How to Use Big Data in Manufacturing
A McKinsey Global Institute report advises, "Despite early advances, manufacturing, arguably more than most other sectors, faces the challenge of generating significant productivity improvement in industries that have already become relatively efficient."
Nobody disputes manufacturing's increasing competitive challenges and quest for improved productivity. However, these challenges are no small feat as most manufacturers have already adopted lean manufacturing, ridded themselves of non-value added activities, optimized the supply chain, streamlined and automated business processes, and squeezed costs to the point where there's just no more fat. So where exactly is the next productivity or business model break-through opportunity?
It's in the next information renaissance being brought by the combination of Big Data and the Internet of Things. These two synergistic and disruptive technologies can uncover and leverage data that is absent, hidden or unappreciated in order to deliver actionable insights capable of changing production processes, customer engagement and business models.
To share how, I'll avoid the technology theories and get to some real world examples with tangible results that can be replicated by just about anybody willing to explore this opportunity further.
Big Data Examples in the Manufacturing Industry
Big data is not a software application, technology tool or something you install. Sure there are big data tools but putting technology ahead of strategy is a recipe for disappointing results.
Instead begin with strategy, and to spur your strategy you will need some creating thinking. Here some examples to stimulate that thinking in a way that can dovetail into designing a big data strategy.
Improve Demand Generation
Small improvements in demand gen and product forecasting models mean fewer stockouts, less idle inventory, decreased cash flow requirements and most importantly, satisfying more customers that want your product – when and where they want it.
However, manufacturing demand gen models that only look at purchase history are generally not considering fluid market conditions and real-time customer demand. Big data offers a unique opportunity to look forward when estimating product demand.
For example, record label EMI turned their business around by shifting demand generation and product forecasting from tenured executives applying gut feel experience to big data analysts tapping into listening patterns on online music streaming services to understand which music is resonating by region and with which subcultures. They are also teamed with music web-based and mobile apps providers, online "second screen" collators and music aggregators to acquire the data and create models which show consumption trends and measure demand in real-time. Sometimes called mood mapping, manufacturers and retailers can buy or acquire consumer reactions which correlate to market demand.
Manufacturers and retailers can also use social listening tools or acquire feeds from social networks to measure consumer feedback volume, reach and sentiment for their products. With live data in hand, they can then interrogate the data by geography, consumer segment and other characteristics that permit modeled interventions. The 24 by 7 constant streams of opinions, likes, dislikes, sharing and social propagation offer manufacturers and retailers a real-time audience to test variations in promotion, pricing elasticity, content, packaging and other variables in a way that correlates testing conditions with consumption and builds a predictable demand generation model.
More New Product Successes
Introducing new products is a high risk venture for manufacturers. Combining online ideation and big data reduces this risk, accelerates product innovation, lowers R&D expense and grows revenues by delivering products that customers most want.
Using online ideation to engage consumers for their ideas and critique the ideas of others, manufacturers are able to test R&D concepts faster and with substantially larger samples of customer feedback. Real-time access to large volumes of customer opinions also allows one-off test conditions and accelerates successive iterations so that new product innovation cycles can be substantially reduced. When consumer feedback can be acquired in minutes instead of weeks, manufacturers are afforded more testing scenarios and valuable What-If analysis.
Again referencing the music industry, music labels routinely arrange for leaked songs to test market acceptance and determine whether the launch of a new tune is ready for prime time or needs some more refinement. While this concept has always been replicable to other industries, the addition of big data allows manufacturers to tap into social media and other online channels for real-time analysis, increase the data population for improved confidence levels and integrate unstructured data to bring measurability to new types of information.
These collective capabilities accelerate market analysis, enable measurement specificity by key variables (geography, demographics, customer segment, etc.) and help manufacturers deliver new products which are more eagerly embraced by customers.
More Reliable Supply Chain
Supply chain interruptions are costly. It only takes one downstream supply chain partner to stumble and thereby impact revenues and damage the manufacturer's relationships with retailers and customers. To mitigate this complex threat, some innovative manufacturers are using big data to cull information that may suggest a supply chain partner is at risk.
Manufacturers have long used structured data such as on-time deliveries, payment cycles or the use of factored receivables to gauge partner health. However, when also including semi-structured or unstructured data from external and online sources – such as published financial analyst recommendations, media reviews, filed litigation or various forms of insolvency protection to name a few – manufacturers can be afforded valuable time for advanced planning and contingency measures when a partner faces financial distress.
