Where we are
Last session you learned to name the cause before choosing the intervention. This session is about the intervention managers reach for most and understand least. Be clear about the honest state of play: a great many improvement plans are written as a formality after the decision to terminate has already been made, and employment-law commentary is increasingly explicit that this is a misuse of the tool which raises rather than lowers legal risk. Disputes now turn on whether the plan's requirements were reasonable and achievable, and whether it was applied consistently to people in similar situations. So the question is not "how do I write the document". It is whether you are running a plan at all, or just a countdown with a form attached - and the lab is built to make that difference visible in two numbers.
Recovery and defensibility are different work 6 min live
Most managers optimise one and assume the other follows. It does not. Coaching and an honest early warning are what make recovery possible; dated evidence and a consistency check are what make the process defensible. A plan strong on one and weak on the other fails in a predictable way.
LiveWhy the bottom left is the most common plan3 min▶
The countdown quadrant is not produced by bad intent. It is produced by a manager who has avoided the conversation for months, is now anxious, and copies a template from an HR portal on a Friday afternoon. The plan then contains criteria nobody thought hard about, no teaching, and a timeline picked because it sounded standard.
- It fails the person - they get a shock, a document, and no instruction.
- It fails the organisation - the record shows a plan appearing from nowhere with no prior warning, which is exactly the shape that reads as pretext later.
- It fails the team - colleagues watch someone go from no feedback to formal process, and quietly conclude the same could happen to them.
The fix is not a better template. It is the six components in Part 2, and most of them are things you do before the document exists.
Self-studyThe Netflix alternative, and its problem2 min read▶
There is a serious argument that you should not run plans at all. In No Rules Rules, Reed Hastings sets out 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. His economic point is blunt: generous severance costs roughly what you would have spent managing the person out anyway.
- What is genuinely right about it: a plan run as a countdown is worse than an honest, well-compensated parting. If you have already decided, the humane and cheaper move is to say so and pay properly.
- What it is exposed to: the keeper test is a manager's judgment with no process around it, and manager judgment carries known bias - we rate people who resemble us more highly. Remove the process and you have removed the thing that catches that.
- What it assumes: a jurisdiction and a balance sheet that allow it. In much of the world you cannot substitute severance for a fair capability process, and a small company cannot fund four months of it.
Hold both: do not run a fake plan, and do not pretend that skipping process is the same as removing bias.
Six components, and which meter each moves 5 min live
Here is the whole design, and the thing worth memorising: they are not interchangeable. Two of them mostly buy recovery, two mostly buy defensibility, and two buy both. Miss the wrong pair and you have a plan that is kind but indefensible, or airtight and useless.
LiveThe two shortcuts to never take3 min▶
Both of these are things reasonable managers do under pressure, and both are visible in the lab as an immediate collapse of both meters:
- Shortening the timeline "to save management time." A two-week plan cannot show a change in the work. It reads as a countdown, the person spends it applying for jobs, and the first question any reviewer asks afterwards is whether the target was achievable in the time given.
- Writing the plan after the termination decision. This is the one that ends up in front of a lawyer or a tribunal. A plan produced as paperwork for a decision already made is a pretext, and it is precisely what gets examined - was it genuine, achievable, and applied consistently? You also destroy whatever recovery chance existed.
If you have honestly already decided, the ethical route is the one in b9: say so, and part properly. Do not dress a decision up as an opportunity.
The PIP Lab ★ live · two meters, six levers, three cases
Pick a case, then build the plan one component at a time and watch both meters. Start with nothing on. Then turn on all six. Then take a shortcut and watch what it costs. Finally switch to the third case with all six still on - that is the one to sit with.
| Move | Recovery | Defensibility | The lesson |
|---|---|---|---|
| Skill case, nothing on | 22% | 8% | A plan in name only |
| + coaching with named support | 61% | 51% | Teaching is what moves recovery |
| All six components on | 88% | 92% | This is a real chance |
| All six, then shorten to two weeks | 32% | 36% | One shortcut undoes most of the work |
| All six, then write it post-decision | 5% | 13% | Pretext destroys both jobs at once |
| Context case, all six on | 31% | 92% | Perfect process, wrong cause |
Turn a countdown into a real plan ★ 8 min · do your own
Here is a plan a Cartwheel manager actually drafted for the analyst whose forecasts arrive late. Read it, find which of the six components are missing, and rebuild it.
