How to give feedback that actually changes things

Most feedback in startups is too vague to act on, too late to matter, or delivered in a way that produces defensiveness rather than change. The feedback that actually changes behaviour is specific, close to the event, focused on observable actions rather than character, and delivered in a context where the person can respond without losing face. That description sounds simple. The practice is one of the hardest skills in management — and one of the least practiced, because most founders are too busy, too conflict-averse, or too convinced that people should be able to figure it out themselves.

The vagueness problem.

The most common form of useless feedback is the general observation: "You need to be more proactive." "Your communication could be better." "I need you to take more ownership." These statements are not feedback. They are characterisations. They tell the recipient that you have formed a negative assessment but give them no information about what, specifically, to do differently.

The recipient of vague feedback does one of three things. They nod and attempt to guess what you actually meant, often guessing wrong. They feel criticised and become defensive, because a characterisation of their general approach feels like an attack on their competence or character. Or they agree with you in the moment and then do nothing different, because they have no actionable information to act on.

The test for whether feedback is specific enough to act on: can you point to a specific event, a specific moment, and a specific action that you want repeated or changed? "In the product review last Tuesday, you presented three options and said you did not have a recommendation. In future, I need you to come to those reviews with a recommendation, even if it is provisional." That is actionable. "You need to be more decisive" is not.

The timing problem.

Feedback is most effective within hours of the relevant event. It degrades rapidly as time passes. By the time a weekly one-on-one arrives, an event from four days ago has faded in both parties' memories, and the feedback risks being received as a verdict rather than a correction — because it has sat, unaddressed, for days.

The reason most founders delay feedback is not strategic. It is discomfort avoidance. Giving feedback in the moment requires accepting a brief moment of social awkwardness. Waiting until the scheduled one-on-one displaces that discomfort onto a future version of yourself who may be in a different emotional state and will have the same aversion to raising it. The feedback often never gets delivered at all.

The discipline to develop is separating the "I noticed something" observation from the "I need to address this fully" response. You do not need to have a complete position on what the person should do differently in order to note that you noticed something. "Hey, I wanted to flag something from the client call — do you have five minutes?" is not a big conversation. It is the micro-conversation that, practised consistently, prevents the accumulation of unaddressed feedback that eventually shows up as a performance review surprise or a termination that blindsides the person.

Character versus behaviour.

The distinction between character feedback and behaviour feedback is the most important structural concept in giving effective feedback, and the one most frequently violated.

Character feedback describes a person's traits or dispositions: "You are disorganised." "You do not communicate well." "You are not strategic." Behaviour feedback describes specific actions in specific situations: "The agenda for Monday's team meeting arrived the morning of the meeting, which meant the team had no time to prepare." "In the last three project updates, the blockers were mentioned but not escalated until after the deadline had passed." "In the board prep discussion, the financial projections were presented without a narrative explaining the assumptions."

Character feedback produces defensiveness because it challenges identity. If someone believes they are organised and you tell them they are not, they will argue with you — not because they are difficult, but because you have made a claim about who they are, and they have contradictory evidence. Behaviour feedback is harder to dispute because it refers to specific observable events, and it is easier to act on because it implies a specific change rather than a fundamental reorientation of personality.

The practical rule: never use adjectives in developmental feedback. "You were dismissive in that meeting" is a character statement. "When the engineer started explaining the technical constraints, you said 'I don't need the details' and moved on — that felt dismissive to the engineer and we lost the context we needed" is a behaviour statement. The second version is harder to write and harder to say, and it is significantly more likely to produce change.

The context that makes feedback land.

The same feedback delivered in different contexts produces different outcomes. Context is not just about privacy (though that matters — developmental feedback should almost never be delivered publicly). It is about the recipient's state of readiness to hear it.

Feedback lands when the recipient is not already activated by another stressor, when there is time to have a real conversation rather than a rushed exchange, when the relationship between the giver and receiver has established enough trust that the feedback is received as caring rather than attacking, and when the giver signals that the conversation is two-directional rather than a verdict being delivered.

The two-directional signal is the piece most frequently omitted. After describing what you observed and what effect it had, asking a genuine question — "What was going on for you in that moment?" "Is there something I'm missing about the situation?" — changes the dynamic from a correction to a conversation. It signals that you have not formed a final judgment, that you are open to information that might change your view, and that you are trying to understand rather than prosecute. Most people, given that signal, become dramatically more receptive.

The question has to be genuine. If you have already decided what you think and you are asking as a formality, the person will feel it. Genuine curiosity about what was happening for the other person is not a technique — it is a prerequisite for the conversation being useful rather than performative.

The specific situations founders face most often.

The situations where founders most commonly give ineffective feedback are: the repeated pattern (the same behaviour happening multiple times, with feedback becoming increasingly frustrated rather than increasingly specific), the senior hire who is not meeting expectations (where the founder is reluctant to give direct feedback because of the hire's seniority and the investment in bringing them on), and the team member who is technically strong but interpersonally difficult (where the feedback avoidance is most pronounced because the person's output is good).

The repeated pattern is addressed by naming the pattern explicitly rather than giving the same specific feedback again. "I want to talk about something I've noticed as a pattern, not just a one-off." Then describe the pattern in behavioural terms with specific instances. The escalation to pattern-level feedback signals that this is no longer about one incident.

The senior hire is addressed by giving the same feedback you would give a junior hire, with adjusted framing that acknowledges their experience: "I know you have done this kind of work many times before, so I am curious whether this was a deliberate choice or something that slipped through." This gives the person a face-saving exit while still naming the issue.

The technically strong but interpersonally difficult team member is the hardest case, because the feedback avoidance is rationalised as "not wanting to dampen their productivity." The correct framing: the interpersonal difficulty is not separate from their performance — it is limiting the performance of everyone around them, and addressing it is the highest-leverage thing you can do for the team.

• 01
  Point to a specific event, not a general pattern. "In Tuesday's client meeting" beats "you tend to." The specific instance is harder to dispute and easier to act on. Build to the pattern after you have established the specifics.
• 02
  Give feedback within 24 hours of the event. Delayed feedback becomes a verdict. Real-time feedback is a correction. If the discomfort of giving it promptly is why you delay, that discomfort is the thing to work on — not the timing.
• 03
  Use verbs, not adjectives. "You interrupted the client three times" is a behaviour. "You are dismissive" is a character assessment. Only one of these is specific enough to act on, and only one avoids triggering identity-level defensiveness.
• 04
  Ask a genuine question after the observation. "What was going on for you in that moment?" transforms a verdict into a conversation. It signals openness and gives the person a way to engage rather than defend. The question must be genuine, not rhetorical.
• 05
  Name the pattern when it has become a pattern. Repeated specific feedback that does not produce change needs to escalate to pattern-level feedback: "I want to talk about something I've noticed across a few situations." Pattern-level feedback signals a different kind of stakes without requiring a performance review.
• 06
  Do not protect technically strong people from feedback about interpersonal impact. The team's performance is limited by their ability to collaborate effectively. Withholding feedback from a high-output but difficult individual is not protecting their productivity — it is subsidising one person's comfort at the expense of everyone around them.

Frequently asked questions.