Learn how to turn feedback into actionable insights using classification, semantic models & dashboards in Power BI to drive real CX improvements
Customer feedback is everywhere. And your qualitative data is some of the most insight rich data, where you can begin to truly understand what your customers need. They’re the source of your actionable insight — the thing that every CX professional needs to create meaningful change.
Actionable insights don’t just describe what’s going on, they give relevant teams the fuel and evidence they need to improve products and services. From our work with housing associations and other regulated sectors, we’ve found that every truly impactful insight has three components:
Our Head of Insights and Innovation, Steve Erdal explains it beautifully…
So if an insight is missing one of those three things, it’s not ready for the boardroom or for your frontline staff to deal with. For example:
❌ “Repairs is the most common theme in our feedback this quarter.”
➤ Okay… but what should we do with that?
✅ “Complaints about delayed repairs have increased by 38% since January — especially in sheltered schemes in the north of the borough. If we can prioritise contractor capacity in those areas, we could reduce complaint volumes by around a third.”
Those two statements would both be received differently in the boardroom. In this post, we’ll walk you through the framework we use to help organisations surface insights like that, and drive meaningful change using Power BI. It’s all about:
Before you can get to shiny dashboards or a semantic model, you need to wrangle the raw feedback — and that starts with classification. Classification is the process of grouping your customer feedback into meaningful, structured themes, and it’s where we see the biggest difference between organisations that are insight-led and those that are just reporting numbers.
Many teams start with manual tagging or inherited taxonomies that weren’t built with insight in mind — maybe they’re based on internal departments, process stages, or complaint categories.
The problem? These structures reflect how your organisation works — not how your customers think. Effectively classifying your data means investing in the right tools and tech that will allow you to classify your data and analyse your feedback effectively, we have an entire guide which explains all the potential feedback analysis options with pros and cons, read it here: 4 ways AI can save your from manual feedback analysis.
Once your classification process is in place — clear, structured, and actionable — you’re ready to build the logic that powers your dashboards and reporting. That’s where the semantic model comes in.
If you watched our full webinar on Understanding tenant feedback in Power BI, we introduced a metaphor about Dug’s Collar. Yes, this guy. Dug wears a voice collar that translates his chaotic dog thoughts into human speech. One moment he’s mid-sentence and shouts “Squirrel!” — but at least we know what’s going on in his head.
That’s what your semantic model does for your qualitative customer feedback. It takes the messy, unstructured comments and translates them into something Power BI can understand. It’s the logic layer between your data and your dashboards. Steph talks about it in our webinar, and we had some attendees with great film taste.
So what is a semantic model?
It’s the logic layer that turns your raw, classified feedback into metrics that business users can work with. It sits between your data and your visuals, and it handles things like:
Once your feedback is clearly classified and your semantic model is translating it into usable insights, you’ve got the engine running. But even the best engine needs a dashboard. This is where it all comes to life, when the insight actually lands in front of someone who can do something with it. It’s about making it obvious what matters, what’s changed, and what needs action. Too often, dashboards are treated as the end product. But in reality, they’re the bridge between analysis and action. Let’s explore how to build dashboards that not only look good, but drive change.
Even the best semantic model won’t matter if your dashboards confuse people. We see a lot of dashboards that are beautiful but unusable — because they’re built for analysts, not humans.
The key is to design visualisations that are:
We’ve worked with housing associations that have implemented some great BI dashboards. One housing association built a Power BI dashboard where frontline managers could click on a theme like “damp and mould” and instantly see scores, volumes, sentiment, and real resident quotes — all in one place. Another set up automated alerts when a theme spiked or fell below target, so action could be taken in the moment, not months later.
It’s not enough to have good insights — you need to make sure the right people see them and know what to do next. Here’s how:
The ultimate goal? Build an insights culture, where feedback isn’t just reviewed, it’s acted on, shared, and embedded into decision-making.
Power BI is a powerful tool, but it’s not magic. If your classification is messy, your semantic model isn’t set up correctly, or your dashboards are hard to use, your insights won’t land. By applying this three-step framework effectively — Classification → Semantic Model → Visualisation, you can move from reports that just describe problems to dashboards that drive real action.
