Feedback and difficult conversations

Feedback is how behaviour changes, and a difficult conversation is feedback at the moment it costs something to give. When feedback fails, it is usually not because it is wrong. The delivery is the problem: too vague to act on, too late to matter, or so cushioned that the message disappears. The competency covers both directions. A manager who gives feedback well and never asks for it has taught the team that information only flows downward, and has cut the channel that would tell them when their own judgment is failing. Feedback is the cheapest corrective a manager has and the one most often put off. The conversation avoided in week two becomes the performance problem in quarter three and the resignation in quarter four, and the rest of the team quietly learns that the pattern is tolerated. The examinations test it because avoidance is the most recognisable mistake in management, and because at every scope the wrong option is usually the comfortable one.

1. Anchor feedback in a situation, a behaviour and an impact. Say what happened, what the person observably did and what it caused. Leave out what they are and why you think they did it: diagnosing character or intent produces defensiveness, and the diagnosis is usually wrong as well.
2. Early and small beats late and large. Frequent small corrections make feedback ordinary, so it stops arriving as a verdict. Feedback saved for the review cycle comes too late to change the quarter it describes and too heavy to be heard.
3. Clarity is the kindness. Softening a hard message until nobody could recognise it passes your discomfort to the person, who will find out the real message later and at higher cost. That is no compassion. Say the thing plainly and privately, with a path forward.
4. Praise follows the same rules. "Great job" is forgettable. Naming the specific behaviour and its effect makes it repeatable, and by contrast makes the manager's corrective feedback credible.
5. Ask for it, and be seen to act on it. Upward feedback only arrives where the reaction to the last piece was visibly good. The first time a manager punishes a messenger, the channel closes for everyone.
6. The truth degrades as it travels upward. Each layer between a leader and reality filters, softens and summarises. The more senior you are, the more deliberately you have to build channels that carry unfiltered signal, and the more your reaction to bad news decides whether you hear any again.

Feedback has more empirical work behind it than most management topics. The practitioner structures in common use rest on practice, though, and have not been tested in trials, so the unit grades them as practice.

Situation, behaviour, impact practice

Sloan R. Weitzel, Feedback That Works: How to Build and Deliver Your Message (2000) sets out the structure. Name the specific situation, describe the observable behaviour and state the impact it had. It works because it removes the two things that set off defensiveness, generalisation ("you always…") and diagnosed intent ("you don't care about…"), and because it gives the person something they can change. It gives structure and does nothing for nerve. A difficult message shaped as situation, behaviour and impact and then never delivered is still avoidance. For a pattern, as against a single moment, name the pattern and back it with several specific instances.

Radical candour practice

Kim Scott, Radical Candor: Be a Kick-Ass Boss Without Losing Your Humanity (2017) offers a two-axis model (care personally, challenge directly) and describes what goes wrong when either axis is missing. Challenge without care reads as aggression. Care without challenge produces what the book calls ruinous empathy: the well-liked manager whose people never learn where they stand. For this unit its value lies in diagnosing the second failure, which is the common one among managers who think of themselves as kind. The model describes a stance and offers no technique. It explains why to have the conversation; the structure above explains how.

The three triggers practice

Douglas Stone and Sheila Heen, Thanks for the Feedback: The Science and Art of Receiving Feedback Well (2014) models why feedback is hard to receive, where most writing looks at why it is hard to give. Truth triggers fire when the feedback seems wrong, relationship triggers when the problem is who it comes from, and identity triggers when it threatens the receiver's sense of who they are. The model helps in both directions. A manager giving feedback can anticipate which trigger a message will set off and shape it to suit. A manager receiving feedback can notice their own trigger firing and separate the reaction from the content. The book is a practitioner synthesis and is graded as one.

Feedback as a system practice

This is a recurring executive practice with no single name: the mechanisms that stand in for the unfiltered signal a leader no longer gets in person. They include skip-level conversations used for signal and kept away from decisions, engagement data that somebody reads and acts on, calibrated reviews of leaders, and norms that make dissent at the leadership table safe. None has a single provenance. All are widely adopted, and all fail in the same way. They exist on paper while the leader's reaction to bad news teaches everyone to stop using them.

The weekly small correction

Corrective feedback is given in the week it is earned, in private and in under five minutes. It is shaped as situation, behaviour and impact and followed by a question instead of a lecture. The habit matters more than any single instance. A team that gets small feedback often stops hearing feedback as a verdict, and when the difficult conversation comes it continues something familiar and does not feel like an ambush.

