Tweets and Widgets – a case study of real time customer feedback

If we believe what we are told by the Swedish statisticians, 90% of their young people aged 16-24 take part in Social Network sites. 91,316 of the whole population uses twitter and twitter use by men and women is 56/44% respectively. 85% of tweets are from individuals and 11% organisations. Interestingly, there appears to be little difference in twitter use between all economic bands e.g. those earning less than 30,000 euros up to those on more than 75,000 per year. Also twitter use by women overtakes men by the age of 45!

Well that’s all very well for people interested in user statistics but what does tweeting offer mature organisations looking to improve services to customers? You might want to take a look at work done at Uppsala University in Sweden, through ROLE technologies. Sweden has a snow problem! Uppsala University decided to track tweets from customers over a three month period and analyse the content using a widget bundle.

The process and findings are fascinating. The widget approach enabled students to use tools to collect, code and present their findings but also provided a medium to share and challenge different interpretations of the findings. It was successful on a technological and learning level and also provided valuable data to the Swedish Railway service.

Now you may be asking `couldn’t this be done without widgets?’. As someone who has earned a living over many years analysing survey data, I would say yes it can, but the level of social interaction that the widget approach offers is certainly a big improvement on traditional methods. For example, the ability to codify, categorise then systematically challenge categories at group and individual level is a real improvement in traditional coding. It encourages group and individual review of customers’ open comments. This in turn encourages organisations to really explore what customers `mean’ when they respond by open text feedback.

I can think of lots of uses for the widget bundle. For example I work with organisations on quality issues. One of the difficulties we experience is defining what we mean by `good’ in specific areas. Many people think they know what `good’ is but it is only in the discussion of detail that differences of opinion emerge and the conditions necessary for `good’ to happen surface.

This widget bundle forces the discussion. I hope to use it as a tool to explore `quality’ in specific learning areas. It can be used by teachers, learners and clients to compare their responses to questions around expectation, effectiveness and approach. Anyway…on to the case study……

The Swedish Railway and the relevance of Tweets

Here is a good example of the use of Widget Bundles that support a) a real business concern b) learning and research analysis skills.

Uppsala University (a partner in the ROLE project) set an assignment for social studies students to analyse `tweets’ sent to and from the Swedish Railway company during the winter 2010/2011. This was a problematic winter for rail traffic in Sweden, with major delays as a result of the extreme weather.

The first widget (Widget 1) contains all the tweets to and from the Railway. The students were invited to categorise all the tweets. Thy did this by selecting a tweet, linking it to a category and they also had the option to add comments.

Widgets 2, 1 and 3 from Left to right.

The aim of the exercise was for the group of students to individually categories all the tweets and experience the conflict that arose from interpreting open text responses. Individual students could assign a tweet to a number of different categories. This is one of the major ongoing problems of analysing open text responses. Open text is subject to interpretation!

Another Widget (Widget 2) shows the tweets on a timeline. It is possible to view tweets by 5 minute intervals, day or month. The timeline shows individual tweets and authors. If you select an author, all tweets by the same author will be selected.

Tweets are grouped into categories and shown on a pie chart. (Widget 3) When you select a category of the pie chart, all the corresponding tweets in the tweet list are highlighted.

To support the analysis of the tweets, students were invited to join a forum. The Forum Widget (Widget 4) shows a thread of forum posts. It also contains references such as tweets. If users select the reference, all the other widgets will relate to the same tweet. The aim of the forum was to get concensus on the interpretation of tweet `meanings’ and categories.

Widgets 4 and 5

There is also a `content viewer’ widget (Widget 5). This shows all relevant content when working in other widgets.

The demonstration shows how a bundle of 5 widgets work together. Each widget performing a distinct but related function to the other widgets in the bundle. New widgets can be developed to add more functionality to the bundle and each widget can be pulled out and used independently or as part of another bundle.

If you would like to download the widgets or create your own, please contact Glyn Owen at and all codes etc will be made available to you.

Case study translation for this article provided by Uppsala Univeristy. Thanks to Matthias Palmer – Software Engineer, Uppsala University.

You can view a screen cast of the Widget Bundle on the Uppsala University testbed.

The source mp4 at:

N.b. The screen cast is in Swedish but much of the text is in English and the demo is good enough to illustrate how the widgets work together to support an analysis of customer tweets.

This entry was posted in News. Bookmark the permalink.

Comments are closed.