Ask a manager what "performance management" means and most will describe a calendar event, a form, a rating, a slightly awkward conversation in December. That's the artefact. The thing it's meant to produce is something else entirely: a person who knows what good looks like, can see how close they are to it, and feels able to adjust. When that loop runs fast, the annual form becomes a formality. When it runs slow, no form can rescue it.

The quick version

  • A feedback loop has three parts: a clear goal, a signal about how you're doing against it, and the room to change course. Break any one and the loop stops working.
  • Frequency beats formality. Specific feedback close to the work changes behaviour; a once-a-year verdict mostly explains it.
  • Feedback isn't automatically good. In the best meta-analysis we have, more than a third of feedback interventions made performance worse, usually by aiming at the person instead of the task.
  • So the move is to shorten the loop and lower the threat: set a specific goal, give small frequent task-focused signals, and make it safe to admit the gap.

The idea in depth: a loop, not a verdict

Borrow the term honestly and a feedback loop is a control system: a target, a sensor that measures the gap to the target, and an action that closes it. Run that on a person and the target is a goal, the sensor is feedback, and the action is the adjustment they make. Performance management is just the discipline of keeping all three present and fast. Most broken systems are missing a part, a fuzzy goal, no signal until review season, or a culture where the signal arrives but nobody dares act on it.

flowchart LR
  A(["Clear goal
(what good looks like)"]) --> B(["Do the work"])
  B --> C(["Specific signal
(feedback on the gap)"])
  C --> D(["Adjust
(needs safety + skill)"])
  D --> B
  C -. "stale, vague, or unsafe" .-> E(["Loop stalls"])
A working feedback loop closes the gap between intended and actual; weak signals or low safety break it. Leaders Loop

Start with the goal, or the feedback has nothing to push against

Edwin Locke and Gary Latham spent decades building goal-setting theory, and the central finding is unusually clean for social science: specific, difficult goals produce higher performance than vague "do your best" goals or easy ones, provided the person is committed, has the ability, and gets feedback on progress (Locke & Latham, American Psychologist, 2002, summarising 35 years of work). Across hundreds of studies, specific hard goals beat the alternatives in the large majority of cases. Crucially, the theory says goals and feedback are not two tools, they are one mechanism. A goal without feedback is a wish; feedback without a goal is just noise about a destination nobody named.

So the move is to make the target concrete before you worry about the review process. Replace "be more proactive" with "bring one option I haven't asked for to each of our weekly check-ins this quarter." Now both of you can see the gap, and every conversation has something to push against.

An honest limitation: the same research warns that specific, stretching goals can backfire on novel or genuinely complex work, where locking onto a target too early blocks the exploration the task needs. For a team still figuring out what good looks like, a learning goal ("test three approaches and report what you find") beats a rigid performance number.

Feedback is not automatically helpful, and the data is blunt about it

The most sobering result in this field comes from Avraham Kluger and Angelo DeNisi's 1996 meta-analysis in Psychological Bulletin, pooling 607 effect sizes from over 23,000 observations. Feedback raised performance on average, but in more than a third of cases it lowered it. Their explanation is the useful part: feedback works when it directs attention to the task and how to do it better, and tends to harm performance when it directs attention to the self, the ego, the rating, the comparison to peers. The moment a person hears "you're a 3 out of 5," cognitive effort drains away from the work and into defending or repairing the self.

Feedback that points at the task tends to help; feedback that points at the person tends to hurt.

This is the quiet case against the annual rating. A single score, delivered once, late, by one manager, is almost engineered to trigger the self-focus that Kluger and DeNisi found so corrosive. It's also statistically fragile. When Deloitte studied its own process for "Reinventing Performance Management" (Buckingham & Goodall, HBR, 2015), it found ratings were dominated by "idiosyncratic rater effects", most of the variance in a score reflected the quirks of the rater, not the performance of the rated, and that the firm was spending close to two million hours a year on the ritual. Their response wasn't a better form; it was more frequent, future-focused check-ins.

So the move is to separate the two jobs we usually cram into one review. Keep evaluation (pay, promotion) honest but rare and well-governed; run development feedback far more often, in small task-focused doses, close to the work it's about. A thirty-second "here's what landed and here's the one thing to tighten" after a real meeting will change more behaviour than a paragraph written in December.

An honest limitation: "abolish ratings" became a fashion, and some firms that removed them found managers gave vaguer feedback and employees felt less clear about where they stood. The lesson isn't no judgement, it's that frequency and specificity have to replace the rating, not just delete it.

