Performance management case study: findings and lessons

Most writing about performance management is about the conversation. Very little is about the document, and the document is what the conversation is made of.

This page is about the record: what goes in it, when it gets written, and what happens when it is assembled afterward from memory. Nothing here is legal advice, and where a formal step is contemplated the rules that apply are set by your jurisdiction and your own organization's policy.

What to take away

• Write the note within a day of the thing it describes. A note written three months later is a reconstruction, and a reconstruction assembled by someone who has already reached a conclusion will support that conclusion.
• Record what was observed and what was agreed, separately. Merging them produces a document that reads as judgment and cannot be checked.
• A record kept only when things go badly is a record of your attention, not of someone's performance.

What a note contains

Four fields, and it fits in a few lines. Longer notes do not get written.

The date, and the date of the thing. These differ, and the gap matters. A note written the same week is evidence; the same words written in month four are a summary of an impression.

What happened, observably. What was done, what was produced, what was said. Not what it showed about the person. The test is whether someone who was not present could picture it.

What was said about it, and by whom. Whether it was raised at the time, in what terms, and what the response was. This is the field most often missing, and its absence is what makes a later conversation land as a surprise.

What was agreed, with a date. Specific enough to be checked. "Will improve" is not an agreement; a named change with a review date is.

Why the memory version fails

Three failures, and they are structural rather than a matter of anyone being careless.

Memory reorders. Once you have formed a view, the incidents that support it come to mind readily and the ones that do not are harder to retrieve. This is ordinary and it is not corrected by trying harder; it is the effect known as confirmation bias, and the only real defense against it is contemporaneous writing.

Memory compresses. Six ordinary weeks and one bad afternoon become "a difficult period". The compression happens before you notice it, and it is the mechanism behind treating one incident as a pattern.

Memory borrows. Things you heard from others merge with things you saw, and by month three the distinction is gone. A record separates them because at the time it was obvious which was which.

The person on the other side of the conversation experiences all three as unfairness, and they are usually right to.

What does not belong in the record

Opinions about the person. Character descriptions cannot be checked, cannot be responded to, and do not tell anyone what to do differently.

Second-hand accounts recorded as fact. If it was reported to you, write that it was reported and by whom. Where it matters, go and look yourself before it goes in.

Anything about health, personal circumstances, or protected characteristics. These carry legal weight, vary by jurisdiction, and belong to a different process handled by people trained in it. If one is relevant to what happened, that is a signal to stop and take advice rather than to write it down.

A conclusion. The record supports a decision. It is not the decision, and a file that reads as a case for the prosecution has stopped being a record.

What most files are missing

The positives. A file that starts on the day concerns began has no baseline, and a period of good work with nothing written is indistinguishable from a period nobody watched.

The times you did not raise it. If something happened in March and was first mentioned in July, that gap is part of the story and it belongs in the file. Leaving it out makes the record look stronger and makes the conversation worse.

The conditions. Volume, cover, tooling, and who else was away. A drop in output during a period when a team of four was running with two is a fact about staffing, and a file that omits it attributes it to a person.

Your own part. Whether the expectation was stated, when, and whether the person had what they needed. Whether the conditions were yours to fix rather than theirs is a question about how the team is run, and it is in team management.

Keeping records for everybody, not just for problems

A file that exists only for people you have concerns about is a biased instrument by construction, and everyone can see that it is.

The practical version is small: a few lines per person per month, written at the same time for the whole team, recording what they did and anything agreed. Ten minutes a month for a team of six. It has three effects worth having. It gives you a baseline, so a change is visible as a change. It removes the signal that a file has been opened. And it makes the annual conversation about a written year rather than about a remembered quarter.

Public-sector guidance frames this as ongoing monitoring rather than as an event, and the performance management cycle published by the US Office of Personnel Management is a reasonable model of what that looks like when it is treated as a continuous obligation. The wider argument about separating feedback from evaluation is in performance management, and the questions worth asking before either conversation are in performance management questions.

Reading a file you have inherited

When a file arrives with a new team member, read it for three things before you read it for its conclusions.

Look at the writing dates against the event dates. Where they cluster, the file was assembled at one sitting and describes a decision rather than a history.

Look for what is absent. A file with no ordinary weeks in it is a record of concerns, not of performance.

Look for whether the person's response was recorded. A one-sided file tells you what somebody believed and nothing about whether it was tested at the time.

Then form your own view over a period, in writing, from the start. Inheriting somebody else's conclusion and continuing their file produces a document with two authors and one opinion. What a management record is for in general, and the obligations that come with holding one, is in management foundations, and the discipline of describing what was observed rather than what it means is the same one that keeps operational records usable, set out in operating processes metrics.

Common questions

How long should a note take to write?

Two minutes. If it takes longer you are writing a report and it will stop happening within a month.

Should people see their own notes?

The agreements, always, and preferably at the time they are made. The rest depends on your organization's policy, which you should know before you start keeping anything. A record the person has never seen is a poor basis for a conversation, whatever else is true of it.

Is this not a lot of bureaucracy for a small team?

A few lines a month per person is not a system. The bureaucratic version is the one that appears when a problem has already developed and someone has to reconstruct six months in a weekend.

What if my organization has a formal process that requires something different?

Follow it, and check it before you need it rather than during. This page describes what a useful record contains; it does not override anyone's policy, and formal steps in particular carry requirements that vary by employer and jurisdiction.