The “Internet of Things” is made up of connected sensors and devices for just about any type of equipment, from phones to household items to heavy machinery. The Internet of Things makes it possible to monitor equipment in real time, as well as use predictive analytics to identify potential disruptions and malfunctions before they occur. The combination of monitoring and analytics can help businesses save huge amounts of money in avoiding defects and service disruptions.
This post is based on an upcoming white paper called “It’s not about Big Data, it’s about Big Insight” by Achim Granzen, Architect in the Data Insights Center of Excellence, with contributions from Ken Collins, Fidan Boylu, Philip Reilly, Benjamin Wright-Jones, and Delbert Murphy.
With new generations of connected devices appearing everywhere, the new world of data described in the post “Enabling Big Insight with Big Data (Trends and Insights)” is even more rich with possibilities. Whereas customer insights employ a mixture of traditional and new data attributes, the data produced by the Internet of Things can only be described by the flexible attributes associated with the new world of data. The Internet of Things produces data as diverse in format and structure as the devices themselves, detailed but often messy, plentiful, and generated outside organizational boundaries (physical as well as logical boundaries).
Gaining insight from this type of data is nearly impossible if managing and analyzing with traditional approaches, especially when time to insights is considered. However, this data holds information that is critical to an organization, as it can help lower damage and repair of equipment (predictive maintenance), prevent health and environmental hazards (event and incident monitoring), increase quality of services and products (performance management), and help optimize service operations (route optimization).
Sometimes, scenarios touch or even overlap, however it is ultimately the business goal that determines the scenario. Even consumer devices such as mobile phones and tablets are considered part of the Internet of Things, as they are essentially connected devices. Better understanding of the phone user’s – the customer’s – location, preferences, likes and interaction history enables customer insight. Monitoring the handset for early signs of equipment failure and detecting coverage gaps by location-specific signal strength analytics provides operational insight in terms of predictive maintenance and (network) performance management.
A key challenge for any operational insights scenario is the wide, potentially regional or global, distribution of the devices to be monitored and analyzed, Before the availability of large scale cloud-based services for ingesting, storing, and analyzing data, many of these scenarios were cumbersome to realize, and in many cases not feasible or plainly impossible.
Similarly, the huge variety of data types and formats in the past – each device having its own format and structure – made it extremely difficult to process data from such devices, let alone employ predictive analytics on a large scale of attributes and parameters.
Now, key technology capabilities can enable a typical analytics solution for operational insights, including:
An analytics solution for operational insights will mainly deal with externally created data, of a large variety of structure and format. Hence, a cloud analytics platform that enables this is a core component of any architecture. External data can be processed through a cloud based data ingestion, storage and management system, with support for streaming data. Most analytics such as predictive maintenance can be performed directly in the cloud, as it is normally not required (nor feasible) to bring that data on-premises.
Internal data from data warehouses and operational systems, which might be required to augment the data analyzed (for example, serial numbers and other basic equipment identification data) can be uploaded into the cloud analytics platform and stored in either structured or unstructured form.
To close the loop, analytics results need to be made available to the business, for further exploration and visualization by business analysts, or for inclusion in management dashboards and other reporting systems.
Figure 1. Schematic Architecture for a Primary Cloud-Based Analytics Platform
The above describes a typical architecture in a scenario where the majority of data is created outside of an organization. In scenarios where a large share of the data is created and available within an organization (i.e. not geographically distributed), the architecture will more closely resemble a hybrid approach, as discussed in the post “Customer Insight—The New Age of the Customer.”
The key theme in predictive maintenance is risk management, particularly reducing the risk of potential high cost events such as machine part failures or quality issues, by being able to take corrective action before the event occurs.
Predictive maintenance as a business scenario is not a newcomer on the stage, however the recent development and availability of cheap, interconnected sensors opens up a large number of use cases and scenarios, and the availability of cloud-based data ingestion and analysis dramatically reduces the need for complex infrastructure. Both trends make predictive maintenance much more pervasive than in the past, literally touching every household (for example, smart meters, energy use, and so on).
Avoid costly asset downtime and reduce maintenance costs
Contoso was struggling to deal with vehicle failures within its fleet management system. As a result of unexpected malfunction and the time wasted on the maintenance, the company was losing money and resources in correcting the issues after the failures happen.
By using extensive amounts of telematics data that are collected through on-board sensory systems integrated with environmental and product data, Contoso is able to use predictive machine learning methods to perform health tests on the current operating conditions of the vehicles and identify vehicles that require immediate maintenance action.
Reduce warranty claims due to unexpected failures
Contoso was struggling with frequent warranty claims on a certain line of its products due to unexpected failures.
Predictive maintenance now identifies when equipment in the field is likely to fail or need maintenance in order to predict future warranty claims costs and maximize uptime for equipment used to deliver service. By using predictive maintenance, Contoso is able to eliminate bad publicity and minimize resulting lost sales from negative customer product reviews.