The third part of the loop: safety to act on the signal

A perfect goal and a perfect signal still fail if the person can't afford to admit the gap. Amy Edmondson's research on psychological safety, the shared belief that the team is safe for interpersonal risk-taking, began with a counter-intuitive finding in her 1999 Administrative Science Quarterly study: better teams appeared to make more errors, until she realised they were simply more willing to report them. Safety is what lets the sensor in the loop read true. Without it, people hide the gap, and feedback bounces off.

Edmondson is careful, in The Fearless Organization (2019), to reject the soft reading: psychological safety is not lowering the bar or being nice. It is the pairing of high standards with high safety, the combination she calls the "learning zone." High standards plus low safety produces anxiety and hidden problems; high safety plus low standards produces a comfortable, complacent team. Performance management lives in the top-right quadrant or nowhere.

quadrantChart
  title Standards x Safety
  x-axis "Low psychological safety" --> "High psychological safety"
  y-axis "Low standards" --> "High standards"
  quadrant-1 "Learning zone"
  quadrant-2 "Anxiety zone"
  quadrant-3 "Apathy zone"
  quadrant-4 "Comfort zone"
Edmondson's learning zone: high standards only produce performance when paired with the safety to speak up. Leaders Loop

So the move is to make admitting a gap cheap and routine. Edmondson's own advice is to lead with curiosity and your own fallibility: ask "what am I missing?" and "where did I make this harder for you?" before you ask anyone to account for their numbers. A manager who visibly takes feedback well is building the exact condition that makes their own feedback usable. (This connects directly to motivating & inspiring teams, the controlling, scorekeeping flavour of performance management is also what quietly crushes intrinsic drive.)

An honest limitation: psychological safety is a team-level climate, not a switch a manager flips in one meeting. It's also been over-claimed, it predicts learning behaviour reliably, but it is a necessary condition for performance, not a sufficient one. Without the goals and the standards, a safe team is just a pleasant one.

A worked example

The figures and people below are illustrative. Priya runs a six-person support team on the company default: SMART goals set in January, a mid-year "touch base," a calibrated rating in December. By autumn, one analyst, call him Tom, was clearly struggling with escalations, but his last real signal was eight months old and his goal said only "deliver excellent customer outcomes." Priya had a verdict to deliver and nothing to push against.

She rebuilt the loop around the three parts. Goal: she replaced the slogan with something measurable and committed-to, "resolve 80% of escalations without a second hand-off, and flag the ones you can't, within the day." Signal: instead of waiting for review season, she started a five-minute Friday review of two real tickets each, talking only about the work, what the customer needed, what the next move was, never about Tom as a "weak" or "strong" performer. Safety: she opened by walking through a ticket she had botched, so admitting a miss became the norm rather than the risk.

Within a quarter Tom's hand-off rate fell, not because the standard dropped, but because the gap was now visible weekly and cheap to admit. The December rating still happened, but it surprised no one: the loop had already done the real work, and the annual form went back to being a summary, not the system.

Frequently asked questions

Should we just scrap ratings and annual reviews?

Not blindly. The research is against relying on ratings to develop people, they're noisy and they trigger self-focus, but organisations that simply deleted them sometimes ended up with vaguer feedback and anxious employees who'd lost their only signal of where they stood. The defensible position: keep a rare, well-governed evaluation for pay and promotion decisions, and move all the development work into frequent, specific, task-focused check-ins.

How often should feedback actually happen?

Close enough to the work that the person can still picture the specific moment. For most teams that means light-touch feedback weekly or per-deliverable, not quarterly. The principle from Locke and Latham is that feedback has to let someone monitor progress and adjust, feedback that arrives after the chance to change has passed is just commentary.

What's the difference between feedback and a feedback loop?

Feedback is one signal. A loop is the whole circuit: goal, signal, and a change that feeds back into the next attempt. Plenty of managers give feedback into a void, no shared goal it maps to, or no room for the person to act on it. That isn't a loop; it's a one-way broadcast, and it rarely changes anything.

My direct report gets defensive at any feedback. What do I do?

Treat defensiveness as a signal about safety and framing, not just character. Kluger and DeNisi's finding suggests the content is probably pointing at the person, not the task, "you're disorganised" rather than "this doc was missing the decision we needed up top." Shift to the specific task, ask for their read first, and make your own mistakes visible. Defensiveness usually drops when the feedback stops feeling like a verdict on the self.

Does this work for remote and asynchronous teams?

The principles hold; the medium changes. Remote work makes informal signals scarcer, so the deliberate part, explicit goals, scheduled short reviews, written task-focused feedback on the actual artefact, matters more, not less. The risk is that feedback gets saved up for the next call and arrives stale, which is exactly the failure mode to avoid.

Related in the Toolkit

Where to go next