The Rise of Machine Learning to Manage Dark Data

For years we have been wasting time and resources pushing aside the benefits of dark data, unconsciously dismissing the great potential it can offer a business or industry.

Dark data is a type of unstructured, untagged and untapped data that has not yet been analysed or processed. With 80% of data being classified as dark data, there is undoubtedly enough information for companies to leverage to their advantage. Therefore, it is now time to bring light to the unknown gold mine of dark data.

For many businesses, understanding the sheer amount of dark data can be overwhelming and time consuming to manage. Businesses may use excuses like legality issues, workflow disruption or architectural costs as to why it has been reluctant to harness dark data. It may also fear that getting access to dark data can invade valuable time which could be used for other tasks and disrupt employees with new ways of operating. Of course, disruption can be kept to a minimum when implemented correctly with the right tools.

The dawn of machine learning

For most companies today, transforming unstructured data into readable assets involves processes that are mostly manual. To create better value, businesses need to automate these processes and free up resources from mundane tasks, and this is where machine learning comes in. Machine learning is an application of artificial intelligence (AI) that provides systems with the ability to automatically learn and accomplish the equivalent of continuously running programmes in a fraction of the time.

Businesses can utilise machine learning to build models that work in the specific business function and industry. In the case of dark data, the process of learning begins with data observations, in order to look for patterns and make better decisions in the future based on previous examples. Typically the system alerts business users to exceptions, and remembers each time they address those exceptions so that it can offer a solution the next time a similar event occurs. If users keep accepting the recommended solution, the system will learn automatically.

Structural changes are required when implementing machine learning, which costs time and money. But it’s worthwhile long term and the business benefits will guarantee a high return on investment.

Uncovering the benefits

At a first glance, dark data can appear obscure or unhelpful but when its approached correctly, it can unlock business value and boost the bottom line. The key to uncovering dark data’s secrets lies in the ability to understand the relationships between seemingly unrelated pieces

of information. Machine learning plays a critical role in helping businesses interpret information correctly and reveals a host of patterns or insights that would have otherwise been missed.

In a business climate where data is competitive currency, dark data is valuable as it allows businesses to learn about every aspect of their operations. The larger the dataset, the more accurate the analysis. In fact a recent survey started that, 76% agreed that businesses who have the most data is going to “win”. By expanding the scope of information that businesses analyse, new levels of innovation and resilience can be leveraged to create a competitive advantage. Without learning to integrate new forms of data into organisational operations, businesses can quickly become stagnant compared to its competitors.

Dark data can also provide a complete view of consumer habits, product usage and overall performance. The ability to act on insights from this customer data is a key requirement for businesses wanting to compete in the digital age. For example, sales departments could use collected data to create better customer sales profiles with an understanding of their buying behaviour and product preferences. Not only will this allow businesses to develop knowledge on what its customers might be looking to buy and start developing said products to suit their needs but it can also help enhance customer retention.

Bringing light to dark data

The thing about traditional machine learning is that as complex as it may seem, it’s still machine like. It largely needs domain expertise and human intervention, meaning in most circumstances it’s only capable of what it’s designed for. For businesses looking to go one step further and make its data extraction processes fully efficient, deep learning could hold more promise.

In contrast to more traditional approaches focused on text only, deep learning aims to extract relations conveyed jointly via textual, structural, tabular, and even visual expressions by using new techniques to automatically capture the representation (in other words the features) needed to learn how to extract relationships from richly formatted data. The biggest advantage to deep learning is that it tries to learn high-level features from data in an incremental manner. This eliminates the need of domain expertise and hard core feature extraction.

Overall, it’s worth thinking about dark data as unfulfilled value. The key to monetising dark data lies not only in gathering it, but in analysing it to discover patterns and putting the insights to use. By utilising new technologies around machine learning, specifically deep learning, businesses can join structured and unstructured data sets together to provide high-value results which in turn can be used to generate profit. When done correctly, the business benefits will easily outweigh the costs involved in mining dark data. By unlocking and allowing machine learning to uncover dark data, businesses can reveal new insights and knowledge that will yield a greater competitive advantage.


About the Author

Laurent Louvrier, CEO at SuccessData. A seasoned technology executive with experience across financial markets, engineering, business management and product strategy. Focus on identifying business opportunities around fintech and leading teams to design great products to address them. Strong passion for innovation and entrepreneurship.