Tweet it your way? Twitter's Capacity for Consumer Sentiment Measurement - Part I

     I know two Twitter posts in a row (4 if you count parts coming tomorrow and over the weekend) may seem a little excessive, but with all the news surrounding the service right now, I'm writing my ongoing fixation off to the influence of news coverage and marketing zeitgeist. With today's valuation of Twitter at the $1 billion mark (paltry compared to the estimated $8 billion value of Facebook, but impressive none the less), attempting to quantify the various uses for the service seems rather envogue.

     From a marketing perspective, Twitter's capacity to deliver a breadth of consumer thoughts is already well understood. Typing a company's name into search.twitter.com or the homepage link is a rudementary proof of the torrent of thoughts available about a product or business. While some product sectors produce more insight than others (luxury/aspirational products, global brands, recreational products and the like lend themselves to more conversation than the more mundane or basic such as household FMCGs and financial services), the ability to look into the amorphous "crowd" and pull out buzz is invaluable. This utility has spawned not only business interest in the website itself, but has also impacted 3rd party services. Twitter specific services such as Hootsuite harness the reasonably accessible Search and REST APIs within the network (ease of use being something I can attest to having written my own Twitter analytics software),in addition wider buzz tracking offerings have added functionality beyond blogs and forums to include Tweet aggreagation.

But what do customers really think about the Baconator(TM) Hal?

     Anyone who's familiar with the advanced search functionality of Twitter knows that added value stems from the ability to segment messages gathered from the service. Searching by user, user mentions, network position from a fixed point (if you're willing to code a large amount of secondary aggregation data) and geographic location mean that user's messages can gain a relevance. However, with this wealth of possible data, what actionable insights can be generated?

     At the most basic level, number of mentions and time series data on mentions are the easiest data to procure. By simply collecting a period of data on a term, one can compare activities such as news coverage with Twitter activity.Combining message amounts with other competitors in a sector can generate topline statistics on its performance and loosely gauge twitter activity possibilities for sector ("market tweet size" if you will) and company share ("share of tweet" if you won't).

     At a slightly more advanced level, user mentions and user profiling are possible, as well as multiple searches to determine activity by various geographic areas. Combining this with analyzing Twitter profiles, a brand can gain insights on what type of users are talking about the brand (though the value of such topical profiling is limited). Geographic searching allows not only for regional activity comparison (great for larger brands), but also for specializing a search area (great for smaller brands and singular locations such as a local chain of restaurants).

Looking at the UK BK Tweets for the last few months makes me want to have a "Chicken Fries Whopper" for some reason

     Finally, data such as message content, message platform and links can generate insight on user behavior when mentioning the brand or associated content (through seperate link spidering - something that is easier said than done with bit.ly and tinyurl's api limits). Perhaps most evidently, the time consuming process of actually reading every user message can generate a specific picture of trends emerging around a brand. Depending on the volume of messages, this can be unfeasible, confusing or flatly impossible; in situations such as this, I've found that porting messages into a word cloud generator such as Wordle can provide top line figures on atleast the most basic of consumer sentiment.

     For the future, I see the nature of "real-time" search analysis becoming even more reliable and specific. Services such as Foursquare (Hurry up and get to London!) are ushering in consistent geolocation of messages in a system that encourages more detail than simple experiential tweets (i.e. "I'm off to get lunch at Business X" vs. "I just had horrible service at Business X").  Furthermore advances in automated semantic analysis are making the process of analyzing user messages for trends and sentiment quicker and more accurate, something that will be highly appreciated by those attempting to read the breadth of aggreagated content.

    In considering consumer sentiment measurement, perhaps the best way to illustrate current and future possibilities is to provide an example. Tomorrow in Part II, we'll take a real-world example and generate some topline data on Burger King within the UK utilizing a variety of web tools.