Unified Feedback Systems

I don’t know what all those gauges and readouts on an airplane dashboard mean, but I do know that I want the pilots flying the aircraft to see them. Otherwise, they’d be flying around the globe pressing buttons and throwing switches on hunches and guesses. It’s the same with change activation. If a business wants its initiatives to actually, you know, work, they need the gauges and readouts of change: two-way feedback loops. Too many transformation strategies stall mid-air because they're missing one critical piece: live feedback from the ground. 🚫 Not the kind that comes 90 days later in a spreadsheet from HR. 🚫 Not the kind that’s missing in a thousand unanswered surveys. 🚫 Not the kind that's too late, showing up in exit interviews from disgruntled employees already moving on to greener pastures. I’m talking about real, instant, interactive, informal feedback. The kind that can be used to course-correct in real time. I call this the “Triple I” strategy: Instant  Interactive  Informal Here's the thing about feedback: 🧭 It’s a compass. It surfaces what people are thinking right now — what they’re confused about, excited by, or flat-out resisting. 📈 It’s a growth engine. It helps teams learn faster and build smarter next time. If they already know that job security is a major concern for one group, why go through the pain of rediscovering that from scratch during the next initiative? 🧠 It’s organizational memory. A well-run feedback system captures insights that can be used again and again. No need to keep asking the same questions if the answers have already been documented. But here’s the challenge: Most companies don’t have the time, tools, or energy to conduct 1:1s, focus groups, and in-person interviews across tens of thousands of people. And survey fatigue is real. You can only send so many Surveymonkey forms before people start auto-clicking “neutral.” Instead, tap into an activity people already do several times every day: interacting with content. When change comms or capability building initiatives are embedded into a change activation platform with built-in interactive functionality, something magical is unlocked: ✅ Questions get asked  ✅ Concerns are shared  ✅ Colleagues respond to each other  ✅ Change champions emerge organically  ✅ A real-time pulse on what is and isn't resonating emerges  Even better? The data is captured automatically. Comment data becomes reports visualized in-platform with sentiment analysis layered on top. Visibility into what’s trending by audience, location, and job level — across the entire organization — without running a single survey. Access to 24/7, large-scale feedback *that doesn’t feel like feedback.* No forms. No follow-ups. Just natural interaction with change content and powerful data to guide your next move. That’s the kind of loop that fuels real agility and speed. Because strategy without feedback isn’t agile - it’s flying blind. 

I get to talk to dozens of B2B product teams every week. One thing stands out: The best teams aren’t collecting more feedback. They’re connecting the dots faster. Average product teams keep asking “what do users want?” But the real question is: which users, and why? You’re not short on input from users. You’re buried in it. It’s scattered across: • Slack, Gong, Intercom, Zendesk, ... (feedback) • Salesforce, HubSpot, Snowflake (customer context) • Linear, Jira, Notion, GitHub (delivery plans) And yet — when it’s time to prioritize or give CS a status update — most teams are stuck stitching it all together manually. (good ol' spreadsheet entered the chat 👋 ) Here's what the best teams unlock with a solid feedback system 1/ You prioritize what matters Not just “a user asked for X.” But who asked (churn risk? high ARR?), how many asked, and what’s already happening in delivery. 2/ Everyone stays informed — automatically When feedback is linked to both customer data and delivery progress, Sales, CS, and Product teams get the right updates without having to ask. 3/ You find patterns, not noise When feedback is structured by segment, usage, or geography, the product team can be proactive, not reactive 4/ You tie product work to outcomes You can show why something was built, who it helped, and how it impacted expansion or retention. It’s not about collecting more feedback. It’s about making the feedback you already have useful. What’s your team doing today to make that happen?

Don’t let broken feedback systems hold your team back. Here’s Christina’s proven approach to feedback systems that work: I’ve spent years refining feedback processes in workplaces. Here’s what I’ve learned makes them effective: 1. Structured Feedback Loops → Design systems for regular input, not once a year. → Focus on specific actions, not vague suggestions. → Ensure feedback is clear, concise, and consistent. 2. Two-Way Communication → Create safe spaces for honest conversations. → Encourage employees to share upward, not just downward. → Build trust through transparency and active listening. 3. Action-Oriented Insights → Turn feedback into measurable improvements. → Align feedback with team and business goals. → Celebrate progress, not just perfection. 4.Follow-Up Systems → Schedule check-ins to track progress. → Keep the conversation ongoing, not one-and-done. → Close the loop by showing how feedback drives change. Most leaders overcomplicate this. - Start simple. - Listen actively. - Act consistently. That’s it. Want a custom feedback system that fits your team? Let’s build it together.

