Over the last few months, Google revealed just how important machine learning is to its future. In October, along with its Q4 earnings, Google announced further investment into machine learning – suggesting that it would be essential to informing the forthcoming direction of the business and its services. Following this, the company also announced that the software in its new machine learning system, TensorFlow, is going to use open-source code.
Machine learning isn’t just a major focus for Google, other Internet giants including Microsoft, IBM and Facebook are also heavily invested. Yet, more interestingly, this approach is beginning to be echoed by offline-companies too. Automobile manufacturer Toyota recently announced its interest in machine learning – suggesting that it is fast becoming a trend that CIOs in all industries will have to start paying closer attention to, or risk being left behind.
So what is machine learning? Machine learning is the next step in the big data evolution. It uses intelligent, computer-generated algorithms to examine data and determine hundreds of thousands – if not more – hypotheses and their potential results to predict, recommend and eventually make automated decisions. Working from a company’s existing data, it can identify the next best action in any scenario, while actively learning from the outcomes to ensure continued success - be it to develop new or to improve existing services or processes.
For companies like Google and Microsoft, machine learning has always been, and will remain, a critical part of its core business. Many of the algorithms and machine learning technologies were born out of and perfected by the same Internet giants now pledging further investment. Due to these businesses fundamentally relying on identifying correlations in datasets and extracting value from data, these companies were some of the first to train, test and develop such algorithms, and gather real-time feedback.
But now with the explosion of big data from and within all areas of business, companies are entering a new industrial revolution. And it is no longer just the Internet giants that are looking to machine learning to take advantage.
More than big data
Big data is now a major focus within the enterprise. In a recent survey of IT and business leaders by Gartner, 75 per cent of companies said they were investing or planning to invest in big data within the next two years.
Yet while the big data hype promises plenty - enhancing the customer experience, streamlining existing processes, and reducing costs - more often than not, these projects just bring disappointment; a lot of investment and not a very convincing ROI.
Sure, data is an important asset; but what is more important is how it is actually being used.
Unfortunately the current analytics methods being used by the typical enterprise are underutilising the vast amounts of data being generated. Companies are still analysing existing datasets – sometimes even manually - only to then use the information to support human decision-making.
Whether guided by data analysis or not, human decision-making is unavoidably prone to inherent bias based on experience or preferences, and is naturally slower than machines. Unlike machines, humans are limited in their ability to make sense of complex datasets from multiple sources, identifying relationships between data points and in producing hypotheses. In effect, individual decisions are made based on simplifications and generalised visualisations of massive amounts of intricate, interrelated data, obscuring the data’s true value.
This fundamentally limited approach is not even close to the machine learning-led methods of the likes of Google, Microsoft and Facebook. But just because these giants are using machine learning, it does not mean that it remains the preserve solely of the largest companies in the world.
Access big data’s real business value
Machine learning is highly accessible and offers the most efficient way to start deriving real business value out of data. Many operational decisions could – and should - be delegated to machines in order to make processes operate at the speed and precision needed to help grow a business.
For example, marketing departments would be able to better analyse customer data in order to determine how to best personalise and deliver next best offers, or to reduce churn. Similarly with operations, where there are currently many redundant processes still in use, such as manual content moderation in online service businesses, or banks retyping data from hand-written forms, or sending ground inspections to monitor large infrastructure objects for utility companies. All of these tasks and processes are costing many companies huge amounts of money for no reason, other than it being the way it has always been done, and a lack of appreciation of how technology could help.
There’s still a long way to go
Yet there is still a long way to go for the enterprise. Introducing new technologies into a company requires not only a successful technical integration but also a change in culture and business processes – which is often demanding of both attention and time.
This is because the whole idea of working with probabilistic results and trusting decisions made by robots is challenging for traditional companies. It requires establishing a culture of experimentation, and learning to run A/B tests as part of a normal work process.
Unlike for the naturally data-reliant and constantly innovative Internet giants, this will not be an easy evolution for many. Not to mention the likely lack of internal machine learning expertise.
These companies will instead need to learn to work with external providers of machine learning - and build teams that are qualified to manage and compare external solutions; and bridge the technical with the business.
It is no longer just the Internet giants that stand to gain from machine learning technology. As data explodes both in size and focus within the enterprise, CIOs in all industries risk being left behind if they fail to embrace machine learning and algorithms.
Analysing data to inform decision support by humans is no longer enough. To unlock the real business value of big data, companies should apply machine learning to collect, analyse and act upon this data - calculating the most suitable action based on the results from previous activities. This will not only allow companies to make informed business decisions, based on evidence and not opinion, but to enjoy truly measurable and additive results.
Jane Zavalishina, CEO, Yandex Data Factory
Image Credit: Sarah Holmlund / Shutterstock