How to Give Feedback That Actually Changes Something

Two kinds of useless feedback, and most small businesses run on both.

Too vague: "That wasn't quite right." The person now knows you're unhappy and has no idea what to do differently, so they guess — usually by becoming more cautious, which is rarely what you wanted.

Too late: a list of frustrations saved up for a review. By then it's a complaint about something they can't remember clearly, and it lands as an ambush.

The fix is a structure and a much shorter delay.

Three parts

Observation. Impact. Request.

Observation: The client report went out on Thursday afternoon rather than Wednesday.

Impact: They review on Wednesday evenings, so it missed their meeting and I got a call about it Friday morning.

Request: Can we make Wednesday midday the internal deadline, so there's room if something slips?

Each part does a job:

• Observation is factual and checkable. Not "you've been slipping on deadlines" — something they could confirm or correct.
• Impact is why it matters. Without it you're stating a preference, and people reasonably deprioritise preferences.
• Request is what changes. Feedback without a request is a complaint.

Miss the observation and it's vague. Miss the impact and it sounds arbitrary. Miss the request and nothing changes.

Separate the observation from your interpretation

The most common way feedback goes wrong: leading with the story you've built.

| Interpretation | Observation | |---|---| | "You don't seem invested in this" | "The last two drafts came in without the data section" | | "You're being careless" | "Three of the five invoices had the wrong PO number" | | "You're not communicating" | "I found out the client had pushed the date from the client" |

The right-hand column can be discussed. The left-hand column can only be defended against — and you may well be wrong about the cause. Ask for it rather than assuming it.

Within days, not quarters

Feedback decays fast. A week later the person has moved on and the specifics have blurred; a quarter later you're asking them to litigate something neither of you remembers accurately.

The rule: if it's worth saying, say it within a few days. If it isn't worth a conversation within a few days, it probably isn't worth raising at a review either.

This also keeps each conversation small. Feedback given continuously is a series of minor adjustments. Feedback saved up becomes a confrontation, which is why people avoid giving it, which is why it gets saved up.

Drop the sandwich

Praise, criticism, praise. It reliably fails in one of two ways: the message gets cushioned until it's missed, or the person learns that compliments precede bad news and stops trusting praise entirely.

Both are worse than being direct.

Give praise on its own, when it's earned. And make it as specific as the criticism — "great work" is exactly as uninformative as "that wasn't quite right."

The way you flagged the discrepancy before sending saved us a correction email to the client. That's the instinct I want on everything that goes out.

That's usable. "Nice job" isn't.

Ask before you conclude

Half the time the explanation changes the feedback entirely.

The report went out Thursday rather than Wednesday — what happened at your end?

Sometimes the answer is that they were waiting on data from someone else, and the actual problem is upstream. Sometimes it's that nobody ever told them Wednesday mattered, which is a handoff defect rather than a performance one.

Leading with a question costs one sentence and prevents the most common error in feedback: correcting the wrong person.

When the same feedback lands three times

If you've given the same feedback three times, clearly, with impact and a request, and nothing has changed — the delivery isn't the problem.

At that point it's one of three things, and they need different conversations:

• Capability. The person can't currently do this. Training, a smaller scope, or a different role.
• Expectations. They don't agree it matters. That's worth surfacing directly, because you may be wrong.
• Fit. The role isn't right for them.

Repeating feedback a fourth time is avoidance dressed as management. And if it's heading toward a change in someone's employment, the process is governed by employment law that varies by jurisdiction — take advice before acting rather than improvising.

The mistakes

1. Vague. "Not quite right" gives nothing to act on.
2. Saved for a review. Too late to be useful, too much to absorb.
3. The sandwich. Message missed, or praise devalued.
4. Interpretation instead of observation. Invites defence rather than discussion.
5. No request. A complaint, not feedback.
6. Praise as vague as the criticism. People can't repeat what you didn't name.

What to do next

Think of the last thing you noticed and didn't say. Write it as observation, impact, request — three sentences — and have the conversation this week. If it turns out you can't fill in the impact line, that's useful too: it probably wasn't worth raising.

The regular version of this is a proper one-on-one, which is where most feedback should happen before it ever needs its own meeting.