Turn insight into action with predictive analytics

There are a host of uses for predictive analytics

After years of talk, predictive analytics is finding a host of uses. Now the question is whether business has the managers and the data scientists to fully exploit it.

Updated 26 August 2016

By Beverley Head

When a Telstra call centre operator asks “How are you?”, they pay very careful attention to the response.

Data analytics has revealed that 48 percent of those who want to disconnect will respond "not too bad". That raises a red flag and savvy call centre jockeys know they have a limited window to salvage the relationship.

When Telstra first harnessed predictive analytics it was bent purely on boosting customer satisfaction. But according to Belinda Haden, head of customer analytics (APAC) for Telstra’s supplier, Verint, it has also helped deliver an 11 per cent improvement in revenues since deployment.

She says that analysing calls – for content and emotion – has allowed Telstra to identify common problems and fix many of them, while also speeding complaint resolution. Haden says customer calls have fallen 28 per cent while complaints to the telecommunications ombudsman (about Telstra) have dropped by 24 per cent. Predictive analytics harnesses technology to identify patterns in large collections of data.

Often these collections are not neatly in a database but sitting in, say, a pile of audio files – “unstructured data”, as data experts call it. Predictive analytics allows likely outcomes to be identified and represented using specially developed visualisation tools, in real time, so that businesses can respond immediately.

For instance, an app may tell a retailer that a customer is approaching a store. The retailer already knows the customer’s buying pattern from an analysis of past purchases, so they can match that against what similarly profiled consumers have gone on to buy. They can then send a coupon digitally to the customer to encourage them to come into the store and make a purchase based on that insight.

Organisations can also use predictive analytics to tweak their business model in response to shifting market conditions or emerging opportunities.

ANZ’s deployment of the Falcon system was at the leading edge of analytics capability – trawling real-time card transactions to identify anomalies. It can ask: is that really you buying $1000 worth of electronics at 5am in Singapore when you spent $45 at Woolworths in Sydney just yesterday?

When ANZ launched the system in late 2005 it was a market leader, and predictive analytics was more vendor hype than reality for many enterprises. Today, however, the tools and technology have matured and become more affordable, and the computer power required to make sense of the data is plentiful and inexpensive.

The remaining challenges for most organisations are to convince business leaders of the value of predictive analytics, and then to find the (still scarce) data science skills required to turn insight into action, and hence business value.

Making use of Big Data

Any discussion of predictive analytics is now twinned with “big data” – the fire-hose of information available to enterprises.

Consider the Boeing 787 Dreamliner. Every flight feeds half a terabyte of data into Boeing’s Airplane Health Management application. That application analyses the data on the fly, pinpoints patterns, identifies opportunities for preventive maintenance, and then feeds that insight back into Boeing’s back-end information systems to schedule crews or order parts. It streamlines operations, reduces downtime and slices risk.

That’s just machine-generated data. Add to that the information sourced from existing structured business systems and spreadsheets, from call centre conversations, smart devices and smartphones loaded with apps, and from social media.

According to EMC’s Digital Universe study, data is doubling every two years. It deduced that by 2020 there will be 44 zettabytes (trillion gigabytes) of digital data on the planet.
The challenge for predictive analytics is to extract value from that data.

IBM grasped the nettle when it set about building its cognitive machine, Watson. Like most great technology advances, Watson began in a bar.

IBM researchers glanced up at the TV show Jeopardy and started to wonder if it would be possible to build a computer that could win. In 2011, when Watson competed against the human champions of the game show, it did.

Dev Mookerjee, technical lead for the IBM Watson Business Unit in the Asia-Pacific, explains that the system has been developed “to replicate humans’ cognitive framework”. The machine can’t be programmed, but when provided with enough domain data, it can learn to make sense of the information and provide a weighted response to a question.

Watson is now powering an IBM predictive analytics service called Watson Engagement Adviser, which Melbourne-based Deakin University uses to respond to student inquiries. Initially the service is handling simple questions, but eventually the university wants it to provide career or course counselling that may reduce student dropout rates. ANZ Bank, meanwhile, is working out how Watson can support its financial advisers. Australia’s Department of Immigration has also tested Watson.

Mookerjee says four Japanese insurance companies have automated 80 per cent of their health claims using Watson. And in the US, oncologists from Memorial Sloan Kettering Cancer Center have been sharing their expertise with the system, so that Watson can support doctors developing patient treatment paths.

Predictive analytics has matured significantly and is now being deployed across a broad swathe of industries, while the surrounding technology continues its relentless advance.

Specially designed systems, such as the Hadoop software framework, can be used as a platform to analyse enormous unstructured data collections (the largest data collections are now so big they are called “data lakes”.)

SAS, the leader in predictive analytics according to analyst Gartner, is using in-memory processing to analyse data coming in at rates of 400 transactions per second. The falling cost of processing and storage has helped to make such tasks feasible.

Legitimising analytics

While predictive analytics harnesses technology, there is far more to the equation. Global chief analytics officer of EY, Chris Mazzei, describes the situation nicely. 

