"Data is the new oil," declared Clive Humby, a Sheffield mathematician who with his wife, Edwina Dunn, made £90m helping Tesco with its Clubcard system. Though he said it in 2006, the realisation that there is a lot of money to be made – and lost – through the careful or careless marshalling of "big data" has only begun to dawn on many business people.
The crash that knocked out the Nasdaq trading system was only one example; in the past week, Amazon, Google and Apple have all suffered breaks in service that have affected their customers, lost sales or caused inconvenience. When Amazon's main shopping site went offline for nearly an hour, estimates suggested millions of dollars of sales were lost. When Google went offline for just four minutes this month, the missed chance to show adverts to searchers could have cost it $500,000.
Michael Palmer, of the Association of National Advertisers, expanded on Humby's quote: "Data is just like crude. It's valuable, but if unrefined it cannot really be used. It has to be changed into gas, plastic, chemicals, etc to create a valuable entity that drives profitable activity; so must data be broken down, analysed for it to have value."
For Amazon and Google especially, being able to process and store huge amounts of data is essential to their success. But when it goes wrong – as it inevitably does – the effects can be dramatic. And the biggest problem can be data which is "dirty", containing erroneous or garbled entries which can corrupt files and throw systems into a tailspin. That can cause the sort of "software glitch" that brought down the Nasdaq – or lead to servers locking up and a domino effect of overloading.
"Whenever I meet people I ask them about the quality of their data," says Duncan Ross, director of data sciences at Teradata, which provides data warehousing systems for clients including Walmart, Tesco and Apple. "When they tell me that the quality is really good, I assume that they haven't actually looked at it."
That's because the systems businesses use increasingly rely on external data, whether from governments or private companies, which cannot be assumed to be reliable. Ross says: "It's always dirty."
And that puts businesses at the mercy of the occasional high-pressure data spill. Inject the wrong piece of data and trouble follows. In April, when automatic systems read a tweet from the Associated Press Twitter feed which said the White House had been bombed and Barack Obama injured, they sold stock faster than the blink of an eye, sending the US Dow index down 143 points within seconds. But the data was dirty: AP's Twitter feed had been hacked.
The statistics are stunning: about 90% of all the data in the world has been generated in the past two years (a statistic that is holding roughly true even as time passes). There are about 2.7 zettabytes of data in the digital universe, where 1ZB of data is a billion terabytes (a typical computer hard drive these days can hold about 0.5TB, or 500 gigabytes). IBM predicts that will hit 8ZB by 2015. Facebook alone stores and analyses more than 50 petabytes (50,000 TB) of data.
Data is also moving faster than ever before: by last year, between 50% and 70% of all trades on US stock exchanges was being done by machines which could execute a transaction in less than a microsecond (millionth of a second). Internet connectivity is run through fibre optic connections where financial companies will seek to shave five milliseconds from a connection so those nanosecond-scale transactions can be done even more quickly.
We're also storing and processing more and more of it. But that doesn't mean we're just hoarding data, says Ross: "The pace of change of markets generally is so rapid that it doesn't make sense to retain information for more than a few years.
"If you think about something like handsets or phone calls, go back three or four years and the latest thing was the iPhone 3GS and BlackBerrys were really popular. It's useless for analysis. The only area where you store data for any length of time is regulatory work."
Yet the amount of short-term data being processed is rocketing. Twitter recently rewrote its entire back-end database system because it would not otherwise be able to cope with the 500m tweets, each as long as a text message, arriving each day. (By comparison, the four UK mobile networks together handle about 250m text messages a day, a figure is falling as people shift to services such as Twitter.)
Raffi Krikorian, Twitter's vice-president for "platform engineering" – that is, in charge of keeping the ship running, and the whale away – admits that the 2010 World Cup was a dramatic lesson, when goals, penalties and free kicks being watched by a global audience made the system creak and quail.
A wholesale rewrite of its back-end systems over the past three years means it can now "withstand" events such as the showing in Japan of a new film called Castle in the Sky, which set a record by generating 143,199 tweets a second on 2 August at 3.21pm BST. "The number of machines involved in serving the site has been decreased anywhere from five to 12 times," he notes proudly. Even better, Twitter has been available for about 99.9999% of the past six months, even with that Japanese peak.
Yet even while Twitter moved quickly, the concern is that other parts of the information structure will not be resilient enough to deal with inevitable collapses – and that could have unpredictable effects.
"We've had mains power for more than a century, but can have an outage caused by somebody not resetting a switch," says Ross. "The only security companies can have is if they build plenty of redundancy into the systems that affect our lives."
guardian.co.uk © Guardian News and Media 2013
