Effective Feedback Models

Most leaders don’t struggle to give feedback because they lack good intentions, they struggle because they lack the right frameworks. We say things like: 🗣 “This wasn’t good enough.” 🗣 “You need to speak up more.” 🗣 “That project could’ve been tighter.” But vague feedback isn’t helpful, it’s confusing. And often, it demoralizes more than it motivates. That’s why I love this visual from Rachel Turner (VC Talent Lab). It lays out four highly actionable, research-backed frameworks for giving better feedback: → The 3 Ps Model: Praise → Problem → Potential. Start by recognizing what worked. Then gently raise what didn’t. End with a suggestion for how things could improve. → The SBI Model: Situation → Behavior → Impact. This strips out judgment and makes feedback objective. Instead of “You’re too aggressive in meetings,” it becomes: “In yesterday’s meeting (Situation), you spoke over colleagues multiple times (Behavior), which made some feel unable to share (Impact).” → Harvard’s HEAR Framework: A powerful structure for disagreement. Hedge claims. Emphasize agreement. Acknowledge their point. Reframe to solutions. → General Feedback Tips: – Be timely. – Be specific. – Focus on behavior, not identity. – Reinforce the positive (and remember the 5:1 rule). Here’s what I tell senior FMCG leaders all the time: Good feedback builds performance. Great feedback builds culture. The best feedback builds trust, and that’s what retains your best people. So next time you hesitate before giving hard feedback? Remember this: → You’re not there to criticize. → You’re there to build capacity. Save this as your cheat sheet. Share it with your teams. Let’s make feedback a tool for growth, not fear. #Leadership #FMCG #TalentDevelopment #PerformanceCulture #FeedbackMatters #ExecutiveDevelop

The fastest way to make feedback ineffective? Turn it into a monologue. In RA/QA, I see this all the time.
Leaders really care about their teams. They want to be supportive. They want to give context.
So they turn a simple piece of feedback into a long explanation… and the core message gets lost. The intention is there, but the result is the opposite of what they hoped for.
People walk away unsure what actually needs to change. Long feedback creates confusion.
Short, factual, future-focused feedback creates improvement. We spoke about this on our recent episode of future leaders - link in comments. Here’s a structure that works in under 30 seconds - and actually lands: 1. Here’s what I saw Keep it factual. No assumptions. No emotion. “Hey, I noticed in today’s meeting you asked questions that were already covered in the preread sent yesterday.” This is just the observation.
Not a character judgement.
Not a story.
Just: here’s what happened. 2. Here’s why it matters People need to understand the impact of their behaviour. “It gave the impression you weren’t fully up to date, and we spent time repeating information instead of moving the discussion forward.” Impact gives the feedback meaning.
It connects the moment to the bigger picture - team time, decisions, trust, efficiency. 3. Here’s what to do next time Make the path forward simple and achievable. “Next time, please take a moment to go through the preread so we can use our time together more effectively.” This is what effective feedback sounds like.
Not dramatic. Not heavy. Not personal.
Just clarity. Short feedback doesn’t minimise the issue - it removes the confusion.
And when people know exactly what’s expected, they adjust faster and perform better. Most performance issues don’t require a 10-minute monologue.
They require a 30-second conversation delivered with intention. These insights are backed by the expert panel in our latest Future Leaders Session, Mastering Performance. You can watch the full session here: https://lnkd.in/dtEHMta2 And if you want to deepen your leadership capability in 2026, join the next Future Leaders Session: here > https://luma.com/oemwy59f See you there!

