Getting performance management right the first time

Performance management has a bad name. Most people know only its worst version: a form filled in once a year, a rating no one can explain, and a talk both sides wanted to end.

The helpful parts (clear about what good looks like, said while it still matters) are less ceremonial than what gets systematized.

A note on scope. This page is about the practice of setting expectations and giving feedback. It is not legal guidance. Rules covering hiring, discipline, dismissal, pay, record-keeping, protected characteristics, and grievance procedures differ substantially between countries and often between regions within them, and your employer will have its own binding policies. Before you act on anything with formal consequences, take advice from your HR function or your organization's employment counsel. Nothing here should be read as a statement of what the law requires of you.

In the US, performance decisions may also be affected by Title VII of the Civil Rights Act of 1964 and the Americans with Disabilities Act (ADA), enforced by the Equal Employment Opportunity Commission (EEOC), and the Family and Medical Leave Act (FMLA), administered by the US Department of Labor's Wage and Hour Division. These laws do not prescribe one universal private-sector appraisal form or annual review schedule; coverage, facts, and employer policies matter. Before starting a formal process, ask HR or employment counsel which requirements apply.

What to take away

An expectation is a description of an outcome, not of effort

The most common cause of a difficult performance conversation is that the expectation was never actually stated. It was assumed, implied in an interview, or communicated once eight months ago in a different context. Setting direction is one of the standing obligations that cannot be delegated to a process, and it is the first of the four in management foundations.

A stated expectation has a few properties worth checking for.

It describes something observable. "Be more proactive" cannot be observed and therefore cannot be met or missed: only argued about. "Raise blockers before they hit the deadline rather than on the day" can be.

It is inside the person's control. Holding someone to an outcome that depends mostly on other people's decisions produces resentment rather than improvement. If the outcome is shared, say what their part of it is.

It has a time frame. Without one, "not yet" and "not going to" are indistinguishable, sometimes for a year.

They can repeat it back. This is the test that catches the rest. If you ask what the priority is for the next quarter and get a different answer from the one in your head, the expectation does not exist yet regardless of what you wrote.

Write expectations down at the start of the period and revisit them when circumstances change. An expectation set in January and never touched will, by autumn, be measuring something the business no longer wants.

When the plan they were derived from moves, they have to move with it, which is the cadence problem in strategic planning. Guidance from the US Office of Personnel Management on the planning stage of a performance cycle treats this written start as the part that carries the weight.

Feedback and evaluation are different activities

Conflating them is why both get done badly.

Feedback is information about a specific thing that happened, given close to when it happened, so the person can adjust. It is low-stakes, frequent, and mostly forgettable. Its purpose is course correction.

Evaluation is a judgment about a period of work, usually with consequences attached: pay, progression, role. It is high-stakes and infrequent. Its purpose is a decision.

Saved feedback delivered as evaluation is the worst of both. The person hears months-old observations they cannot act on, at the moment of highest stakes and peak defensiveness.

The manager, having avoided the small conversations, compresses them into one, which makes the whole thing feel like an ambush.

The practical rule: nothing in an evaluation should be new. If it is, the failure is in the preceding months, not in the meeting.

Giving feedback that lands

Three things do most of the work.

Be specific about the observable. Describe what you saw and what followed from it. "In yesterday's review you cut across the client twice; they stopped raising objections after that, and we did not find out what their real concern was." That is checkable. "You come across as dismissive" is a conclusion the person can only accept or reject.

Separate the observation from your interpretation, out loud. "Here is what I saw; here is what I made of it; tell me what I am missing." People are far more willing to engage with an interpretation that is presented as one. And you are sometimes wrong: the missed deadline had a cause you did not know about.

Say what should be different next time. Feedback that stops at the diagnosis leaves the person to guess at the remedy, and they will often guess at something more drastic than you intended.

Some practices worth avoiding. Burying criticism between two compliments trains people to dread compliments. Giving feedback about a pattern using a single incident as a proxy makes the conversation about that incident. And delivering it in front of others converts the topic into status.