In most manufacturing environments, there is little or no real-time feedback once the production schedule is planned. Consequently, the production schedule remains purely theoretical. When downtime occurs, requiring shifts in the schedule, there is no immediate way to determine if the next order should start, say, an hour or two later. Real-time updates would solve this challenge: → Visibility into Upcoming Changeovers → Real-Time Updates to Estimated End Times → Consolidated View of Order Status and Progress → Monitoring Planned Activities Against Actual Progress Because ERP systems are primarily transactional, they lack the dynamic response needed to identify and mitigate bottlenecks in the manufacturing process. A UNS closes that gap with an event-driven data model that feeds ERP work orders directly into production control, enabling on-the-fly execution adjustments. How it Works in Practice? When a new order is created in the ERP, the system publishes an Order Creation event to the UNS. If the ERP cannot publish natively via MQTT, a proxy (such as a custom microservice or an IIoT platform) converts the order transaction into an MQTT event. This event typically includes details such as the order ID, product specifications, quantity, due date, and any relevant Bill of Materials (BOM) information. Depending on site preferences and performance requirements, organizations can choose how and when to publish order data: 1. 𝐑𝐞𝐚𝐥-𝐓𝐢𝐦𝐞 𝐔𝐩𝐝𝐚𝐭𝐞𝐬: Publish each new order the instant it is created. 2. 𝐂𝐡𝐚𝐧𝐠𝐞-𝐎𝐧𝐥𝐲 𝐔𝐩𝐝𝐚𝐭𝐞𝐬: Publish changes to existing orders as they occur. 3. 𝐁𝐚𝐭𝐜𝐡 𝐔𝐩𝐝𝐚𝐭𝐞𝐬: Periodically publish the entire order list. Once the order details are available in the UNS, relevant status and progress updates can be automatically merged with operational data streams from other production systems. This unified data model allows stakeholders to see, at a glance: → Which production lines are performing smoothly and which need attention? → How current performance compares to planned timelines? → Upcoming changeovers and the preparation required → Orders that are nearing completion (e.g., over 65% done) and the changeovers that follow Consolidating this information in real-time, the UNS makes it easier for teams to adjust schedules, coordinate staffing, and manage resources based on the most up-to-date production data.

Day 19 / 30 of RecSys 2025 - Unified Survey Modeling to Limit Negative User Experiences in Recommendation Systems This work from TikTok engineers addresses a classic recommender system problem: the over-optimization on positive feedback signals due to the sparsity and bias of negative signals like dislikes or reports. Relying solely on these explicit negative signals fails to capture the full spectrum of user dissatisfaction. The authors propose a system that uses in-feed surveys to create a richer, more direct, and less biased feedback channel, specifically targeting user satisfaction and perceptions of content appropriateness. This provides a more robust dataset for explicitly modeling and mitigating negative user experiences. The core of their technical approach is a personalized survey model built upon the Hierarchy of Multi-Gate Experts (HoME) framework. This is a multi-head architecture where learning tasks are logically grouped. For instance, survey_like and survey_notlike form a "Satisfaction" group, while survey_inappropriate and its sub-reasons (e.g., violent, hateful) form an "Ecosystem" group. A shared group captures common information across all tasks. This structure allows the model to learn both specialized intra-group relationships and generalized cross-group information, improving both specificity and model generalization. The system uses dedicated expert towers for each head and group, with final outputs being a concatenation from the specific expert and a shared expert.

Machine data is only part of the equation for digital operations; don't forget about the people, materials, and flow! I recently started experimenting with Tulip and AWS IoT SiteWise to better contextualize machine data, operator feedback, and context, as well as other operational data sources such as planning and scheduling. It's not enough to know the CNC Mill has a spindle speed of 4,000 RPM... The typical set of follow-up questions from most plant managers include: * Is it supposed to be running? * What work order is it running? * Who is running the machine? * Do they have what they need? To answer these questions, it's vital to contextualize machine data from the PLC alongside operator input and systems data (ERP, PLM, etc.). Otherwise, you only get half the picture of the state of operations. Tulip Integration: Connector Function: I experimented with using a Tulip Connector Function to write data to IoT SiteWise to add the operator context. I was also able to use the same Connector Function to query recent metrics from SiteWise. Tables API: For alerting, I was able to use a Lambda function to write data to Tulip via the Tulip Tables API. This data could include alerts on maintenance or quality as well as insights for the shop floor supervisor. Future Considerations: Adding more predictive analytics to this simple stack could build upon this feedback loop. Tools such as TwinThread could add to the compelling value proposition. Cost Notes: I assumed 100 machines per plant sending 5-10 data points per minute (More frequent data would be processed at the edge). * The cost for API Gateway and Lambda is pretty negligible. * The IoT SiteWise cost comes to about $1 - 1.5k per month but can vary based on data transformation and integration with other services. Overall, closed-loop feedback systems like this could really enable true OEE... and by that I mean Overall Employee Engagement and Overall Enterprise Effectiveness. ;) Let me know what you think and how you've explored closed-loop feedback systems in manufacturing. If there's interest, I can publish architecture details and the Tulip Connector details too!