Tired of your LLM just repeating the same mistakes when retries fail? Simple retry strategies often just multiply costs without improving reliability when models fail in consistent ways. You've built validation for structured LLM outputs, but when validation fails and you retry the exact same prompt, you're essentially asking the model to guess differently. Without feedback about what went wrong, you're wasting compute and adding latency while hoping for random success. A smarter approach feeds errors back to the model, creating a self-correcting loop. Effective AI Engineering #13: Error Reinsertion for Smarter LLM Retries 👇 The Problem ❌ Many developers implement basic retry mechanisms that blindly repeat the same prompt after a failure: [Code example - see attached image] Why this approach falls short: - Wasteful Compute: Repeatedly sending the same prompt when validation fails just multiplies costs without improving chances of success. - Same Mistakes: LLMs tend to be consistent - if they misunderstand your requirements the first time, they'll likely make the same errors on retry. - Longer Latency: Users wait through multiple failed attempts with no adaptation strategy.Beyond Blind Repetition: Making Your LLM Retries Smarter with Error Feedback. - No Learning Loop: The model never receives feedback about what went wrong, missing the opportunity to improve. The Solution: Error Reinsertion for Adaptive Retries ✅ A better approach is to reinsert error information into subsequent retry attempts, giving the model context to improve its response: [Code example - see attached image] Why this approach works better: - Adaptive Learning: The model receives feedback about specific validation failures, allowing it to correct its mistakes. - Higher Success Rate: By feeding error context back to the model, retry attempts become increasingly likely to succeed. - Resource Efficiency: Instead of hoping for random variation, each retry has a higher probability of success, reducing overall attempt count. - Improved User Experience: Faster resolution of errors means less waiting for valid responses. The Takeaway Stop treating LLM retries as mere repetition and implement error reinsertion to create a feedback loop. By telling the model exactly what went wrong, you create a self-correcting system that improves with each attempt. This approach makes your AI applications more reliable while reducing unnecessary compute and latency.

If your feedback isn't changing behavior, you're not giving feedback—you're just complaining. After 25 years of coaching leaders through difficult conversations, I've learned that most feedback fails because it focuses on making the giver feel better rather than making the receiver better. Why most feedback doesn't work: ↳ It's delivered months after the fact ↳ It attacks personality instead of addressing behavior ↳ It assumes the person knows what to do differently ↳ It's given when emotions are high ↳ It lacks specific examples or clear direction The feedback framework that actually changes behavior: TIMING: Soon, not eventually. Give feedback within 48 hours when possible Don't save it all for annual reviews. Address issues while they're still relevant. INTENT: Lead with purpose and use statements like - "I'm sharing this because I want to see you succeed" or "This feedback comes from a place of support." Make your positive intent explicit. STRUCTURE: Use the SBI Model. ↳Situation: When and where it happened ↳Behavior: What you observed (facts, not interpretations) ↳Impact: The effect on results, relationships, or culture COLLABORATION: Solve together by using statements such as - ↳"What's your perspective on this?" ↳"What would help you succeed in this area?" ↳"How can I better support you moving forward?" Great feedback is a gift that keeps giving. When people trust your feedback, they seek it out. When they implement it successfully, they become advocates for your leadership. Your feedback skills significantly impact your leadership effectiveness. Coaching can help; let's chat. | Joshua Miller What's the best feedback tip/advice, and what made it effective? #executivecoaching #communication #leadership #performance

User Feedback Loops: the missing piece in AI success? AI is only as good as the data it learns from -- but what happens after deployment? Many businesses focus on building AI products but miss a critical step: ensuring their outputs continue to improve with real-world use. Without a structured feedback loop, AI risks stagnating, delivering outdated insights, or losing relevance quickly. Instead of treating AI as a one-and-done solution, companies need workflows that continuously refine and adapt based on actual usage. That means capturing how users interact with AI outputs, where it succeeds, and where it fails. At Human Managed, we’ve embedded real-time feedback loops into our products, allowing customers to rate and review AI-generated intelligence. Users can flag insights as: 🔘Irrelevant 🔘Inaccurate 🔘Not Useful 🔘Others Every input is fed back into our system to fine-tune recommendations, improve accuracy, and enhance relevance over time. This is more than a quality check -- it’s a competitive advantage. - for CEOs & Product Leaders: AI-powered services that evolve with user behavior create stickier, high-retention experiences. - for Data Leaders: Dynamic feedback loops ensure AI systems stay aligned with shifting business realities. - for Cybersecurity & Compliance Teams: User validation enhances AI-driven threat detection, reducing false positives and improving response accuracy. An AI model that never learns from its users is already outdated. The best AI isn’t just trained -- it continuously evolves.

