Where we are
Three sessions of building things - ladders, levelling, promotion. This one is about what happens when it does not work out, and the framing that separates good organisations from anxious ones. A single underperformance case is a management task. A pattern is a system output, and systems produce patterns for reasons: hiring bars that were never written down, roles rewritten by reorgs without anyone re-examining who is in them, managers with no time to coach, promotions into scope people had never held. Every one of those is a decision made upstairs. Before you tighten your improvement-plan policy, it is worth asking which of your cases your own design is generating - because tightening the policy makes those cases faster and more defensible without making them fewer.
Two of the four causes are yours 7 min live
Your managers will learn to diagnose will, skill, fit and context in the practitioner track. Read the same four from the top of the organisation and something becomes obvious: only two of them are really a manager's to solve.
LiveWhat your plan policy incentivises3 min▶
Whatever your policy says it is for, managers read it as a set of incentives, and two effects are near-universal:
- If a plan is mandatory before any exit, the plan becomes a formality. Managers who have decided will run the process to get through it, which produces exactly the pretext problem that raises your legal risk rather than lowering it. Employment-law commentary is increasingly direct about this: plans written as paperwork for a predetermined decision are a misuse of the tool, and disputes now turn on whether requirements were reasonable and achievable and whether the plan was applied consistently to people in similar situations.
- If a plan is heavy and slow, managers avoid opening one at all - and instead tolerate the situation for another year, which is worse for the team and worse for the person, who loses the chance to fix something while it is still fixable.
The design goal is a policy where a genuine plan is light enough to start early and rigorous enough to mean something, and where a manager who has honestly already decided has a legitimate route that is not a fake plan. That route is the exit conversation, and it is b9.
Self-studyThe hiring bar is a performance-management decision2 min read▶
The cheapest place to prevent underperformance is before someone starts, and it is usually the least examined.
- Unwritten bars produce inconsistent hires. If "good enough" lives in each hiring manager's head, some teams hire above the bar and some below, and the difference shows up two years later as a performance pattern in specific teams.
- Hiring under pressure imports future cases. A role open for five months gets filled by whoever is available in month six. That is a rational trade, but it should be a recorded one, with a ramp plan attached.
- Track it. Underperformance cases by hiring route, by hiring manager, and by time-to-fill. If cases cluster in roles that were filled in a hurry, you have found a system cause you can act on.
Your exposure depends on where they sit 6 min live
Employment law is local, fact-specific, and moves. This is a map of the shape of the obligations, not advice - and if you operate in more than one of these places, your global policy has to be written to the strictest one or written per jurisdiction on purpose.
LiveThe keeper test: the strongest argument against plans4 min▶
Take the opposing case seriously, because it is well argued. In No Rules Rules, Reed Hastings describes Netflix's keeper test - would you fight to keep this person if they resigned tomorrow? - and argues that improvement plans rarely help anyone improve and mostly delay a firing by weeks or months. The economics are direct: generous severance costs roughly what you were going to spend managing the person out anyway, and it is more honest.
- What is right about it: a plan run as a countdown is worse for everyone than a clean, well-compensated parting. If you have decided, saying so and paying properly is the humane option.
- What it removes: process. The keeper test is a single manager's judgment with nothing checking it - and manager judgment carries known bias, including rating people who resemble us more favourably. Strip out the process and you have not removed bias, you have removed the thing that catches it.
- What it assumes: a jurisdiction where severance can substitute for fair process, and a balance sheet that can fund it. Neither is universal.
- The synthesis for your policy: do not mandate fake plans, do fund honest exits, and keep the evidence and consistency requirements regardless of which route a case takes.
Read your own case log as a system output ★ 12 min · think it through
Take the last ten formal performance cases in your organisation and classify them by cause. The distribution tells you what your design is manufacturing.
For each case, name the cause honestly: skill, will, fit, or context. Use the evidence, not the file's framing - files tend to record everything as a will problem.
Note what preceded each one: a reorg, a role rewrite, a manager change, a rushed hire, a promotion into new scope.
Count the bottom row. Fit and context cases are your decisions arriving late with someone else's name on them.
Pick the most common upstream event and decide what changes in how you do it.
| Cases | Cause | Preceded by | The system fix |
|---|---|---|---|
| 4 of 10 | Context | The systems reorg that rewrote three roles | Re-examine who is in a role when the role changes. Nobody had |
| 2 of 10 | Fit | Promotion into scope they had never held | Stop promoting on tenure; see a3 and a4 |
| 2 of 10 | Skill | Roles filled in a hurry after 5 months open | Attach a real ramp plan to any rushed hire |
| 2 of 10 | Will | Manager change, no handover conversation | Require a proper handover when a manager changes |
Decide your actual policy ★ 8 min · your org
Four decisions that belong to you and cannot be delegated to a template. Make them explicitly, because if you do not, your managers will infer them from whatever the form requires.
Is a plan mandatory before an exit? If yes, accept that some plans will be formalities and decide how you will detect that.
What is the minimum a plan must contain? Insist on the teaching component, not just criteria and dates - most templates omit support entirely.
Where is the redeployment route? A fit case needs somewhere to go. Without an internal route, every fit case becomes an exit by default.
Who checks consistency? Someone outside the reporting line should be able to see that similar cases were treated similarly, before an exit happens rather than after.
The consistency check that stopped an exit. Cartwheel introduced a rule that no performance exit could proceed without an HR partner confirming how comparable cases had been handled. The first time it ran, it surfaced that two analysts with near-identical issues had been treated completely differently - one coached for six months, one moved to a plan in three weeks - and the difference tracked to which manager they had, not what they had done. Nothing about that was discoverable from either individual file. It is the cheapest control in this session and it works precisely because it looks across cases rather than at one.
Try it yourself - this week ◐ 30-45 min total
- Classify your last ten formal cases by cause, and count how many were organisational in origin.
- Read your own plan template and check whether it requires any teaching or support. Most do not.
- Decide whether you have an internal redeployment route for fit cases. If not, note that every fit case is currently an exit.
- Introduce a consistency check outside the reporting line, before exits rather than after.
- Ask whether any recent reorg rewrote roles without anyone re-examining who was in them. That is where the context cases come from.
- Re-verify the Acas Code status and your local guidance before you publish any policy, and get counsel to review it.
Sources covered (cited)
Taught from published employment-practice guidance and the improvement-plan literature. This session covers:
Three questions before you go 🎯 ◐ 90 seconds
1 · Six of your last ten performance cases followed a reorg that rewrote roles. What does that tell you?
A single case is a management task; a pattern is a system output. Fit and context cases are organisational decisions arriving late with someone else's name attached. Re-examine who is in a role when the role changes.
2 · What happens if you make an improvement plan mandatory before every exit?
Managers who have already decided will run the process to get through it. Disputes turn on whether requirements were reasonable and achievable and whether the plan was applied consistently - so a plan written as paperwork for a predetermined decision increases exposure. Give managers a legitimate honest-exit route instead.
3 · What does the keeper test remove, alongside the fake plans it correctly avoids?
The argument against countdown plans is strong. But the keeper test is one manager's judgment with nothing checking it, and it assumes a jurisdiction and a balance sheet that permit substituting severance for fair process. Do not mandate fake plans; do keep evidence and consistency whichever route a case takes.