Most website projects fail because too many voices create chaos instead of clarity. After managing dozens of e-commerce builds across fashion, beauty, and lifestyle brands, we've identified the single biggest factor that separates smooth launches from complete headaches. It's not budget constraints or technical complexity - it's feedback management. Here's what typically happens: a brand kicks off their website project with enthusiasm. The marketing manager has opinions about the homepage. The founder wants to weigh in on product pages. The operations director questions the checkout flow. Before you know it, the design team is drowning in contradictory feedback, conflicting priorities, and endless revision cycles. We learned this lesson the hard way early on at Studio Almond. A premium skincare brand came to us for a complete Shopify rebuild. Three weeks into the project, we were receiving feedback from six different stakeholders. The marketing team wanted bold, attention-grabbing visuals. The founder preferred minimal, clinical aesthetics. The customer service manager insisted on prominent FAQ sections everywhere. The result? Paralysis. Delays. Frustration on all sides. That's when we implemented our dual-lead system. Now, before any project begins, we ask clients to assign two key roles: a project lead and a creative lead. They can be the same person, but the creative lead has one critical responsibility - ensuring all feedback from the client side is consolidated, aligned, and cohesive. This person becomes the single point of truth for creative decisions. The creative lead collects input from all stakeholders, synthesises it internally, and presents unified feedback to our team. That same skincare brand? After restructuring with a clear creative lead, we delivered their new site two weeks ahead of schedule. The most successful launches we've managed all have one thing in common: clear creative leadership that transforms multiple voices into a single, powerful brand narrative. What's your feedback structure for creative projects?

Your CEO just asked what customers think. 5 teams are about to give 5 different answers. Your customer feedback lives in 5 different places. Surveys. Reviews on your site. Social media mentions. Support tickets. Call center transcripts. Each one is managed by a different team. Each one lives in a different system. Each one gets analysed separately - if it gets analysed at all. And then your CEO asks: "What are customers saying about our new product?" Nobody has the full answer. Because nobody's looking at all the sources together. You need a single brain. Something that connects to all the different channels, pulls all the data into one place, and makes sense of it as a unified whole. Not five separate analyses. One analysis across everything. This is what we built at Chattermill. Because we realised early on that fragmented feedback creates a fragmented picture of your customer. And that fragmented understanding leads to bad decisions. The companies that are winning with experience-led growth aren’t the ones with the most sophisticated analysis of their survey data. They're the ones who can see the complete picture of what customers are saying across every single channel. And then turn that analysis into meaningful action.

A core challenge in product discovery is unifying feedback. Here’s a breakdown of how you can approach it in a super simple way with automation and using your existing tools/channels: 👉 The Goal: stream all feedback into a single, dedicated channel. Lets say a Slack channel. 👉 The Setup: identify your key feedback sources, this could be a tool like Pylon for internal requests, HubSpot for sales feedback, a feedback form and ad-hoc insights, or Discord for community conversations. 👉 Use an automation tool like n8n as your glue. For each source, set up a simple workflow: - From Pylon: A webhook listens for new feature requests and sends a formatted message to Slack. - From HubSpot: The HubSpot trigger node watches for specific properties on deals or contacts and routes that feedback to the same channel. - From a Feedback Form/NPS: A webhook listens for new submissions, pulling out the score and comment. - From Discord: A Discord trigger node listens to a dedicated feedback channel, forwarding new messages. What you get: 👍 A unified view: all feedback, regardless of its source, appears in a single stream. This removes the need to constantly check multiple platforms. 👍 Contextual awareness: a standardized message format provides immediate context, allowing your team to quickly understand where the feedback is coming from. 👍 Actionability: once in Slack, your team can discuss, react, and, critically, link this feedback directly to Jira. This turns raw data into a clear input for your product backlog. This process isn't about finding the perfect tool; it's about building a system that brings context and clarity to the messy but essential world of product discovery. Go try it - those tools have free trials and are super easy to use. Don't wait until you lose more user feedback due to chaos of tools. #ProductManagement #ProductDiscovery #WorkflowAutomation #BuildingProcess

A leading telco used to  spend weeks preparing Voice of Customer reports.  They were outdated before anyone even read them. So they made a data strategy change: 🔷Unify all unstructured customer text (surveys, app reviews, calls, chats) in a database 🔷Thematically analyze and tag each record with consistent and precise themes 🔷Build dashboards that show themes that are positive and negative drivers Suddenly, frontline managers and executives had real-time access to what customers were saying. They could filter by region, product, or call reason and act faster. Within months, they: ✅ reduced reporting time by over 80%, ✅ spotted recurring issues sooner, and ✅ saw internal engagement with VoC insights rise dramatically! Moral of the story: unifying feedback in the data warehouse speeds action & creates real-time feedback loop that makes stakeholders sit up and take notice.