Following the launch of our new Topics engine Audiem co-founder Ian Ellison explains how this new capability can transform your workplace insight.
“Our office isn’t agile. Can you help us make it work?”
This was a common question that James and I used to hear when we ran our consultancy and were called in to support a physical workspace change. But after spending some time working with a business, you won’t be surprised to hear, that a far more complicated interconnected system would typically emerge. It would always involve the people, their physical workspaces, and the tools and technology they were using. And all of this – not just the ‘agile office’, as initially suspected – worked in interrelated ways to influence personal and collective performance, for better or for worse.
So when we launched Audiem we were determined to enable workplace teams to unravel that interconnected system. Step one was mapping out that system that eventually became the Workplace Mix™, our academically validated framework that represents workplace as sub-themes of people, spaces, technology, and their business impacts.
To inform that framework we developed technology that could capture the corporate conversation in the words of employees, not in tick boxes of agreement. Since then, we’ve been able to process hundreds of thousands of employee free text responses, with deep insights navigable through the lens of the Workplace Mix™. But we had a hunch there were even more insights to draw from this vast collection of messy data.
By using our Workplace Mix™ framework, Audiem’s AI analyses text data to identify which comments fit into specific themes and their related sub-themes. It even spots the links between them – where one thing is influencing or causing another thing, according to the evidence. To use a research term, our AI is deductively analysing text data using our proprietary Workplace Mix™, checking each comment for fit within the framework.
We also spotted another way to explore workplace experience data. In research terms, the opposite of deductive is inductive. This is where a researcher seeks emergent findings, with no preconceived framework or agenda. Consequently, it can sometimes unlock the most revealing and unexpected insights. Which meant we needed to identify emergent topics in the text data too.
This is why we’ve developed our Topics engine. A way of making sense of a data set with the topics of conversation specific to the workforce in question. Combining this with the Workplace Mix™ can result in a powerful cocktail of workplace insight.
Let’s take a look how.
Topics, a quick overview
Topics are groups of sentences that are all talking about the same thing. The key is topic modelling, a form of text mining used to identify and cluster related patterns of words in a body of text. Different ways to do this effectively have preoccupied academics for decades – long before AI was available to help. But for humans, trying to do this accurately and quickly when working with large amounts of text data is nigh on impossible.
Audiem’s first beta topic modelling tool was useful, but the labelling approach wasn’t intuitive. Sometimes it wasn’t immediately obvious what a topic category was about without reading individual comments to be sure. But not anymore!
You can find info about topics on the Headlines, Explore topics and Topics descriptions pages of the Audiem dashboard, as well as alongside the Workplace Mix™ in the Workplace with topics page. The Headlines and Explore topics pages allow you to see the frequency and sentiment of different topics comments. The default view shows the most frequent at the top, so you can get a feel for key topics in a dataset:
Explore topics also shows the groups of talking points within a topic. Some topics may have tens of sentences associated with them. Others may have hundreds or even thousands, depending on the size of your dataset. Most datasets have a cluster of major topics, followed by a long tail of minor topics. Using filters to explore the data will reveal more detail – for example, different locations or teams might have different topics of relevance, like the example of Commuting at the London location here:
Topics descriptions is a great place to get a feel for what each topic is about, with summary descriptions of all topics on the left-hand side.So here, exploring Commuting again:
Remember that Audiem analyses a text dataset BOTH deductively, using the Workplace Mix™ framework AND inductively, looking for topics. Talking points can appear in one or both of these concurrent analyses, depending on what they are about. You can see this most clearly on the Workplace with topics page:
And finally, a quick pro tip: use the search tool in the top right of most of the pages to dig into specific issues. For example here in the Topics descriptions page you can see how ‘commute’ doesn’t just feature in the Commuting topic, but also in a contributory way in some other topics too:
Changing the game
A Chief People Officer once said to me “All facilities managers ever do is come into the boardroom with presentations that show how much we don’t use things. They show us nothing but statistics, then use them to justify taking things away. Their data never tells us anything about what is actually going on, or why!”
Anyone who has struggled to find the common threads of meaning in pages and pages of written text knows just how challenging it can be. It is the poisoned chalice of qualitative research. Nevertheless, we wanted to enable any workplace professional to confidently present tangible evidence to their leaders and colleagues, moving beyond general statistics and unfounded assertions. We wanted to arm Audiem users with quantifiable, relational impacts and actionable insights, promoting better decision-making and ultimately, improved workplaces.
We hope this article gives you the confidence to use Audiem to dig even deeper into your workplace experience data.