While manufacturers are captivated by big data, and how to collect and analyze it, getting a handle on “small data” is the critical first step.
As Jim Heppelmann, president and CEO of PTC, and Professor Michael Porter of the Harvard Business School define in their HBR article, How Smart, Connected Products Are Transforming Competition, the combination of intelligence and connectivity gives products new capabilities and allows them to share massive amounts of data about their condition, environment, and performance.
Today, for example, sensors, embedded microprocessors, and connectivity within mining equipment generate vast streams of data and allow fleets of machines to operate with increasing safety, efficiency, and autonomy (see Strategic Choices in Building a Smart, Connected Mine).
Given the potential value of this wealth of product data, companies may be tempted to start collecting and analyzing as much information as possible in a swing for the fences. As the cost of adding sensors to products and storing data falls, there seems to be little downside to simply amassing data and then searching in the haystack for anything of value.
Big data has three common characteristics, as defined by Gartner, high-volume, high velocity, and considerable variety, which all have an important impact on the time and effort required to extract insight and then to move from insight into action. There are three challenges to the big-data-first approach:
- First, the more data, the more complex and uncertain the process. Companies must centralize and index data from products and other sources and hire data scientists with deep product and process knowledge—who are scarce and in high demand—to do the analysis needed to generate actionable insights. This work is highly process- and device-dependent and there is no “one-size-fits-all” solution.
- Second, there are significant security and privacy risks from amassing product data. Target filed its data breach costs at $148 million, but with the resignation of its CEO in May 2014, and subsequent drops in profit and customer sentiment, the actual costs were much higher. Product data clouds will become new targets for hackers, and depending on the type and volume of data being stored, risk can increase.
- Third, most companies find the ultimate limiting factor is not technology, but the cultural and organizational change required to transform business processes based on new data insights. It takes time for the organization to trust product data and view the insights generated from complex analysis as truth. The time required is multiplied as the complexity of the analysis, and thus uncertainty of the process, increases.
So where to begin?
Start small and focus on uncovering specific use cases across the enterprise. Stakeholders in every business functions—from R&D and marketing to sales, service, and operations—must each identify the connected product data and other system data that would help them make better decisions faster and enable more intimate customer relationships.
For example, small data from a smart, connected product can signal when and how a specific part has failed, allowing the dispatch of the right technician with the right part, thus improving the first-time fix rate.
Small data can be effective in automating business processes—like dispatching a service technician or automating the reordering of product consumables,—and gaining insights into the use and performance of smart, connected products like capturing who purchased your product or when a specific feature is activated. Simple rule-processing and the ability to interactively provide feedback and updates to smart connected products in the field will provide dramatically shorter time to value than a larger, big data project, and continue to deliver value for companies and customers over time.
Case in point: A manufacturer of sophisticated blood testing equipment used a physical limit switch as a last resort to prevent severe damage to the motion control system. In theory, that switch should never be hit, but by implementing a small data system that included notification whenever one was activated, the manufacturer was surprised to find this happened quite frequently.
The issue was easily solved by remotely updating the calibration procedures of the motion control subsystem, and the next generation product was redesigned. Something as basic as simple event detection and remote software updates avoided expensive service calls and improved customer satisfaction.
By starting small, manufacturers can capture the opportunities that exist from small data and allow this new way of thinking to permeate the organization. Then, as companies mature in their capabilities and understanding, they can expand their use of data and analytics to provide new business insights (e.g. seasonal sales of certain products in a vending machine located in a specific location), optimize business processes (e.g. optimize aircraft flight routes to minimize fuel consumption), create highly differentiated offerings and define entirely new business models (e.g. moving from an equipment sale to a ‘cost per valid result’ in the blood testing industry).
The key is understanding and creating value from data is going to be an iterative process and one must start with a platform that makes it easy to adapt and add new applications as the business needs evolve over time.