Feedback fuels High Performance High performance doesn’t happen by chance—it requires a culture where feedback flows freely, consistently, and constructively. Without rich, ongoing, and actionable feedback, growth and performance stall. I was reminded of this recently while listening to Kim Scott’s excellent interview with Guy Raz on the Wisdom from the Top podcast (link in comments—totally worth a listen, especially for her distinction between managing Superstars vs Rockstars). Drawing from her time at Google with Sheryl Sandberg and her experience as a startup founder, Scott developed the Radical Candor framework—a simple yet powerful model for creating feedback-rich environments. At its heart, Radical Candor is about balancing a polarity: caring personally AND challenging directly. It helps us avoid feedback pitfalls like “Ruinous Empathy,” where we avoid difficult conversations to spare feelings, and “Obnoxious Aggression,” where bluntness is delivered without care. Why does this matter for high performance? Because when teams lack actionable feedback, issues fester, decisions are delayed, and growth slows. Feedback isn’t just a nice-to-have; it’s a critical driver of results and development. Yet, in my experience, many leaders struggle with this skill. Delivering feedback effectively is not innate—it’s learned and practiced. If you’re looking to build a feedback-rich environment, start with a simple exercise: Draw the Radical Candor 2x2 grid on a whiteboard and ask your team to identify your dominant feedback style as a group—not what you wish it was or think it should be, but how you truly experience the team’s feedback culture. Use this as a springboard for an honest and constructive discussion. - What changes would help us operate in the Radical Candor quadrant more consistently? - What agreements can we make to ensure feedback is both frequent and constructive? And, if the team is unable or hesitant to share their view on the feedback culture - well, that’s also data. Organizations that embrace a feedback-rich culture unlock their full potential. The real question isn’t whether feedback is hard—it’s whether you’re willing to accept the cost of not having it.

I was 11. Fresh off a football match that went… Terribly. I froze. Barely called for the ball. Kept my head down the whole game. On the drive home, my dad didn’t say too much. Just this: “You kept hiding in space, hoping they’d pass to you. But if your team can’t see you, they won’t use you.” That was it. One moment. One behaviour. Why it mattered. At the time, I thought he was just being kind. (And maybe a little smug.) Years later, in a Uni lecture, it hit me: He’d nailed one of the best feedback models out there...  Without ever hearing of it. Turns out, great feedback is clear, specific,  and science-backed. Here are 6 proven ways to give feedback that lands  without losing trust: 1. SBI Model → Situation: When and where → Behaviour: What you saw → Impact: Why it mattered (My dad’s comment? A textbook SBI.) 2. Radical Candour → Care personally. Challenge directly. → Miss either one, and trust doesn’t stand a chance. (Top-right quadrant or bust.) 3. FeedForward → High performers don’t want a post-mortem. → Give them the next step, not just a replay. 4. The 5:1 Ratio → 5 positive interactions for every 1 critique. → Feedback only sticks if the relationship can carry it. (Make deposits before you withdraw.) 5. Ask–Tell–Ask → Ask what they think. → Tell them what you saw. → Ask what they’ll try next. 6. CEDAR Model → Context. Examples. Diagnosis. Action. Review. → When the stakes are high, this one delivers clarity. Feedback isn’t about being brutally honest. It’s about being precise. So it actually lands. That’s what my dad got right. (Needless to say, I never did get much better at football.) ✅ It was short ✅ It was specific ✅ And it stuck Because when feedback is framed well, it doesn’t just  get heard. It gets remembered. And acted on. ♻️ Repost for your network (and look ridiculously clever while doing it.) Follow 👋 David Meade Keynote Speaker for science-backed strategies you can use this week.