Positive feedback follows the same rules and is more neglected. "That was good" is nearly useless; the person cannot tell which part to repeat.

Goals: what they are good for and where they distort

Goals are useful for direction and for making trade-offs explicit. They are unreliable as a measurement system, for a reason worth understanding.

Any measure with consequences gets optimized, including in ways you did not intend. That effect is Goodhart's law. Tickets closed per day rise when tickets close prematurely. Lines of code, meetings attended, calls made: each proxy for value gets pursued instead of the value.

This is not cynicism about people. It is what happens when you tell someone exactly what will be counted.

Some ways to reduce the distortion:

SMART goals and objectives and key results (OKRs) are two common ways to make targets explicit. Neither replaces checking quality or adjusting goals when strategy changes.

Also decide in advance what happens when circumstances change. A goal invalidated in month two by a strategy change should be renegotiated, not marked as failed at year end.

Ratings, distributions, and calibration

Many organizations require a rating. If yours does, a few things are worth knowing about how they behave.

Forced distributions, requiring a fixed proportion of people in each band, are attractive because they prevent grade inflation. They are usually indefensible at team level, because they assume performance is distributed identically in every group.

A small team of strong performers and a small team of weak ones get the same shape imposed on them. Where you have no choice about the mechanism, be honest with people about what the rating reflects: a position within a distribution, not an absolute judgment of their work.

Calibration meetings, where managers compare their assessments before finalizing them, do address a real problem: different managers apply different standards, and the generous one's team gets better outcomes. The risk is that they become an advocacy contest in which visibility beats contribution.

Two things help: require evidence for each case rather than characterization, and pay attention to whose work is structurally less visible, such as people doing maintenance, support, or enabling work that shows up in other people's results. Noticing that work before it disappears is a daily habit rather than a review one, and it is covered in team management.

If you are on the receiving end of a rating you disagree with, ask what specifically would have produced a different one. A rating no one can attach to concrete examples is a signal about the process, and worth raising through whatever channel your employer provides.

The conversation when performance is genuinely a problem

This is where the practice ends and formal procedure begins, and where you should stop improvising.

What is generally within good practice: raising the concern early and directly, in private; being specific about what is not being met and what meeting it would look like.

Asking what is getting in the way, and meaning it, because the answer is often something you can fix. Unclear priorities, a missing skill nobody arranged training for, an obstacle they thought they were supposed to work around, or something in their life that is not permanent.

Agreeing a concrete next step and a point at which you will both look at it again, then writing down what you agreed, so neither of you is relying on memory.

Do not decide alone whether a formal process is triggered, what warnings mean, what documentation is required, or what timescales apply. Also not alone: what fair procedure is, and anything touching dismissal, demotion, or pay reduction.

Employment law, which varies by jurisdiction, and your employer's policies govern these. A wrong procedure can cause serious problems for the individual and the organization. Talk to HR or your employer's legal advisers before you start, not after.

When performance changes suddenly, the cause is often outside work. Health, disability, or family obligations may apply. That is another reason to involve people who know your jurisdiction's rules.

In any organization with a formal process, a manager who has already had the informal conversations is much better placed than one whose first step is a formal letter.

Running the review conversation itself

Assuming the substance has already been said in the preceding months, the meeting is mostly about making sure both people leave with the same understanding. A few mechanics that make a difference.

Send anything written in advance. Reading a document about yourself while its author watches is a bad way to take in information, and it guarantees the first response is emotional rather than considered.

Start with their view. Not as a courtesy, because their self-assessment tells you where the gaps in understanding are, and it is much easier to address a gap you have heard stated than one you are guessing at.

Deal with the consequences early, not at the end. If there is a pay or progression outcome, the person is waiting for it and will not absorb anything until they have it. Holding it to the end means the developmental conversation is delivered to someone who is not listening.