Improve inventory management of spare parts by predicting failures
Contoso was having problems with predicting the inventory for its expensive parts that may have long lead times, leading to high inventory cost and risk of low inventory levels. The company needed just-in-time inventory management to reduce costs.
Using predictive analytics capabilities, the manufacturer can determine if certain products are likely to fail and then analyze the financial implications of the failure. The analysis also shows where the failures will occur and what the demand in a given region will be for the replacement parts. The manufacturer can then ensure that the correct supply of replacement parts can be available at the appropriate time.
Managing the performance of an organization, or of a part of its business, goes much beyond the classic financial reporting, best described as financial performance management. Using both descriptive and predictive analytics capabilities to track, analyze and improve the performance of a system (such as component production or discrete manufacturing) has a direct impact on the bottom line of an organization where every element of “waste” can attribute to significant costs.
Improve Production Quality Assurance and Yield
Contoso was having problems with its certain line of products coming out defective with some parts or the products not meeting quality standards. These needed to be recycled or fixed, which led to loss of time and effort and caused delays in the assembly and shipment of the products.
By using an analytics approach to examine the sensor data collected from the assembly line, Contoso is able to accelerate root-cause analysis to determine the source of the problem and sustain quality standards by understanding the key predictors in the assembly of the products that eventually leads to defects.
Safety is the key objective for event and incident management and monitoring, with potentially severe implications not only financially but in many cases to human lives. Especially in the natural resources and energy industries, health, safety and environmental (HSE) is a key concern in every aspect of the operations: problems can affect the lives of thousands, even millions of people.
Similarly, organizations secure locations, buildings, and events not only to prevent financial implications (intrusion, theft, damage) but also to address the safety of the public as well.
Better protect Health, Safety and Environment
Contoso was unable to achieve a near-real time view of their equipment condition, which is distributed over a large geographical region with difficult access. This led to situations where equipment malfunctioning was going unnoticed for a considerable period of time, causing severe HSE issues. With HSE being a key performance metric for Contoso, it was important to get an accurate picture of events, and mitigate any event that has happened.
Using a cloud-based remote equipment monitoring system, Contoso’s operations control center is able to get accurate and timely information on all of their equipment, visualized in an easy-to-understand format, and with capabilities to take direct action. As a result, potential hazards to the health and safety of their workers, as well as to the environment, can now be addressed much quicker, resulting in lower severity and an overall improved HSE KPI.
Improve Situational Awareness, Security and Tracking
Contoso City was challenged in analyzing data from various streams of public safety systems in a timely and consistent manner to present law enforcement and crime prevention officers with a concise description of a crisis situation, preventing them from acting quickly and precisely.
Using a consolidation and comparison system which can ingest a large variety of data types from video to license plate readers, and employing monitoring, visualization and analytics capabilities, Contoso City can now provide real-time analytics and improved situational awareness for the men and women on the front lines of law enforcement and crime prevention, which helps to further enhance public safety outcomes for Contoso City citizens.
Going well beyond the classic “Travelling Salesman” problem, route optimization applies to a multitude of scenarios where things have to be moved from A to B in the most optimal, most cost-effective, or most secure way: it applies to moving people (transportation from buses to planes), discrete goods (from parcels to containers) and resources and utilities (gas, water, electricity).
Besides “Where” (to optimize the transport mechanism or medium) – the How is equally important. This is particularly evident when dealing with power grids and networks, as there are many more elements involved (power plants, transformers, distribution stations, meters) than just the power lines.
Optimize Power Grids and Networks
With ever increasing pressure to conserve energy and reduce energy loss and peak loads, Contoso was struggling to analyze load patterns and distributions, and encouraged selected consumers to shift to off-peak hours.
By employing a network of smart meters connected to a cloud based data ingestion and analysis system, Contoso can monitor energy consumption and the effectiveness of off-peak special offers in real time, and take other corrective measures as required. The system also allows Contoso to analyze and predict electricity usage patterns much more detailed and accurate than ever before.
Optimize Traffic and Goods Movement
Contoso Airlines was challenged to increase their revenue per seat-kilometer (RPSK), primarily by optimizing ticket pricing, capacity and scheduling; and additionally by increasing ancillary revenue (non-ticket revenue). For example, the ability to accurately forecast seat upgrade demand for a given flight – based on a combination of factors such as flight, region, period of the year, loyalty program status, early/late purchase, etc. – would allow them to set the price of the premium seat accordingly. Contoso was unable to do a proper analysis based on data, as they lacked analytics and data preparation capabilities.
Using a cloud-based analytics solution leveraging Machine Learning, Contoso Airlines is now able to optimize the decision-making process for schedule optimization and for ancillary revenue. For example, they can priority rank customer attributes according to their relevance on customers’ propensity to upgrade their existing seat.