What are 'Themes' in Wordnerds?

What are 'Themes' in Wordnerds?
Why can't I classify my data just using 'Topics'?
Context Themes
Keyword Themes
Upload Themes

What are 'Themes' in Wordnerds?

Great question! Themes are a big part of using Wordnerds. Themes are used to create custom classifications so that you can group your data in the way you want.

Themes can be based on what is being discussed in the verbatim content itself (using context or keyword themes - more on those shortly) or based on metadata you already have attached to the verbatim content.

Once a 'Theme' is set up in your project, you can...

See the volume and sentiment of the Theme in your data
Trend your Theme over time
Apply a Theme historically to get a retrospective view on a new issue
Use your Theme as a filter and a way to slice your data during analysis
Group Themes into 'Theme Categories' to size and trend at a category level
Group Themes or 'Theme Categories' into 'Frameworks' that act as a lens through which to view and analyse your data

Why can't I classify my data just using 'Topics'

Well, you can! But... we wouldn't if we were you.

Unsupervised topics are great (like, really great) for..

Getting an immediate feel for what's in your data, no training/setup required
Finding surprises - the things you didn't know to listen for

BUT they do not robustly size or classify issues.

'Topics' will surface the issues in your data at a given time, based on words and phrases commonly used together. However, the human job of deciding how these should be grouped and how your data should be classified has not been done.

For example, "expensive", "poor value for money'", "price... too much" could all be 'Topics' that the platform surfaces for you, but if you want to know how many customers are talking about pricing (which these 'Topics' all are, in different ways), you would train a context theme to do that.

Context themes

Context themes are usually the best way to classify your data. You can train context themes to robustly size issues, regardless of the specific language being used.

Context themes are great for...

Issues that could be described in lots of different ways (that's most things)
Issues where the context in which the language is used is crucial to understanding what is being discussed (that's also most things)

Still not convinced by context themes?

Someone talking about a train being busy may use the phrase "the train was busy", but they could also mention that they couldn't get a seat, or say something more unpredictable like "we were packed in like a tin of sardines". Someone could also use the word "busy" to talk about something else entirely, such as being busy themselves.

If your theme misses mentions of the issue you want to classify because people used vocabulary you didn't predict, or because they used the vocabulary in ways you didn't predict (or both!), then you won't have robustly and accurately sized the issue, and your classification won't be going the job you need it to.

Keyword themes

Keyword themes allow you to build rules that include keywords and phrases, as well as specifying exclusions, so that you can classify your data based on specific language being used. But (there's a but!)... this relies on you being able to accurately predict how your customers speak. And humans aren't always that predictable. Therefore, while there are situations where keyword themes do make the most sense, for example, if you want mentions of a specific location or product name, keyword themes can get really complicated for classifying 'issues', and are often not the best option (even if they are still pretty good).

Keyword themes are great for...

They can sometimes be a better way to classify mentions of things (rather than issues) - e.g. a product name, or a location name
Situations where what you're classifying is only described using specific words, and when those words are not used in other contexts

Upload themes

Upload themes are great for analysing your data based on the metadata already associated with your customer feedback - for example, the NPS scores or groups given alongside open-ended survey responses, complaint stage, demographic information, location or business unit - anything! While these are different to the context and keyword themes that you create, you still choose what upload themes to use, and they can play an important part in your analysis.

Upload themes are great for...

Categorising data based on information you already have and don't need to find in the verbatim itself
To add a metric you will use for analysis purposes - for example, the score given alongside an open-ended survey response

What are 'Themes' in Wordnerds?

This article explains what 'Themes' are in Wordnerds, the different types of 'Themes', and how they can be used to classify your data