Social media is a marketer's dream. I'm not just talking about for advertising, this post is about stats. Statistics. Mathematics. Market research. Social and behavioural science. Stuff which fascinates me for some inexplicable reason.
I had some interest in using social media for data mining, and so I did some research. I found that the capabilities already out there go well beyond what I imagined. Prepare to be gobsmacked:
- Companies can use Twitter to profile people by geography, sentiment, gender and more. For example, data mining using the term "cat lover" might reveal where best to set up a cat accessories store. This example is a bit left field, but you can see where I'm going with it. Clients from financial services, healthcare, retail, politics, TV and news media companies have started using social media to gain a better understanding of their industry. Twitter has proven to be a surprisingly accurate indicator of levels of influenza in the community (and so data mining could be useful for predicting epidemics, tracking contamination etc). It is an excellent tool for political analysis and may even be a predictor of stock market activity and volatility, or perhaps a trading strategy.
- With Twitter Sentiment, anyone can search a brand and find out both positive and negative comments relating to it, as well as a graph depicting the percentages of positive vs negative sentiment. I did a few test searches and I'm not sure how accurate, or extensive this app is, but it's certainly a sign of things to come. Another tool, Twist, creates graphs of the frequency that a brand is mentioned at any given time during a day, month etc.
- Social networks are a goldmine for data for analysing behaviour and emotions. One study linked particular words with positive and negative moods. For example, posts using the word "awesome" or "fantastic" or "happy" would indicate a positive state, whereas words such as "angry" or "upset" would indicate a negative state. Obviously there is some scope for inaccuracy in an individual post, but once you mine large quantities of data you can observe interesting trends. They found that people tend to be happier in the morning and during weekends (no surprises there!). The messages revealed that people wake up happy and slowly grow more disgruntled and sour as the day goes on, though their affect usually rebounds in the evening (again, this makes sense!). You can see it here in graphical form.
- Did you know that Sydneysiders are the rudest Tweeters, followed by Brisbanites? Those in Adelaide are the most affectionate and those in Canberra the most aggressive and sad. No wonder I find Canberra somewhat depressing...
- Tweets or other posts provide a valuable time stamp, not only on moods but also behaviour. Marketers can use this sort of data to determine when best to target people with advertising, when to prompt people to buy and when to prompt them to share product information. For example, a UK analytics company did a study and found that pass-along value tweets are most frequently seen at 11:30am on Mondays, conversational tweets are heaviest on Tuesdays and news tweets are heaviest at 2:00pm on Tuesdays. I bet you didn't know that!
- Social networking offers endless posibilities for personality analysis and psychological profiling. Using Tweetstats, you can find out someone's most commonly used words (this is all about measuring the cognitive and emotional properties of a person based on the words they use). There are a number of other websites offering free personality analysis via your Twitter account. For those with the capabilities (such as analytics companies, marketing agencies and even the police if they invested in this sort of skill) you can see how a person's Facebook or Twitter account has the potential to provide a window into the person's mind. I envisage a crime show based on social media analysis, to find criminals and predict who a criminal's next victim will be...
- If marketers want to target certain types of people, for example to seed a product, they can use a simple tool like Twitalyzer to find out a person's (a) Influence score, which is basically your popularity score on Twitter (b) signal-to-noise ratio (c) one's propensity to ‘retweet' or pass along others' tweets (d) velocity - the rate one's updates on Twitter and (e) clout - based on how many times one is cited in tweets.
- For people using Twitter as a publicity or marketing tool, TweetEffect analyses which tweets make them gain or loose followers. Obviously there are other factors at play so this app has limited accuracy, but a more accurate tool will no doubt emerge in the future. Other research reveals what types of tweets get re-tweeted most often, and what users can do to get more re-tweets.
- Social media may even prove useful for companies to predict demand and control stock levels and manufacturing volumes. For example, they can find out if people like a newly released product, and how many intend to purchase it.
Some of above might seem obvious, but the point is that analysing data in this way may reveal some underlying and previously ignored trends and effects. In the book Supercrunchers: How Anything Can be Predicted, the author reveals some mind blowing ways that large data sets (such as social network posts) are used as predictors of future behaviour and how they are used to uncover cause and effect relationships. For example, casinos can calculate the maximum amount that a customer can lose and still come away feeling like they had a positive gambling experience thanks to a 'free' meal at the casino's restaurant. The blurb of this book asks the question: Why should you be worried if you are receiving good customer service? I definitely want to get my hands on it, it sounds fascinating, and worrying.
Is there anything about us that they won't know? Social media data analysis is a rapidly growing industry. Technology is rapidly changing the way that businesses operate, and well as the way we do business. Watch this space.
I never knew you could analyse Twitter data to that extent. Thanks for that post. Twitter Sentiment is interesting. I am off to play!
ReplyDelete