The draft: "Priya must improve the quality and timeliness of her forecasting work over the next 30 days. Performance will be reviewed at the end of the period."
Score it against the six. Criteria she can meet? No - "improve quality" is not measurable. Evidence from before? None cited. Coaching and support? Not mentioned. Early warning? This is the first time she hears it. Timeline? Picked, not reasoned. Consistency? Never checked.
Rebuild it with all six, and notice how much of the rebuild is not writing at all - it is arranging teaching and checking your own record.
Then ask the b6 question you should have asked first: is this actually a skill gap? If she has never been taught the forecasting method, the plan is instruction with a review date, not a warning with one.
| Component | The draft | Rebuilt |
|---|---|---|
| Criteria | "improve quality and timeliness" | "Weekly forecast delivered by Tuesday 12:00, with variance under 10%, for four consecutive weeks" |
| Evidence | none | Six dated examples from the last two quarters, with what was late and what needed rework |
| Support | none | Two hours a week with the senior analyst on the forecasting method, plus a worked template |
| Early warning | this is the first she hears of it | Raised in the 1:1 in month two, in writing after, with the offer of help then |
| Timeline | 30 days, unreasoned | Eight weeks - four to learn the method, four to demonstrate it independently |
| Consistency | never checked | Checked against how the same gap was handled for two other analysts at the same level |
The plan that worked, and why nobody noticed. The rebuilt version above ran at Cartwheel and Priya is now the person who trains others on the forecasting method. Nothing about that outcome looks like performance management - there was no formal drama, and the file records an eight-week plan that closed successfully. Successful plans are invisible, which is exactly why the tool has such a bad reputation: the ones everyone remembers are the countdowns, because those are the ones that end in an exit and a story.
Try it yourself - this week ◐ 30-45 min total
- Find any improvement plan you have written or received. Score it against the six components and place it in one of the four quadrants.
- For any current concern, write the honest early warning you could give this week - the thing that means no future plan is ever a surprise.
- Take a live case into the lab and be honest about which cause it is. If it is a context case, write what you would fix instead of the plan.
- Check your own consistency: has a similar gap at the same level been handled the same way? If not, that is a finding about you, not the person.
- Ask HR what your organisation's plan template actually requires, and which of the six it leaves out. Most templates cover evidence and criteria, and say nothing about teaching.
- If you have honestly already decided, do not write a plan. Read b9 before you do anything.
Sources covered (cited)
Taught from employment-practice guidance and the improvement-plan literature. This session covers:
Three questions before you go 🎯 ◐ 90 seconds
1 · Which pair of components mainly buys recovery rather than defensibility?
Teaching and honesty move recovery; evidence and consistency move defensibility. In the lab, adding coaching to the skill case takes recovery from 42% to 61%, while the consistency check barely moves recovery and takes defensibility to 92%. A plan needs both kinds of work.
2 · In the lab, the context-mismatch case with all six components on reaches 31% recovery and 92% defensibility. What does that mean?
You cannot improvement-plan your way out of a problem that was never the person. When the role was rewritten, the team lost people, or the manager is the issue, a defensible plan just fails more tidily. Diagnose the cause first - that is b6.
3 · Why is writing the plan after deciding to terminate worse than not writing one?
A plan written as paperwork for a decision already made raises legal risk rather than reducing it: disputes turn on whether requirements were reasonable and achievable and whether it was applied consistently. In the lab it drops the skill case to 5% recovery and 13% defensibility. If you have decided, part properly instead.