Now that the era of big data is in full flow, any organisation that wants to succeed in today's hyper-competitive market must now be implementing data-driven decision-making. Many organisations have started this journey with 'descriptive' analytics – the use of data mining techniques to find out what has happened, and root-cause analysis to get to the crux of why something happened.

Once they start to get comfortable with this kind of historically-based data-driven decision making, organisations can start to grow in their analytics maturity, embracing predictive analytic techniques and asking more forward-looking questions, such as 'what is my sales forecast?' or 'how can I expect this trend in pricing to grow over the next three months?'

As Luis Bajuk-Yorgan, senior director of product at TIBCO Analytics explains, 'prescriptive' analytics is the next stage of analytics maturity in which you begin making decisions based not only on individual predictions, 'but on an aggregated view of predicted relationships with known constraints, which yield recommendations on the best possible decision to make, given known, realistic limitations (such as cost, not overwhelming your customers with multiple contacts, organisational bandwidth, etc.).'

> See also: Predictive analytics: what if you could access tomorrow's stocks and shares today?

This decision-based analytics gives businesses the ability to best take advantage of a future opportunity. For example, questions like 'what marketing offering will my customers best respond to, while maximising my profit?', 'what price point will yield the best combination of sales and profitability?' and 'what number of distribution centres will minimise delivery time while maximising profitability?'

'It’s of no surprise then that prescriptive analytics is continuing to rise up the Gartner Hype Cycle as many early adopters are realising the competitive advantages that can be gained from this analysis,' says Bajuk-Yorgan. 'Armed with prescriptive analytics, businesses are able to improve their confidence in business outcomes.'

Though business optimisation in itself is many decades old, it's only now that businesses have been able to reach the next logical point. In 2013 Gartner called prescriptive analytics 'the final frontier for big data, where companies can finally turn the unprecedented levels of data in the enterprise into powerful action.

As Gartner analyst Lisa Kart explains, prescriptive analytics in itself is not the end goal, however, rather than a broadening of the set of tools available to businesses.

'We do advice organisations doing descriptive analytics, business intelligence (BI) and reporting to increase their skills, and then move to predictive and prescriptive – but you're not dropping the others,' says Kart.

'Prescriptive analytics is about applying logic and mathematics to data, with the goal to specify a prefered course of action- unlike other type of analytics the output is a decision,' she continues. 'Thats where it's really different. It's about trying to find the best decision, where best is defined by you, whether that's lowest cost, most efficient process, higher revenue, or one that meets customer needs. Fundamentally the focus begins with the business decision.'

Few and far between

As Gartner has found, it's very difficult to estimate just how many companies are currently using prescriptive analytics. This is because the ones that do aren't yet taking a 'joined up approach.'

'Just the other day I was talking to an organisation that was using predictive in marketing, and they wanted to know how to use prescriptive to define actions and offers and design their marketing campaigns,' Kart recounts. 'They didn't know I was already talking to another group in the organisation that had adopted it within supply chain and customer service. A lot of the time that's the case- very few organisations have adopted prescriptive analytics in the way they've adopted predictive, and they don't spread it to the whole enterprise.'

According to Gartner's Hype Cycle, which tracks the maturity of various technologies, prescriptive analytics hasn't yet made it to the peak yet and is still in its early stages. The analyst house estimates that around 5-10% of organisations are currenly using it.

'Interest is picking up on the topic though,' says Kart, 'I've noticed when I speak at conferences people have become a lot more aware of it.'

APT or Applied Predictive Technologies is one software company helping major firms get started. Many of them, such as pharmacy retail chain Boots, are using repeated experiments to find out the best layouts for their stores.

'Using the prescriptive method they can estimate store by store the different kinds of layouts that will work bestto achieve their particular goal,' explains Rubert Naylor, VP of APT. Broadly speaking in the UK, the method is being applied today across three big issues – marketing, advertising and promotions.

'It helps to plan capital investments and ask questions such as 'does it pay out to put in a kiosk in my store for direct consumer access to inventory?' and 'does it work in every store?' says Naylor. 'It also applies to things like store hours, labour allocation, training modules so on.'

The metric that most firms are interested in is return on investment (ROI), about more than that, can be applied to many things, such as optimising logistics and creating a better customer experience.

Parcel delivery service UPS is optimising its routes for drivers so that they can save a few miles in each day- adding up to close to 50 million dollars a year in savings, and vastly improving customer service.

'It's about trading off your objects things like revenue and loss, customer satisfaction, efficiency, and customer service,' says Kart. 'When people think about the benefits, UPS could've probably done a better job by optimising for fuel, but theyre also trying to optimise for customer service, so that's when some of the real benefits come in, being able to incorporate these tradeoffs and multiple objectives and find a solution, where best means many things.'

But for most organisations, prescriptive analytics is still very much in the ‘lab’ stage. And as Hugh Cox, chief data officer for Rosslyn analytics explains, it is only very few large companies that have access to these labs and the computational power needed to run the complex algorithms associated with prescriptive analytics.

'For most companies, it will continue to be a concept that is at the hype stage, and not tangible and usable in everyday business,' he says. 'The focus for them, and their budget will be on ‘deep learning’; black box-driven decision-making which effectively bridges between predictive and prescriptive analytics. The Google Self Drive car is making use of deep learning, and we will see more examples throughout 2015 in enterprise. However, while deep learning will enable organisations to uncover the ‘what’; the ‘how’ and ‘why’ will still remain elusive for many.'

A data rainbow

As Gartner's Kart explains, developing prescriptive analytical capabilities are less of a huge leap up and more of a spectrum, starting with 'decision management'- the ability to make data insights more actionable by combining them with business rules.

'Optimisation is the higher end,' says Kart, 'This involves applying analytics that recommend a course of action and that can also be achieved by combining data with predictive analytics. It's not that it requires a whole new toolset- there are tools out there focused on making prescriptive easier because there are things that will solve the optimisation problem, or do simulation to let you try 'what if' scenarios. In short, there area lot of ways you can piece together what you have today with predictive analytics and business rules to really start with prescriptive analytics.'

Having the capabilities already in-house is certainly useful, says Kart, but it's not a pre-requisite for companies.

'Building the capabilities in house is just one option, but when people are getting started in a new areas, it makes a lot of sense to call in experts in that area, maybe service providers, or hiring in someone with that kind of experience to know the pitfalls and what to look for,' she says.

But the main thing is to work closely with the decision makers themselves.

'When I help people get started in prescriptive analytics, I tell them to focus on decisions in the organisation- sufficiently complex decisions where you have a lot of data and are already doing some really good descriptive or predictive analytics but you need help turning those predictions into decisions,' says Kart. 'But fundamentally the decision makers have to understand the objectives and constraints of what's being done, and be able to frame that decision.'

> See also: Traditional big data strategy is ancient history

As Cox from Rosslyn analytics argues, there is still a technological leap needed to go from rule-based networks to being able to effectively explain what is happening in real-life situations. Consequently, prescriptive analytics is still somewhat limited.

'Given the exponential increases in computation power through the cloud,' he says, 'it's now possible for these labs to visit previously dormant algorithms and use them to drive insight and progress. Amazon, for example, is installing the equivalent of the computational power that was required to run their whole $7bn business in 2004 every day.'

The next breakthrough in prescriptive analytics, says Cox, will be to make it available for the wider market.

'Components of prescriptive analytics will be available for use in 2015 and there will be progress in terms of advances and what is available,' he says. 'We will see some increase in prescriptive analytics used by big consultancy projects partnering with academics for continued development.'However there won’t be an off-the-shelf prescriptive analytics solution available for general purchase and use – that is still a long way off.'