Businesses using Twitter to learn about their customers have to fish for useful information in a swamp of daily posts.
To glean something useful from this torrent of data, it helps to add context to tweets such as: what is the age and sex of the person tweeting, where do they live, and are they giving your product a kicking or singing its praises?
IBM has partnered with Twitter to add this context to tweets and allow companies to cross reference that data with sales, inventory, and other business information, as well as third party reports such as weather forecasts.
Once this added information has been sprinkled onto tweets, IBM believes firms can start mining users' timelines for insights that boost the bottom line, answering questions such as, 'Why did we see a spike in sales in this period and how do we repeat it?' or 'How are we letting customers down?'
From today, data that combines tweets with this additional information will be available via APIs through IBM's cloud-based Bluemix service.
IBM will also work with firms to build analytics dashboards and visualisations that bring together these enriched tweets and business data, and make them available via IBM's Watson analytics platform. These tools will be aimed at letting business professionals with limited technical skills answer questions about how the company is performing.
"We take raw unstructured data and put it into a form that people can make decisions around very easily," said Alistair Rennie, GM of IBM's analytics business.
"I might say 'I want to find out what's trending in my industry and how does it differ by gender and location?'.
"You can get that information very easily and then use predictive analytics to correlate where you have stores, what kind of inventory you've got, and what would be the best steps to take to increase sales in each region. You have empirical evidence about what your next best step would be."
The services provide access to tweets going back to 2009. With about 500 million updates every day, that's a lot of posts. However, users can restrict data. For example, a restaurant could limit itself to seeing tweets with the word 'pizza' or 'bagel', from within a certain range of dates.
More than 100 firms have been trialling the services and IBM singled out a couple of examples of how it benefited businesses.
A telecoms operator was able to better identify when customers were likely to switch to rival providers after it discovered those who complained about service interruptions brought on by bad weather were more likely to switch. IBM says this insight increased the accuracy with which they could predict customer churn by five percent.
Retailers also identified a link between customer dissatisfaction and staff turnover, as loyal customers valued the relationship they had with staff and felt disappointed when they left.
Much of the data added by IBM is approximated based on analysing the tweets themselves and other sources of information - for example the user's Twitter profile alongside the content, pattern and timing of their tweets to decide on their location. To help with this process, IBM is using its System U software, which Rennie says can look at a spectrum of tweets and calculate a demographic profile with "pretty good precision".