A couple of months ago we were briefed by our client Kuoni, to use social media to promote their Holiday Health Experiment competition and research project. To support the theme of investigation into the UK public’s health, and their need for a holiday, we decided to investigate this based on what people were tweeting about. From our findings we created an infographic to visually explain the data in an engaging way. This was offered as an asset to those promoting the campaign, to add an additional layer of insight into how people feel and talk about holidays and health.
So, we discovered the following:
- People were 61% more expressive of emotions (feeling or being happy, sad, unhappy, healthy, ill, stressed or relaxed) in winter months than they were in summer months.
- The most common term when people said they wanted/needed a holiday was an unhappy face :(
- Of our major cities, Manchester is the most stressed (/unhappy/sad/ill), with 1.5 stressed/unhappy/sad/ill tweets for every one relaxed/happy/healthy tweet.
- At the other end of the scale, there’s 1.7 relaxed/happy/healthy tweets to one stressed/unhappy/sad/ill tweet in Bradford.
- Manchester is also the city most wanting a holiday, with twelve times as many people there expressing the desire for a holiday in comparison to Belfast, where people least expressed such a need.
- People in Manchester also expressed their need for a holiday more than twice as much as any other major UK city.
- People said they wanted/needed a holiday almost twice as often in Summer than they did in Autumn. (Could this be because theyʼve already been on holiday by Autumn?).
The method in more detail is below, and here’s the infographic designed by Ollie Aplin (click to enlarge):
More about the method:
We designed a list of possible investigative questions, which were refined based on what data was actually available, and what we could possibly interpret from people’s tweets. For example, we could derive that people are saying they are happy, but we couldn’t really derive that that meant they were truly happy.
Then we designed queries to bring up relevant conversations. Some of these ended up being extremely long and complex to ensure relevance. For example, most people saying “I’m sad” were saying it in the context of being a loser rather than being unhappy, so we had to exclude that. We chose common positive and negative feelings about health and happiness and designed around those.
Then we found a buzz monitoring tool which would provide us with historical tweets, going back a year (to account for seasonal variation), restricted to UK accounts. We used Radian6 which was the only tool available which would provide 100% of historical tweets. All other tools would at best provide historical (more than a month that is) tweets that happened to overlap with their other clients’ queries, which would add bias. However, it turned out that a single data dump was limited to a certain volume of tweets, so we ended up doing a lot of manual exports (due to a volume limitation on data export) and choosing a representative sample for the geographical stuff.
Steve then assigned each geocoded tweet to its nearest city, from the top 20 biggest cities in the UK. Then he got population data for each city and normalised the tweet data per city. (Otherwise London would come out top for everything). He was then able to create a ratio of happy/sad category tweets, and see how they mapped to cities.
Overall we analysed 2,907,099 tweets, all from UK accounts, over a period of 12 months.
But over the past year, the volume of accounts and therefore tweets on Twitter has been increasing, which would affect seasonal trend data. Don’t worry, we accounted for that by normalising the data based on the varying UK Twitter traffic as an indicative ratio, which we got from Google Trends for Websites.
It was very interesting doing this project, particularly looking at people’s (expressed) feelings as data. Reminded me of the old We Feel Fine project which I used to love (and still do). What’s also great is that we now have a wealth of tweets about feelings from people in the UK, which we could do more data analysis of, if we get the chance.