurance and lose the conversation in the process.

When the response comes, listen for what's underneath the deflection. "It's been busy" is a surface answer. So is "I'll do better." Ask one more open question and wait. If the diagnosis is will, the real reason is usually a missed promotion, a relationship that broke, a project that drained them, or something happening outside work. If the diagnosis is skill, the real reason is usually a specific piece of the work they don't know how to do and haven't admitted. Neither answer comes out unless you make room for it.

The plan that follows the conversation is what makes the whole thing real, and the discipline here is specificity. A time-bound improvement plan needs four sections. Measurable milestones at 30, 60, and 90 days, each tied to observable output or behavior rather than effort or attitude (so "ship the migration with no rollback" rather than "show more engagement"). A weekly check-in cadence, with a clear answer to who runs the agenda and what triggers an off-cycle conversation. Named support, meaning the specific scaffolding you're providing (a pairing, a scope adjustment, an obstacle you'll remove) that maps to what they told you in the conversation. And explicit consequences, stated cleanly, for what happens if the milestones at 30, 60, or 90 days are missed.

The plan must read as humane and firm at the same time. Vague milestones fail your report, because they have no idea what good looks like. Vague consequences fail the company, because the plan won't hold up if it's later escalated. Specificity is the kindness here, not the cruelty.

The single takeaway: underperformance is a diagnosis problem before it's a conversation problem, and the conversation is a structure problem before it's an empathy problem. Everything above is theory until you run it on a real case, so what comes next is a quick pattern-spotting exercise on skill-versus-will vignettes, a peer pressure-test of the conversation you'd actually walk into tomorrow, and a draft of an improvement plan tight enough that both your report and HR would call it fair.