Separate the summary from the plan. What happened is one conversation. What changes next is another, and it often works better a week later, when the first has been digested.

Write down what you agreed, and send it. Two people remember the same conversation differently, particularly a tense one, and the divergence surfaces months later when it is hard to resolve.

Where your organization has a required form and process, follow it, but the form is a record of the conversation, not a substitute for one, and filling it in together on screen is a reliable way to have neither. Your employer may use records such as an appraisal form, a self-assessment form, a written warning, or a performance improvement plan (PIP); use the required versions rather than assuming a universal template.

Development is not the same conversation as promotion

These get merged, and the merge damages both. Promotion depends on organizational factors (whether a role exists, budget, structure, timing), that are often outside both of your control. Development is about capability, and it is available regardless.

Running them as one conversation means every discussion about growth becomes a negotiation about title, and someone who cannot be promoted this year hears that there is nothing to work on.

Some things that help:

Be honest and early about what is not available. If there is no route to the next level in the current structure, say so. People make reasonable decisions with accurate information and resentful ones with vague reassurance.

Ask what they want to be able to do, not what they want to be called. The answer is more actionable and frequently different from what the title conversation would suggest.

Development happens through work, not through courses. The main lever you have is what you assign and what decisions you hand over. A training budget is a small supplement to that, not a replacement.

Name what would need to be true. If progression is possible but not yet, describe the specific gap. "More strategic" is not a gap; "you have not yet led something where the outcome depended on people who do not report to you" is.

Assessing work you cannot see

Some contributions do not appear in anything you measure: the person who prevents incidents rather than resolving them, who does maintenance that keeps things from breaking, whose main output is other people's output. There is a structural bias against all of them, because visible work is easier to evidence in a calibration meeting.

Practical counterweights. Ask the team who they rely on, and take the answer seriously as evidence. Deliberately look at what did not go wrong in an area someone owns. And when you write an assessment, describe the mechanism rather than the artifact: "kept the release process stable through two system migrations" is evidence, even without a number attached.

Be aware of who this bias tends to fall on. Enabling, coordinating, and maintenance work is not distributed randomly across a team, and if the same people always end up doing it, that is worth examining on its own terms.

Self-assessment and upward feedback

Asking someone to assess their own performance is useful mainly as a diagnostic of alignment. Where their view and yours match, the conversation is short. Where they diverge sharply, that divergence is the conversation, and it is nearly always a sign that expectations were unclear.

Upward feedback, asking your team what you should do differently, produces little of value unless you have made it genuinely safe, and you cannot declare it safe. What you can do is act visibly on something small, so the first person who says something sees a result. Until then, expect politeness.

A 360-degree feedback exercise gathers perspectives from a person's manager, peers, and, where relevant, direct reports. Treat it as input for discussion, not as a rating by itself.

What to do this quarter

If you want to improve performance management without redesigning it:

Common questions

How often should feedback happen?

Close enough to the event that the details are still fresh for both of you. That means days, not months. Formal reviews can happen on whatever cycle your organization runs; the feedback cannot wait for them.

Should ratings be linked to pay?

Linking them focuses attention on the rating, not the development conversation, so people negotiate rather than reflect.

Separating them can preserve a more useful conversation, but pay decisions still need a basis. The compensation structure is usually not a manager's to change, so whatever your organization does, be straight with people about which conversation they are in.

What if I inherit a team with no records of past performance?

Start the clock now. Set expectations for the coming period explicitly and in writing, and say plainly you are assessing from here.

Judging people against standards they were never told about, based on a predecessor's undocumented opinions, is unfair. Depending on where you are and what follows, it is potentially a problem for your employer as well. Ask HR how to handle the transition.

Is it worth doing any of this in a very small organization?

Clarity is worth it. Paperwork mostly is not. Two people who know the priorities and talk about progress have the substance.

Small organizations need a written record when something serious arises. The informality that works day to day leaves you exposed in a dispute. Ask an employment adviser what records to keep in your jurisdiction, before you need them.