A peer 1:1 is wrapping up, or there are ten minutes left after a review, and the feedback comes out like this: "Great job on the PR, but you're slow." Your teammate says thanks, nods and goes back to work. A week later they remember the "great job." They remember that there was a "but." Neither of you can say what was supposed to change, and that includes the person who said it.
The habit here is small. Before the conversation, write one request in SBI form (situation, behavior, impact). At the end, ask your teammate to restate the request in one sentence. If they can't, the feedback didn't land. The fix is to rewrite the sentence, not to keep talking.
Why "great job, but…" never turns into action
That sentence has two problems: the sandwich and the adjective.
The sandwich is criticism wrapped in praise. The Center for Creative Leadership (CCL) is blunt about it in its guidance on difficult conversations: "Don't sandwich negative feedback between positive reinforcement." Praise that was earned gets its own moment. Jensen and Baumgartner, writing in HBR in 2021, explain why the format is so tempting: it "makes correcting people easier for you, but it rarely achieves the goal of helping someone improve." People with strong egos find the criticism goes down too easily. Self-critical people choke on it. Everyone else just finds the mixed message confusing. And the praise gets wasted, because after a few rounds "great job" turns into the noise that tells people a "but" is coming.
The adjective is the second problem. CCL separates attribution feedback, which labels the person ("you're undisciplined"), from impact feedback, which describes the effect without assuming a motive or assigning blame. "Slow" is attribution. It doesn't say slow compared to what. Your teammate can agree, disagree or take offense, but they can't repeat a behavior you never described.
Our read at DevDojo (editorial, not a study): "great job, but…" is a sandwich with an adjective for filling. Your teammate keeps the bread and the label. That doesn't prove every sandwich fails in every context. It only shows that this one, said the way it's usually said, leaves nothing anyone could repeat.
The SBI mold in one sentence
CCL's SBI model fits in one instruction: "clarify the situation, describe the behavior, then share the impact." The same page asks for a situation with a time and place ("last week" is too vague), a behavior stated as fact and not opinion, and an impact without judgment.
| Field | What goes in | What stays out |
|---|---|---|
| Situation | When and where: "on yesterday's 11 a.m. call" | "Last week," "lately" |
| Behavior | What could be seen or heard: "the comment arrived at 6 p.m." | "You're slow," "you're disorganized" |
| Impact | What happened to the work, or to you | "That's a lack of commitment" |
Two examples to illustrate. Neither comes from a real case. A late review:
> "On yesterday's 11 a.m. call, we agreed the review on the checkout item would be done by 3 p.m.; the comment only came in at 6. I couldn't start the handoff, and the deploy slipped to tomorrow."
A handoff that didn't happen:
> "At today's 11 a.m. handoff, you left without saying where item X stood or who owned it next. I couldn't pick it up, and the client went without an update until late afternoon."
Compare those with "Great job on the PR, but you're slow." There's no situation, no visible act and no impact. Just a compliment and a label.
After the SBI, you can add a request for next time. The 2021 HBR piece suggests closing with what to do going forward, and one sentence is enough: "Next time, send the review by the agreed time, or let me know by 2 p.m. if it's going to be late." SBI still has three parts. The request is just the last line.
Three things to watch during the conversation:
- Don't ask "why did you do that?" The same HBR article notes that a backward-looking why sounds like a scolding. Ask what can happen next time instead.
- Don't guess at intent. If your teammate wants to explain why, listen. You don't need to show up with a theory.
- Leave room for a reply. SBI is short precisely so the other person has time to talk.
And if the "great job" was true, it deserves its own sentence, at another moment, with the same structure. "Great work!" doesn't say what to repeat. "In Tuesday's presentation you opened with the migration risk, and the team decided in ten minutes" does.
The replay test: can your teammate restate the request?
This step is a DevDojo heuristic. It's not part of CCL's SBI. After you've said your sentence, ask:
> "Before we wrap up: in one sentence, what do you understand I'm asking for before our next conversation?"
- Pass: your teammate gives back the situation, the behavior and the request, even in their own words: "When we agree on a review time, I deliver by then or give you a heads-up before 2."
- Fail: all that comes back is the adjective ("you want me to be faster"), the praise, or a "got it" with nothing inside.
If it fails, rewrite the SBI once, usually by making the behavior more concrete, and ask again. When your teammate answers "be faster," the temptation is to bring up last week's delay as evidence. Now you have two complaints and zero requests.
Notice what the test is actually grading: your sentence, not your teammate. If all they can repeat is the adjective, that's because the adjective was all you gave them.
What this conversation is not
It's a peer-to-peer request about one recent, observable episode, this week. In the book Feedback That Works, CCL presents SBI as something you can use with a boss, a peer or a direct report. You don't need to be a manager to use it. You also don't become one by using it: there's no development plan and no quarterly follow-up.
It's not a career conversation either. The piece on the promotion cycle: "be more senior" with no written bar is about getting the written bar out of your manager, and the one on standup: the AI diff you cannot narrate in 60 seconds is about narrating a change against the Sprint Goal. This one is a peer asking for something the other person can repeat back.
If your teammate pulls it toward status ("is this going to count in my review?"), go back to the sentence: "This isn't about your performance review. It's just yesterday's code review and the timing next time." The list of everything that has bothered you since March also stays out. One episode only.
DevDojo's recommendation is to treat "great job, but…" as unfinished feedback until there's an SBI sentence and your teammate can restate the request. If they only repeat the adjective, rewrite once. If the label is still all that comes back after the rewrite, stop there and bring the revised sentence to the next 1:1, without opening a second round on the spot.
This week's next step
Today, pick the most recent observable episode you wish had gone differently and write the SBI: a situation with a time and place, one visible act, the impact and, if you want, a short request. One topic.
Use that sentence in your next 1:1 with that teammate, still this week. If you don't have 1:1s, the same sentence fits in ten minutes of peer review. Before you leave the conversation, check:
- your teammate restated the request in one sentence;
- you didn't open with "great job, but";
- you named a time and place;
- you described a visible act, not an adjective;
- you didn't add a second episode.
If all five pass, the feedback was delivered. If all your teammate remembers is that you thought they did a great job, you paid a compliment, not feedback.