I used the sandwich approach when giving constructive feedback to engineers. I'd start with the good things, discuss what needed to change, then finish with something good again. I wanted the conversation to end on a positive note.
I later moved away from that approach. My concern was that the person could leave unsure how seriously to take the problem. Was everything fine overall? Was there something they needed to change? Finishing on a good note didn't answer those questions.
Why I stopped using the feedback sandwich
Kim Scott criticizes the feedback sandwich in Radical Candor, arguing that wrapping criticism in praise can obscure the message.
My objection is to making it the default structure, regardless of what you actually need to say.
Good work deserves recognition too. You can explain what someone did well and why it mattered without using it to introduce a problem. A discussion about something that needs to change doesn't require an unrelated compliment at the end.
Give feedback while the person can use it
I gave feedback between formal one-on-ones. Sometimes that meant a private conversation at the office; sometimes it was over coffee on a terrace. We didn't need to wait for a calendar invitation to discuss something that needed attention.
I wanted people to know where they stood and what they needed to work on. By the time we reached a performance appraisal, the feedback should already have been part of our conversations.
Waiting carries a cost: the person continues without knowing what you think needs changing. Then you ask them to account for work they had no opportunity to adjust.
Continuous feedback can also become excessive. My advice is to discuss something when it matters and there is enough privacy and time to do it properly. You don't need to interrupt someone every time a thought occurs to you.
Explain what you mean and why it matters
I tried to be specific about what had happened and explain why it mattered.
If you think something needs to change, you should be able to explain what you're referring to and why it matters. Otherwise, you're asking the person to work out what is in your head.
The Center for Creative Leadership's guidance on feedback makes a useful distinction between describing observable behavior and making judgments about a person. It also recommends asking about their intentions rather than assuming you know them.
For an engineering manager, I'd add one check: can you explain whether you're discussing an expectation, a consequence of the work, or your own preference? Those deserve different conversations. Having the manager title doesn't make every preference a requirement.
Before you talk, separate what you observed from what you inferred. If your argument depends on what you assume the person thinks or feels, you have something to ask about. You don't yet have an established fact.
Give the person room to explain what you haven't seen. A discussion can reveal an unclear expectation, a constraint you missed or a disagreement worth taking seriously. You may still decide that something needs to change, but you should be able to explain that decision after hearing their perspective.
Adapt the conversation to the person
I adapted how I gave feedback because people and personalities differ. The same delivery doesn't suit everyone.
You can adjust your pace, give more context or leave room for questions while keeping the point clear.
Pay attention to where the conversation gets stuck. More explanation may help if the expectation is unclear. If the person needs time to consider what you've said, agree on when to return to it. You can ask what would help them discuss the issue rather than guessing from your impression of their personality.
An easy conversation that leaves both of you with different ideas about what happens next hasn't helped much.
Agree on what should happen next
Before you finish, make sure you both understand what is expected to change. Describe it in terms of the work or behavior, and ask how the person sees the next step. A nod is easy to mistake for agreement when you're the one with authority.
Discuss any support they need from you and when you'll revisit the subject. The timing should fit the change. If you're discussing something that should happen differently the next time a situation arises, months of silence won't help. A skill that takes practice will need progress checks over time. Both people should know what you'll look for.
Follow what happens after the conversation
I checked understanding both during the discussion and through the work afterwards. I saw changes, returned to the subject and kept following progress. Feedback was something we could revisit and adjust.
If nothing changes, I'd look again at the original feedback before assuming the person ignored it. Was the expectation clear? Did you understand their perspective? Was the change realistic? You may need to explain more clearly, change your judgment or address something that is making the work difficult.
Following up also gives you a chance to recognize progress. If you've asked someone to change something, pay attention when they do.
Ask for feedback yourself
I also asked for feedback and tried to stay involved when receiving it. I considered that important for my own development as a manager.
I asked for anonymous feedback too. I wanted people to feel able to be honest and direct, including about things they might not tell me face to face.
If you're unclear, inconsistent or making something harder, people need a way to tell you. Asking is a start; your reaction affects whether they'll want to do it again.
My advice is to ask about a part of your work the person can actually comment on: how you explain priorities, run a meeting or make yourself available when they need a decision. A specific subject gives them somewhere to start. In a direct conversation, ask for an example if you don't understand the feedback.
You can take time to consider what you've heard. You don't have to promise a change during the conversation. Let people know what you decided, including your reasons if you disagree or can't act on the suggestion. Otherwise, they have no way to know whether you considered it at all.
This is part of the broader change I wrote about in moving from engineer to engineering manager: your own work is no longer the only thing you have to pay attention to.