What if feedback wasn’t something people feared… but something they valued? 💬✨ Too often, feedback feels like criticism. Vague. Personal. Poorly timed. But when done right, feedback becomes fuel for growth. 🚀 The difference? A structured, intentional approach. Here’s a powerful 5-step framework for giving feedback that actually works: 1️⃣ Prepare First Before you speak, pause. ✔️ What is your intention? ✔️ What specific examples can you reference? ✔️ Is this the right time and context? In the Critical 3 Academy Framework, this aligns with Intentional Thinking. We teach leaders to regulate emotion, clarify purpose, and communicate from strategy—not impulse. 2️⃣ Start With Connection Feedback without relationship feels like attack. ✔️ Ask permission ✔️ Acknowledge strengths ✔️ Show genuine care This reflects our Emotional Intelligence Toolkit: psychological safety increases receptivity. When people feel respected, they listen differently. 🧠💛 3️⃣ Be Specific, Not Vague ❌ “You’re always late.” ✅ “The last 3 meetings started 15 minutes late, which delayed the project timeline.” Specificity reduces defensiveness and increases clarity. Critical communicators replace generalizations with observable data. 4️⃣ Focus on Behavior, Not Character ❌ “You’re lazy.” ✅ “Task X wasn’t completed by the agreed deadline.” Behavior is changeable. Identity attacks are not. In the Critical 3 model, we emphasize separating facts from assumptions—a cornerstone of powerful decision-making and leadership credibility. 5️⃣ Collaborate on Next Steps Feedback is not a verdict. It’s a conversation. ✔️ Brainstorm solutions together ✔️ Set clear future goals ✔️ Define accountability When people co-create solutions, they own them. Ownership drives performance. 📈 Here’s the truth: Specific + Timely + Kind = Growth. Feedback is a gift 🎁—but only when delivered with skill. Inquiry-driven leaders ask: 1️⃣• What outcome do I want this conversation to create? 2️⃣• How can I strengthen the relationship while addressing the issue? 3️⃣• Am I correcting—or developing? Within the Critical 3 Academy Framework, powerful communication integrates: 🔹 Cognitive clarity (facts over emotion) 🔹 Emotional regulation (respond vs react) 🔹 Strategic collaboration (solution-focused dialogue) When these three intersect, feedback transforms culture. Imagine teams where conversations build capacity instead of eroding trust. Imagine leaders who elevate performance without diminishing people. That’s not idealistic. It’s trainable. Feedback isn’t about being right. It’s about helping others rise. 🌱 Follow and share if you found this helpful. Check out my featured post. Image credit: SketchedWisdom #LeadershipDevelopment #EmotionalIntelligence #CriticalThinking #CommunicationSkills #ProfessionalGrowth #ExecutiveCoaching #PsychologicalSafety #OrganizationalCulture #Critical3 #PerformanceLeadershi

