Published on the 20/09/2017 | Written by Donovan Jackson

Decision support ‘in the now’ plays a part in Formula One win…

Fresh from the side-lines of the Singapore Grand Prix where the team it co-sponsors notched up an unlikely win and third place, TIBCO’s APAC head of sales Erich Gerber said streaming analytics is the next big thing for a wide range of industries. But before getting to that, he wanted to talk racing.

“The team, Mercedes, is a real analytics shop. Without downplaying the value of the driver, the team has 20 data analysts on site for every race, and a further 40 to 60 in the UK when a race is underway. This team is constantly looking at a wide range of data sources direct from the track to figure out the best tactics,” Gerber told iStart.

This most recent GP is the perfect microcosm, if one in somewhat rarefied air, to demonstrate the value of streaming analytics. The win for Mercedes was an outside chance as Lewis Hamilton started well back on the grid on a track where overtaking is difficult. In a split second, a crash took out the leading contenders; the constant stream of telemetry data, 60-odd eggheads and Hamilton’s skill came together to put him across the line first. The team went from a strategy of damage limitation, to one of targeting the podium – and it had the smarts to pull it off.

So, what’s the big deal with ‘streaming’, then, and how does it differ from the analytics we know and love? Gerber pointed out that like the catch-all term ‘IT industry’, today ‘analytics’ is a broad definition within which lies multiple specialisations (all the way on up to another broad term, AI). “The key words are ‘real time’. Streaming means you are picking up and acting on things as they happen, rather than having the traditional rear-facing view of what has happened.”

The biggest difference with streaming, then, is the immediacy of analysing data and acting on it as it happens.

The value of data, it is commonly accepted, deteriorates over time. In multiple settings – say, a Grand Prix race, a fraudulent credit card transaction, a component found to be lacking in quality going into a production line – the ability to grab streams of data as they happen, analyse and contextualise, means the ability to act before problems, losses or missed opportunities occur.

All of which sounds wonderful. But, iStart wanted to know, are there any conditions precedent which are required for an organisation to implement a streaming analytics solution? Why yes, said Robert Merlicek, TIBCO’s APAC CTO; “You’d need modern applications which make data available to be consumed by a streaming analytics solution. Traditional IT systems tend to be batch-oriented with ‘data processing’ windows; more modern ones can provide real time data. That, coupled with the rise of connectivity and web services means additional data can be accessed for further insight and context.”

There is, of course, ample opportunity to introduce some of the more whizz-bang stuff going on in the analytics field, including one of the subsets of AI: machine learning. Gerber said the idea behind machine learning in streaming analytics is to address the challenge that the human interface isn’t always the best one when querying whether an issue is really an issue or just a false positive.

“When an event is detected as likely to happen, with a probability of say 98 percent, you need to monitor what happens afterwards as you could be wrong and if so, you need to recover the situation,” he explained. “So, you’d use machine learning to improve and refine the model based on previous actions.”

In the useful example of a credit card transaction, a false positive can mean an angry/embarrassed customer. It is also a useful example because the traditional way of checking if a transaction might be legit or not is to pick up the phone – but that process will take too long to prevent a loss if something untoward is happening.

But while the use cases might be broad (‘any organisation with a wide consumer base, particularly in the business to consumer environment’) with even the likes of Melbourne Airport putting streaming analytics to work to gain a better picture of its operations, it isn’t yet a mainstream technology, said Gerber. “It isn’t yet implemented widely; there is always first resistance to do it, then the adoption hype curve kicks in. The real difficulty isn’t how to implement streaming analytics, it is getting together a task force to tackle problems and decide on how to approach them.”