The difficult conversation, prepared

Before a hard conversation the manager writes down three things: the specific instances, the impact on the team or the work, and what they are asking for. If the instances cannot be written down, the conversation is not ready, because the feedback is still an impression and has not yet become an observation. The conversation opens with the behaviour, states the impact, asks for the person's account and closes with a named expectation and a date to look again. Opening with reassurance delays the message, and closing without the expectation leaves the message deniable, so it does neither.

Triangulating second-hand signal

Feedback about someone sometimes arrives through others, especially about a manager from their reports. The practice is to verify before acting, using skip-level conversations, direct observation, peer input and team-health indicators such as attrition and engagement. Taking raw hearsay to the person burns the sources and invites denial. Dismissing it because it has not come through a formal channel abandons the people who provided it. The feedback eventually given is the manager's own, based on evidence, and does not identify who spoke.

Asking for it

At a fixed interval the manager asks each report one concrete question about their own management, such as what to do more of, what to do less of, or one thing the manager does that gets in the report's way, and then reports back what they changed. Everything depends on the report-back. Asking without being seen to act teaches people that the question is ceremonial.

Protecting the messenger

When bad news arrives, the practice is the visible reaction: thank the person, ask what they need and act. The behaviour is the same whether the messenger is an engineer or a vice-president. Nothing a senior leader does has more effect on the quality of their information.

At team scope the object is an engineer's behaviour, and the manager gives feedback directly and often. At organisational scope the object becomes a manager's leadership, which nobody can observe by reading status reports. The manager of managers has to gather evidence at one remove and give feedback about how someone leads as well as what their team ships. At executive scope the truth has largely stopped arriving unprompted. The work then is to institutionalise what lower scopes do in person, and to deliver hard messages to powerful people without flinching and without humiliating them. The pattern is in How Judgment Scales; leading the managers who receive the feedback is PL-6 Leading through leaders.

• Gives frequent, small, specific corrections so that feedback stops feeling like a verdict. team
• Delivers a hard message plainly and respectfully, in private, with a path forward.
• Names the behaviour and its impact, never the person's character or presumed motive.
• Solicits upward feedback and visibly closes the loop on it.
• Praises specifically, so the praise is repeatable and the corrections are credible.
• Failure mode: saving feedback for the performance review. team
• Failure mode: softening a hard message until it is unrecognisable, then being surprised the behaviour continued.
• Failure mode: giving feedback while visibly frustrated, so the emotion drowns the content. team
• Failure mode: "you're careless", which targets the person instead of the behaviour.
• Gives managers feedback, shaped as situation, behaviour and impact, about how they lead as well as what their team delivered. org
• Triangulates second-hand signal before acting on it and protects the people who provided it. org
• Names the real gaps of a well-liked manager whose delivery has slipped; likeability is not the bar. org
• Failure mode: letting a manager's own account stand in for the reality of their team. org
• Failure mode: relaying skip-level input so identifiably that people stop talking. org
• Failure mode: treating second-hand signal as either a verdict or noise. org
• Builds mechanisms that surface bad news early and rewards the people who bring it. exec
• Delivers hard feedback to senior, powerful people with the same directness owed to a junior engineer. exec
• Failure mode: softening messages in proportion to the recipient's seniority. exec
• Failure mode: punishing messengers and then wondering why surprises keep arriving. exec
• Failure mode: mistaking the absence of complaints for organisational health. exec

Directness versus relationship. The fear is that a plain message will damage the working relationship. Usually the reverse holds. The fragile relationship is the one in which hard things cannot be said, and people trust a manager's praise because they know the same manager will criticise. Whether to deliver is settled; the judgment lies in how.

Timeliness versus preparation. Feedback given in the moment is the most useful and the most likely to be badly shaped, while carefully prepared feedback arrives late. Scale resolves it. Small corrections happen in the moment, difficult conversations are prepared, and a difficult conversation never waits beyond the week in which the instances were gathered.

Evidence versus urgency. Second-hand signal about a manager demands action and cannot be acted on as it stands. The harm continues while you verify; acting without verifying destroys sources and may be wrong. The discipline is to verify quickly and deliberately instead of choosing one over the other.

Candour versus safety. A team that often hears hard truths can become a team that braces for them. The counterweight is the care axis. Hard messages land within a relationship in which the manager has visibly invested in the person, praised specifically and asked for feedback in return.

Filtering versus fidelity. As information travels upward some summarising is necessary, and every bit of it costs fidelity. A leader who demands raw detail drowns in it, and one who accepts summaries ends up managed by them. The judgment is in deciding which signals must arrive unfiltered and building the channel for those.

Over two weeks, in three separate skip-level conversations, a manager of managers hears that one of their managers takes credit for the team's work in leadership meetings and is dismissive of people in private. That manager runs a high-output team and is well regarded by peers, and their own account of the team in weekly one-on-ones is uniformly positive. Nothing has reached a formal channel. Two of the three engineers who spoke asked not to be named.

