Detecting Underperformance & the Improvement Plan

This is the part managers most want to skip and most need to get right. Handled well, an improvement process either genuinely turns someone around or reaches a clean, fair, unsurprising exit. Handled badly — late, vague, undocumented — it produces surprise, resentment, and legal exposure. The discipline here is turning a feeling that something is off into evidence, then into a written plan with success criteria.

The signals of underperformance are consistent across teams: delivery that repeatedly misses commitments or stalls without the person flagging it; ownership gaps, where someone works quietly in a corner, waits to be told, and doesn't close loops; weak judgment surfacing in design and incident reviews; unresolved collaboration friction that erodes team trust; and a flat trajectory despite feedback. None of these is damning on its own. The manager's job is to notice the pattern while it is still small enough to fix.

A feeling is not actionable; a dated, factual record is. Log concrete instances in a running one-on-one document as they happen — what was expected, what occurred, the impact — in neutral language. Then calibrate with a peer manager or the skip-level to check you are not over-reading one person or under-reading another, and to keep the bar consistent across the team. And always separate ramp from underperformance: a genuinely new hire or someone on a genuinely new problem deserves runway, and "still learning the system" must be consciously distinguished from "not going to meet the bar."

When the pattern is real and evidenced, the improvement plan makes it explicit and fixable. A good plan states, in writing, seven things:

1. The gap — the specific pillars or behaviors below bar, with dated examples.
2. The bar — what "meets expectations" looks like at this level, in observable terms.
3. Success criteria — measurable enough that both sides can tell at the end whether it was met. Not "improve communication" but "owns feature X end-to-end and flags blockers within a day."
4. A timeline — a defined window with explicit start and end dates (commonly thirty, sixty, or ninety days; a coaching-tier window often runs three to six months).
5. Support — what the manager and company will provide: scoped work, mentoring, clearer priorities, fewer competing demands.
6. Check-in cadence — weekly or biweekly, each one written down with what was discussed and where things stand.
7. The stakes — what happens if the criteria are and are not met.

The framing is everything: this is a genuine, resourced attempt to help the person succeed, not a paperwork prelude to a decision already made. Some people really do turn around, and the plan is how.

Every check-in dated and written. The plan acknowledged by the employee. Evidence attached. This is not bureaucracy for its own sake — it is what makes the process fair to the person (they can see exactly where they stand) and defensible for the company. If the person is on protected leave, on a sponsored visa, or otherwise in a special situation, involve people/legal before starting the clock and plan the timeline around it. Getting the sequencing wrong here is the most common and most costly mistake in the entire firing half of the system.

A closing note that connects back to hiring: most underperformance that reaches an improvement plan was screenable at the interview. A team that will not fail a candidate on a weak fundamental answer ends up managing out later what it should have gated earlier. The improvement plan is the expensive, humane, last-resort version of a decision the loop is supposed to make cheaply up front. Run the loop honestly and you run this part rarely.