"Analytics is not a technology issue,” says Mazzei. “It’s a strategy and operational issue. Analytics is changing how organisations make decisions and take actions. Data by itself has limited value, but when managed as a strategic asset, data can change how organisations compete and win.”

But is data valued by boards and the C-suite?

Not sufficiently, according to independent IT consultant Dez Blanchfield. “CIOs [chief information officers] understand it but board members may not understand it yet,” he says. 

“The value of big data won’t be understood by CEOs until it’s monetised, and big data is often misunderstood by CFOs as a risk or a cost rather than an asset.”

He points to industry disrupters which have almost no physical assets but an abundance of data assets. “The world’s most valuable retailer has no inventory – Alibaba. The fastest growing banks have no money – Society One. The world’s largest taxi company owns no taxis – Uber. And the largest accommodation provider owns no real estate – Airbnb.

“Data isn’t an IT issue … it’s a bona fide business issue and needs to be legitimised inside boardrooms,” says Blanchfield.

Analytics for SMEs

For Boeing or Aera (see “Analytics Meets the Oil Well”, below) the value of data can be tracked all the way to the bottom line. But what about smaller businesses (SMEs) – does predictive analysis have a role to play there?

Paul Franks, general manager financial services and risk at SAS Australia, was previously a partner at Deloitte, a firm which he says habitually uses data insights to shape the business. He believes that access to modern cloud-based tools puts the technique in reach of many more businesses. Further, he argues that data analysis should resonate particularly for accountants.

“The challenge accountants have is that a lot of the services they offer are fixed-price and commoditised, so this is all about a quest for value to their clients. How can they become a more meaningful coach or adviser? How can they use data to guide or inform them?

“A lot of clients in SMEs, even corporates, will tell you they have good accountants to fill out their tax returns, but not a good coach to provide them with competitive insight.” 

Franks says the challenge for accountants is to use data to identify and serve up meaningful information, for example using data visualisation to show clients how they compare to like firms in the same sector, and then using insights from the data to shape decision making.

“The typical model is ‘I’ll give it a go and see how it goes’. This is about getting evidence to give you confidence about the decision you’re making, and is an opportunity for accountants to offer more value. The technology is much more accessible, and with cloud deployment is in reach for more organisations.”

Data analytics also gives accountants a better handle on their own business, Franks adds. 

“The problem with practice benchmarking is it can be up to 12 months after", he says. "Organisations need to respond to changing conditions and market needs. The whole concept here is that data can be used forensically.

“It could be used to stress-test a business, to look at financial indicators and compare with a benchmark, then come up with a profile that allows you to compare your business to other businesses and identify any delta between where you are and other businesses are. We are all busy in the business; we should spend some time on the business.”

Now find the experts

The stumbling block for many businesses will be finding specialist skills. Exploiting predictive analytics requires data science skills. But data scientists are a rare breed, and data scientists with specific domain experience are even scarcer.

Several Australian universities are now offering postgraduate courses in data science and there are around 50 international university courses currently on offer, and that number is rising rapidly. And you don’t need fancy data technology to predict that graduates from those courses will find themselves in great demand.

Analytics meets the oil well

California-based oil and gas company Aera Energy, jointly owned by Shell and ExxonMobil, has deployed predictive analytics to save US$50 million in one particular oil field.
In a presentation at the SAS Forum in Australia, Paul Barnes, Aera’s former technology manager of innovation, said the company analysed data to learn that when a well failed, for whatever reason, it cost the company US$31,000 a day.

Data analysis revealed that production wells commonly fail because of a tubing split or pump failure, or when a well wasn’t fully pumped off.

By analysing sensor data, Aera refined the operation of the oil field, spotting and remediating looming problems before they led to a failure.

That cut the number of well failures per day from eight to three, generating US$50 million savings per year, not including the inventory loss associated with previous downtime.

The race for soccer data

Eddie Moore, CEO of Football NSW, has an accounting degree and knows the numbers. His organisation deals with 227,000 winter players, 23,000 summer players, 12,683 registered coaches, 8000 futsal players and around 5000 referees.

The not-for-profit organisation has a turnover of A$14 million, a staff of 57 and an IT department of one.

In the past, Moore managed by spreadsheet. It meant that although registrations were completed in March, the data wasn’t summarised until August, just as the winter competition peaked. “Now it takes 10 minutes with SAS analytics,” Moore says.

This delivers Moore a much finer-grained understanding of the football community, allowing the organisation to develop better retention and communications strategies and fine-tune its sponsorship pitches.

Moore explains that when it pitched to Rebel Sports, it was able to use visualisation tools to show where players lived, how close they were to a Rebel store, and later tell the company what proportion of players had used their Rebel coupons and how much they had spent. It’s using a similar data-rich approach in its discussions with local government about the need for new and better sports facilities, linking its data with demographic data to provide an evidence base about where investment should be targeted.

Read more:


Like what you're reading? Enter your email to receive the fortnightly INTHEBLACK e-newsletter.

September 2018
September 2018

Read the September issue

Each month we select the must-reads from the current issue of INTHEBLACK. Read more now.

CONTENTS