Both reflexes are wrong. Taking the accounts to the manager directly, as received, exposes the engineers who trusted the skip-level, and invites a denial that cannot be answered without naming them. Dismissing the accounts because they are informal, or because the numbers are good, abandons the people who gave them and tells them the channel is decorative.

The judgment is to verify before acting, and to act as soon as verification allows. The manager of managers widens the evidence without exposing anyone. They attend the next two leadership meetings where the manager presents and listen to how the work is attributed. They look at the team's attrition and internal-transfer requests over the past year, ask the manager's peers an open question about working with the team, and bring forward their remaining skip-levels with that team by a week. Within a fortnight things they have seen for themselves confirm the picture. The manager presented as their own a design that the skip-levels attributed to two engineers, and two transfer requests from the team are pending.

The feedback given is then the manager of managers' own. In a private conversation they name two specific instances they witnessed and state the impact: the team's work is invisible to leadership, and its people are trying to leave. They ask for the manager's account and do not cite the skip-levels. They set an explicit expectation about attribution and about how the manager speaks to their people, with a date to revisit it against fresh evidence. They also tell the manager that this feedback concerns leadership, that it is separate from delivery, and that good delivery does not offset it.

Two things are deliberately left undone. The engineers who spoke hear only "I heard you, and I am acting on it" about what happened to their words, which closes the loop without the detail that would expose them. And the manager's output is not treated as a reason to wait. The version of this story that was avoided is the one in which the manager of managers waits for a formal complaint that never comes, because the people who would have made it have already left.

• PL-4 Performance management: where the conversation goes when feedback has not changed the behaviour, with expectations, support and honest consequences.
• PL-6 Leading through leaders: the managers who receive feedback about their leadership, and how their judgment is developed.
• CC-1 Psychological safety and productive conflict: the safety that decides whether feedback, especially upward, is given at all.
• CC-2 Coaching versus directing: coaching as the form feedback takes when the goal is judgment more than a specific behaviour.
• CC-5 Values, integrity, and ethical judgment under pressure: the costly right call when the person who needs the hard message is senior, valuable or feared.

1. Rewrite "you need to communicate better" as a situation, a behaviour and an impact.
   Answer

   "In Tuesday's design review, when the platform team raised the migration risk, you moved on without responding; they left believing the risk was accepted, and it was not." It gives a specific moment, an observable act and a named consequence, and says nothing about the person's character.
2. Why does softening a hard message until it is unrecognisable count as a failure of kindness rather than an excess of it?
   Answer

   The person will discover the real message later, from consequences, without having had the chance to change. The manager has spared themselves the discomfort and passed on the cost.
3. Three engineers tell you in skip-levels something damning about their manager. What do you do first, and why not confront the manager immediately? org
   Answer

   Verify through your own observation, peer input and team-health data, without exposing the sources. Confronting the manager with hearsay burns the people who trusted you and invites a denial you cannot answer. The feedback you eventually give has to be your own and based on evidence.
4. What is the single most consequential thing a senior leader does for the quality of the information that reaches them? exec
   Answer

   How they visibly react to bad news. Thanking the messenger and acting keeps the channel open for everyone watching; punishing a messenger once closes it for a long time.
5. Name the three triggers that make feedback hard to receive, and say how knowing them helps the giver.
   Answer

   Truth, relationship and identity triggers. The giver can anticipate which one a message will set off and shape it to suit: more evidence for a truth trigger, a different messenger or an explicit repair of the relationship for a relationship trigger, and a framing that separates the behaviour from the person for an identity trigger.
6. A well-liked manager's delivery has slipped for two quarters and their team is happy. What is the feedback, and what is not the bar? org
   Answer

   The feedback names the specific delivery gaps with evidence and sets a clear expectation and a review date. Likeability is not the bar. A happy team that does not ship is a leadership gap, and avoiding the conversation because the manager is liked is the failure this unit describes.
7. What replaces the unfiltered signal an executive no longer receives in person, and what makes each replacement fail? exec
   Answer

   Skip-level programmes, engagement data that somebody reads, calibrated leadership reviews and norms for dissent at the leadership table. Each fails in the same way: it exists on paper while the leader's reaction to the bad news it carries teaches people to stop using it.

• Sloan R. Weitzel, Feedback That Works: How to Build and Deliver Your Message (2000): the situation, behaviour, impact structure.
• Kim Scott, Radical Candor: Be a Kick-Ass Boss Without Losing Your Humanity (2017): care personally, challenge directly; ruinous empathy.
• Douglas Stone and Sheila Heen, Thanks for the Feedback: The Science and Art of Receiving Feedback Well (2014): the three triggers; receiving feedback well.