The onset of the fourth industrial revolution brought about by the emergence of the Industrial Internet of Things (IIoT) has led to the Hi-Tech and Manufacturing environment undergoing a massive upheaval.
However, these cybernetically connected physical systems have caused an explosion of data, presenting more challenges than opportunities for Hi-tech & manufacturing companies worldwide. Employees and workforces around the world continue to collate and store data in a range of homegrown formats. This has led to the creation of data silos, which has further lessened the overall value of production, data, process data, factory data, and the like. This has also caused businesses to halt their efforts to transition to being a data-driven company, with Forrester reporting that while 74% of organizations say they want to be data-driven, only 29% have successfully connected action analytics.
Actionable Insights: The Missing Piece of the Data Puzzle
Over time, the increasing amount of data sources will cause data overload, something which has already begun to take shape.
According to the World Economic Forum, 2.5 quintillion bytes of data are produced every day. This blistering pace at which data is gathered and stored has grown so fast that the best estimates say 90% of the data in the world today has been created in the past two years. This massive explosion of available data sources reduces the visibility of information data, which leads to inefficiencies and errors. With organizations struggling to generate value due to being unable to make sense of their data, they are slowly resorting to a hierarchical data structure that can help them derive actionable insights.
The hierarchical data structure, often called a data pyramid, has data at its foundation, information above it, inference in the middle, and actionable insights at the apex. To understand what actionable insights are, it takes knowing the subtle differences that separate these key terms beforehand.
The data that forms the base of the pyramid is the raw and unfiltered information, usually in the form of text and numbers, and primarily exists in databases and spreadsheets in computer-friendly formats. Information is called prepared data, or simply data that has been filtered, aggregated, processed, and structured into a human-friendly format to gain greater context.
Inference, or the knowledge part of the data pyramid, is a free-flowing mix of framed values, experience, expert insight, contextual information, and grounded intuition that offers a framework and an environment to examine and integrate new information and experiences. At the apex comes actionable insights, which are generated by evaluating information continuously and forming conclusions.
The Importance of Actionable Insights in Business
Actionable insights can go a long way in producing real-world benefits for companies. According to a recent Mckinsey Survey, high-performing organizations attribute at least 20 percent to earnings before interest and taxes (EBIT) over the past three years to their big data and analytics capabilities. Such was the case for Intel. Intel’s factory equipment live-streams IoT-generated data into their MES integrated, big data solution. The analytics solution uses data to recognize patterns and detects faults through visualization apart from the standard deviation of established metrics. On a continuous basis, patterns emerge, and tendencies are recognized to spot situations where immediate attention is required to prevent system breakdowns. Such predictive maintenance reportedly reduces reaction time from 4 hours to 30 seconds and cuts costs. In 2017, thanks to big data and IoT, Intel predicted saving $100 million.
Big data can also enable process optimizations. Take the case of AlayaCare, a HealthTech homecare company. AlayaCare analyzed different yet connected data sets to produce an algorithm that successfully predicted seniors’ negative health events at home. The senior patients would take a series of vitals every day through connected devices. The algorithm then recorded and analyzed the vitals and combined that with the clients’ ICD-10 diagnosis, age, and gender. They successfully reduced hospitalizations and ER visits by 73% and 64% amongst a chronically ill patient set.
Big data and related insights can also lead to a permanent change in seemingly docile business practices. In the case of Caterpillar Marine, this is exactly what happened. As a standard after-sales procedure, they were requested to show the analysis of how hull cleaning impacts fleet performance. Their big data solution compared sensor data from ships with and without cleaned hulls. Then, correlations were found between the fleet performance and the investments made in hull cleaning exercises and practices. With this data, Caterpillar concluded that their client needed to clean hulls more often (every 6.2 months, not two years) and that related investments paid off.
The insight that can be gained can become even more complex when you include human sentiment in the mix. But by following the same principles of driving change through actionable insights, even human emotions can become a source of data for use. Take the case of Amazon comprehend. The program uses sentiment analysis to break down customer reviews on the aspect model to rate the review as positive, negative, or neutral. This allows for the analysis of thousands of customer reviews to give the seller some actionable insight into which aspects of its product are well received or need improvement.
How to Use Data for Improving Business Outcomes
Harnessing unstructured data is the first step in using data to improve business outcomes. To harness the potential of unstructured data that is in place in your organization, the following needs to be established:
Managers need to have a clear understanding of the desired outcomes. This will help business stakeholders remain focussed on the data points needed to make key decisions and avoid becoming bogged down by the masses of data available, most of which may not be relevant for them. From there, companies can expand to additional objectives, eventually scaling up to a fully data-centric entity. Once objectives are decided, we have to take stock of the data already being generated in the system. The next step is to contextualize the data with metadata. Metadata is important to provide searchable and contextualization to data. This helps in making data relevant and usable across platforms as well. After the data has been sanitized and validated, the arguably most valuable part can begin: drawing insights and taking action. By making data visualizations and leveraging the power of NLP and AI, we can make the insights compelling, and teams can transform those insights into strategic business decisions and actions.
Using AI, ML, and NLP to extract insights:-
With access to useful data, the next goal is deriving insights from your data – and this usually requires a lot of effort. AI and ML help in creating new methods for doing so. Using training sets, AI can be primed to take advantage of recurring data to derive insights, which traditionally takes a large amount of manual labor to generate. Similarly, data quality can be achieved by using ML as well. Where algorithms can be used to detect outlier values, duplicate values, and missing values, helping normalize data. Prescriptive analytics is another area that is already being used to reap benefits. Traditionally, big data decisions were based on past and present data points, generally resulting in linear ROI. Prescriptive analytics, leveraging AI, provides company-wide, forward-looking strategic insights helping to advance the business.
NLP adds another layer of complexity and enhances our analytical capabilities. The combination of NLP and machine learning enables organizations to gain insights from data sources that are traditionally unstructured (e.g., emails, memos, chats, legal documents, internet forum interactions) in a way that was not possible before. Using methods such as text clustering, topic modeling, named entity resolution, and relationship extraction, data that has been previously dark transformed into insightful information.
Big data has tremendous potential in the Hi-Tech and Manufacturing world due to the agnostic nature of the technology and the domain itself. Some examples of the usage are below:
Understanding performance across multiple metrics
Across Hi-Tech and manufacturing companies, various metrics determine a company’s performance in terms of cost, efficiency, and responsiveness. Big data helps connect these datasets and variables to present a complete picture.
Managing supply chain visibility and complexity
Today supply chains are bigger and more complex with multiple variables, both internal and external. Simple logistical problems become complex when you add in the variability of transportation routes, traffic patterns, and vehicle population. Big data helps handle the scale of these operations and the complex interdependencies of these variables.
Big Data can detect machine performance in real-time, allowing companies to set standards of production. As production continues, big data analysis allows for the recognition and prediction of patterns, hence making it possible for enterprises to detect anomalies and predict production downtime, resulting in cost savings as well.
Data is only a piece of the puzzle. In addition to data, generating actionable insights depends on the harmony between the disciplines of operations executives, unit managers, process engineers, IT teams, and the like. Forward-thinking companies have already started preparing for their journey to reorient their business processes through advanced analytics. As the information age matures and organizations progress into the fourth industrial revolution, they will stand to see much higher benefits by analyzing, understanding, and leveraging data to inform their decision-making process and make their transition to an intelligent world more seamless.