Administration Updates: Ensuring You Are Getting the Best ROI From Brandwatch
By Nick TaylorOct 15
How has living through a pandemic changed consumer behavior and perceptions?
However, we’ve recently started work on a project to make our author analysis even better, even more insightful and even more useful.
Yesterday, we released the first step of this new author analysis work. Now, within Brandwatch you will be able to see gender and ‘account type’ (that’s whether it’s an individual, or an organisation/company/brand account) for your Twitter data.
And, as always, you will be able to dive into that data to see what makes up the stats, unlike many other tools. We want you to be confident in the analysis, which is why we always let you click into the charts to see those tweets behind the numbers, so there are no shady un-clickable charts here.
How do we determine gender and account type?
The gender analysis is based on first names, using a huge dictionary of (currently around 44,000) first names. Where a name is ambiguous – i.e. it could be either gender, or is not in the library – we will not classify it, preferring to focus on ensuring the precision and accuracy of those that are classified (similar to our approach to sentiment analysis accuracy).
In addition to the dictionary, we also have a list of words, such as ‘mum’, ‘father’, housewife’, etc that, when used in the Twitter bio, can be used to classify the gender.
The account type analysis is slightly more complicated. It’s based on the full name and the language used in the Twitter bio, with a machine learning system that can detect words and phrases denoting whether that account belongs to an individual (such as @tales_of_cake: me!) or an organisation or company (such as @Brandwatch: us!).
It does this by analysing the occurrence of words and phrases such as ‘official’, ‘I’, ‘we’, ‘our’, ‘we are’, ‘I am’, etc, whilst also using the name dictionary used for gender and an additional dictionary of company names.
The analysis will continue to improve as we grow our library of names and the machine learns, and we’re also looking at pulling in other information to help classify the gender or account type.
In the coming months we’ll also be introducing the ability for users to assign or change gender and/or account type for mentions, meaning that you will be able to refine the results even further, plus adding the ability to backfill the data to get gender and account information for historic mentions.
Gender and Account Type in action
For a quick demonstration of this new feature, what better than to use yesterday’s history-making Wimbledon final as a basis for the analysis. We set up a basic query tracking ‘Wimbledon’ for the purposes of showing what the gender and account type functionality can be useful for.
Looking at this morning’s tweets about the great win by Murray, we can see that males are slightly more interested in tweeting about Wimbledon.
We can then dive into those mentions to see the tweets from, say, the males:
We can also break it down by account type, which reveals that the large majority of tweets are from individuals, though lots of organisations and companies got involved too. Clicking in to those organisations we can see that a wide range of companies, charities and publications got involved in the good feeling, from Mecerdes-Benz to Glasgow News.
Of course, these new classifications mean you can also filter your Twitter mentions in other components. Say, for example, you only want to see tweets by individuals as you’re focusing on customer mentions, preferring not to see tweets by other organisations. Or, you might want to just see what the big organisations and news accounts are saying.
You can now easily do so in the mentions component or in other components such as charts, revealing just the data you want.
You could also filter you author lists and tables by a specific gender or account, allowing you to see the top authors of a specific gender or from, say, just organisations. Here are the top male Wimbledon tweeters:
And here are the top organisations:
Sentiment, categories, over time … the possibilities are endless
There are plenty of other uses. For example, you could take a look at how the topics of conversation differ between males and females. Perhaps you could look at how sentiment differs between males and females (below), or plot your categories against gender or account type to see which topics or types of comments are coming from whom.
You’ll also be able to plot the data over time, so you can see how the gender or account type split changes over time.
Where to find it
The possibilities are endless, but I think that’s quite enough charts for today; we’ll let you explore and discover more for yourself.
Current users can find Gender and Account type in filters, charts and tables under ‘Twitter’ in the ‘new’ Brandwatch. If you’re not a Brandwatch user, but wish you were, get a demo.
If you do anything particularly snazzy with the data and would like to share it, do let us know (send us a tweet). We always like to see examples of how people are using Brandwatch.
This is just the first step in our new Author analysis capabilities. Over the rest of the year we’ll be expanding analysis to other networks in addition to Twitter and expanding to our other supported languages, plus adding more capabilities for even more powerful, in-depth author and audience analysis. We’re very excited, and hope you are too.
Watch this space for more updates in the coming months!
Offering up analysis and data on everything from the events of the day to the latest consumer trends. Subscribe to keep your finger on the world’s pulse.