Give feedback like you mean it. Don’t wrap it in a 💩 sandwich. We all need people to give us feedback it’s how we learn. The way it is given is crucial to how it is received. Wrapping “negative” feedback in praise ↳ ‘aka The Feedback Sandwich’ ↳ ‘aka The 💩 Sandwich” always starts from a good intentions. Here’s how it usually goes: 🥪 Nice comment (praise) 💩 Real feedback (criticism) 🥪 Another nice comment (encouragement) The goal? To make tough feedback more “palatable.” It sounds nice in theory.  Less defensive, more digestible. Your team, however, starts to dread hearing  “You’re doing great… BUT…” They stop trusting the praise.  They brace for the “but.” The Feedback Sandwich model fails because: ❌ It confuses the message (what exactly should I fix?) ❌ It dilutes real recognition ❌ It makes people suspicious of positive feedback ❌ It sidesteps discomfort instead of building trust So what should you do instead?  Give feedback like you mean it. Not wrapped. Not disguised. Just clear and human. 𝗛𝗲𝗿𝗲 𝗮𝗿𝗲 𝟯 𝘀𝗶𝗺𝗽𝗹𝗲 𝗺𝗼𝗱𝗲𝗹𝘀 𝘁𝗵𝗮𝘁 𝗵𝗲𝗹𝗽: ✅ 𝗖𝗢𝗜𝗡 𝗠𝗼𝗱𝗲𝗹 Use when giving constructive feedback with clear next steps: → 𝗖𝗼𝗻𝘁𝗲𝘅𝘁: What happened, where, when → 𝗢𝗯𝘀𝗲𝗿𝘃𝗮𝘁𝗶𝗼𝗻: What you noticed → 𝗜𝗺𝗽𝗮𝗰𝘁: Why it matters → 𝗡𝗲𝘅𝘁 𝗦𝘁𝗲𝗽𝘀: What should improve ✅ 𝗦𝗕𝗜 𝗠𝗼𝗱𝗲𝗹 Use when you want to name a behavior and its effect: → 𝗦𝗶𝘁𝘂𝗮𝘁𝗶𝗼𝗻: When and where it happened → 𝗕𝗲𝗵𝗮𝘃𝗶𝗼𝗿: What the person did → 𝗜𝗺𝗽𝗮𝗰𝘁: What result it had ✅𝗦𝗧𝗔𝗥 𝗠𝗲𝘁𝗵𝗼𝗱 Great for reviewing performance or reflecting after a task: → 𝗦𝗶𝘁𝘂𝗮𝘁𝗶𝗼𝗻: What was happening → 𝗧𝗮𝘀𝗸: What they were responsible for → 𝗔𝗰𝘁𝗶𝗼𝗻: What they did → 𝗥𝗲𝘀𝘂𝗹𝘁: What changed or improved Clarity builds trust. Not praise dressed as protection. 🍞 What’s your preferred model to give feedback? —————————— ♻️ Repost if you’re building a no-fluff feedback culture. 🔔 Follow Justin Hills for practical leadership insights.

Weekend Read Training small, specialized language models faces a fundamental tradeoff in how they learn from feedback. Reinforcement learning tells the model whether its final answer was right or wrong, but nothing about where it went astray along the way. The model gets a single bit of information after generating hundreds of tokens. Off-policy distillation provides detailed feedback on every token, but only on the teacher's examples. When the student makes an early mistake the teacher never would, it has no guidance for recovery. On-policy distillation bridges this gap. The student generates its own text, and the teacher scores every single token. Think of a chess coach watching YOUR games and grading each move, rather than just announcing if you won or showing you grandmaster games from positions you will never encounter. The efficiency gains are remarkable. On AIME'24 math problems, starting from the same 55% baseline (Qwen): • RL: 67.6% accuracy, 17,920 GPU hours • On-policy distillation: 74.4% accuracy, 1,800 GPU hours Kevin Lu at Thinking Machines replicated these findings, improving from 60% to 70% with 9-30x cost reduction compared to scaling supervised fine-tuning to 2 million examples. The method also enables effective personalization. When Qwen3-8B underwent domain-specific training, its IF-eval score dropped from 85% to 45% as it gained specialized knowledge but lost general capabilities. On-policy distillation recovered performance to 83% while simultaneously boosting internal QA performance from 18% to 41%. The mathematical foundation explains the effectiveness. RL provides O(1) bits of feedback per episode. Distillation provides O(N) bits per episode. For a 1000-token response, that represents 1000x more learning signal. The technique uses reverse KL divergence, making it mode-seeking and resistant to reward hacking. This creates a powerful training pattern: acquire new knowledge through domain-specific training, then recover general capabilities through distillation. More details : https://lnkd.in/ge7XeA2u