From Data to Insights: Using Themes to Get A Middle Step Overview Of Your Data


girl walking three steps. Lower step saying data, middle step saying with question marks implying themes, third top step saying insights

Data analysis is that element of qualitative research where human brilliance truly gets to shine. It’s not just data collection, but a process of looking at data from a series of different perspectives, ultimately leading to insights generation (our favourite part). 

Yet, the journey from data to insight is an arduous, lengthy and challenging one. As you probably know, you’re going to need multiple takes to group up, scan and reshuffle data before you see patterns emerging, and sometimes that spark of brilliance might still elude you as you trot along your path to inspiration. 

Better still, sometimes you have gathered up a lot of information, but you would like to get a fresh, unbiased and sweeping opinion of your data set, without the hassle of having someone going through it in depth.

In short, what you need is a middle step that enables you to grasp tons of data at a glance, to sort it according to emerging patterns and to spot undiscovered ones, with the ultimate goal of speeding insights generation even further.

That’s what inspired us to create Themes, a brand new feature that will empower you to sweep through your data at a glance and to collect it automatically into thematic “buckets”. To get the full picture of our new feature, we went straight to its maker, Jökull Snæbjarnarson, CTO at Sonar, and asked him a few questions. Enjoy!



What kicked off the production of this feature? Did you feel there was a specific need for it?

I think the Themes feature is coming more from our platform users than from us. We kicked off production when we aggregated our users’ feedback. Here, we noticed they were asking for a middle step in their data analysis, both by telling us explicitly and by observing how they compiled and analysed data right before generating insights. 

Indeed, sometimes insights were created just as data repositories, as a way to gather quotes and observations containing similar themes. With our new feature, users can now create a middle step where they can polish data and convert it to insights in a much clearer and cleaner way. 

However, that’s just part of the picture. As we were looking at this trend, a few questions popped up. “What more can we achieve with this space? How can we get users one step closer to insights? Lastly, what can we do with data other than providing qualitative nuggets?” 

Therefore, we thought of this feature also as a way for users to create more insights, by having our AI give its own spin to data patterns, potentially revealing trends users wouldn’t be able to see by going through their data task by task.


Do you think having a middle step is going to help researchers speed up their studies?

Yeah, I hope so. The common thread on any feature we make is to “get better insights, faster“. And in this feature, there’s definitely an element of that. By having the initial themes automatically generated out of your data set, you will gain a massive head start in your analysis. In a sense, I see this feature as an evolution in our automation of data processing.

Looking at our platform, where creating automatic quotes minimises the need to go to look at all of the transcripts, adding the themes minimises the need to go to the quotes and observation page. 

Furthermore, you now have a chance to create thematic buckets where you can store your users’ quotes and observations you would like to look at better later, or half-finished insights that require additional refinement.

However, the main goal of this feature is to provide you with an aerial overview of your data, a rapid summary that gets automatically generated for them by tracing patterns in their data set, which you can then enrich with your own take even before starting analysing data.


Aerial overview – could you elaborate on this any further?

With our new feature, you can freely zoom in on individual quotes and observations, or zoom out for a sweeping overview of what’s in your data set at a glance.

If you were to think in Google Maps terms, you could say this feature allows you to move from Map view to the Street view and back. I think that’s something that has been missing from the product.

This is especially useful if you want to onboard teammates into your projects. Without a strong footing into your study, they have a hard time probably knowing what users are talking about – unless they go through all the data. 

But with this feature, at least part of our goal, they can get an overview of what’s going on inside of your study, without you having to look at every single quote or every single transcript. Kinda like Netflix.


Like Netflix?

Yeah! If Netflix worked just as a repository with episodes and movies showing up in order of upload date, it would take you tens or more scrolls to find something you’re interested in. However, since Netflix’s algorithm creates genres – or themes –  based on what you are watching, as soon as you log in you get a home page with thematic carousels, filled with movies or series that might be interesting to you. From there, you can dive deeper into a specific genre and figure out what you want to see.

Likewise, our Themes feature gives you at a glance thematic buckets of quotes and observations: once you find an interesting theme, you can click on and read them one by one. 


Getting into the nitty-gritty of the feature, how does the AI process the information to cluster it into themes?

Our platform makes sure you get themes of a certain size – 5 quotes minimum per theme – but you want to have similar subjects within each theme as well. If users from your study are saying the same or conceptually similar things, then they go to the same theme. For instance, users might comment about the high price of a product and use sentences like “the price is too high”, or “my pay is X and my expenses are Y, so I cannot afford it”: these would be clustered into the same theme. 


And how does the AI assign a title to each theme?

Each theme is titled based on what makes it unique from the others. In other words, the title tries to explain what’s inside of the theme. However, there is no guarantee this will work every time. Therefore, to bolster up each theme, the first quote that you see in each theme is the most relevant quote for that theme. That said, platform users can change the title – or any other element of the theme for that matter – at any time.

Explore Hidden Gems from your Users – with Themes 

Design Pain Points

Triangulation in UXR: Why Mixing Multiple Research Methods Leads to Deeper Insights

AI Customer research

How can AI help optimize and simplify user research? Takeaways from our webinar

Collect Pain Points

Ask the Expert – How to Write Mind-blowing Survey Screening Questions


Get the latest news from Sonar


About Us
Cookie Policy
Privacy Policy


Langebrogade 4
1411 Copenhagen

Phone: +45 7734 8685
VAT: 33510608




Get in touch

Book a demo
Write to us