This technique can also be used to avert public relations nightmares brought on by partners using illegal or culturally unacceptable business practices. For example, consider recent consumer backlashes directed toward garment makers who (allegedly) unknowingly had partners using child labor or dreadful wages in distant lands. In the tech industry, consider Apple's missteps by not identifying early or responding timely to the Foxconn fiasco born from subpar labor conditions in China.
Big data offers a new opportunity to keep close tabs on your suppliers and creditors.
Some Internet of Things Examples
The Internet of Things describes a world where everyday objects are embedded with sensors and radio tags which give them network and Internet connectivity for capturing and transmitting data. RFID tags are the favored wireless transfer technique, but other tagging technologies include barcodes, QR codes, near field communication (NFC) and digital watermarking.
Despite being what is likely the broadest catch-all phrase ever coined in the technology industry, the Internet of Things offers real value to the manufacturing industry. Sensors can produce extremely high volumes of machine generated big data and manufacturers who trace their products sensor-based data can understand customer preferences, product utilization, wear and tear, operational problems and host of performance measures that facilitate product maintenance, replacement and innovation.
Let's get to some practical examples.
Product Innovation
I had a chance to meet a driver from the Lotus F1 team at the last Convergence conference. He shared with me that in Formula One racing there are actually two races going on – the race on the track and the race off the track. The off track race is one that lends many lessons in how manufacturers can apply machine generated big data to improve their products and services.
These race cars are laden with sensors that detect flexing, vibration, load, wear, temperature and many other measures which impact machine performance. This data gives teams the opportunity to not just understand how various gear impact performance, but to use the data for modeling improved performance. In fact simulation testing using real data applied to various combinations of components has delivered the single biggest impact to improved race car performance and winning more races.
Comparing on the track equipment performance measures with more stagnant product attributes such as component age, composite materials, manufacturing methods and so on enables detailed understanding of equipment deterioration pace or break point thresholds as well as optimization opportunities. This type of rich data analysis can help most manufacturers find ways to make better products at less cost, improve product reliability or become more environmentally friendly.
These lessons have been extrapolated to many car manufacturers who now embed their vehicles with sensors and microprocessors which capture data for maintenance and repair purposes as well as R&D innovation.
Other types of manufacturers can apply similar methods to detect usage patterns, variables which negatively impact performance and conditions which contribute to hazards and then model this data to experiment and determine materials and process techniques which increase performance or useful life.
Adding to this big data equation, the analysis will be improved by including customer attributes such as region (i.e. cars driving in the snow or in the desert), road conditions (the smooth roads of one area compared with the potholes of another) or purpose (a pickup truck used for farm work compared to one in an urban setting). Customer use cases, environmental conditions and product patterns may materially affect modeling results.
The ability to perform simulations using machine generated big data enables manufacturers of sophisticated equipment to model different product components, swap parts, change configurations or test other variables in order to predict consequences and outcomes. This learning not only accelerates innovation resulting in superior products, but lends itself to downstream processes such as warranty policies, field service operations or even product recalls.
Product Lifecycle Management
Machine based sensors give manufacturers real time visibility to recognize how a product is being consumed, when maintenance is needed, when the product is approaching the end of its useful life or when the useful life is over.
Machine sensors populating data logs allow manufacturers or service providers to identify low oil levels, high temperature readings, excess waste, unusual vibrations, extra noise, declining production levels or runtime patterns which suggest performance degradation. With this type of information the regular dispatching of technicians for maintenance visits based on period intervals can be replaced with real-time maintenance delivered when it's actually needed.
Real-time machine health information that shows how products are degrading also gives service providers a window to schedule maintenance during low impact hours or operations. This advanced planning represents a sea shift change from being notified only when a machine becomes inoperable and unplanned downtime along with a hit to production output are simply part of the scramble.
Xerox is a good example of a manufacturer that monitors millions of copiers, printers and other devices located at customer sites. Rather than depend on customers to deliver accurate equipment feedback or incur the high costs to dispatch technicians for onsite assessments, the machines transmit log data to a central data warehouse at Xerox. Automated business rules identify equipment in need of maintenance or situations to escalate for human consideration and a predictive algorithm determines which equipment may stop operating in the future. Remote maintenance management has resulted in dramatically fewer trips, and ensuring the right people with right tools and parts are in-hand for each trip.
Products ranging from race cars to printers, or consumables to industrial equipment, can deliver streaming data which enable manufacturers to manage their products health and utilization, increase customer value and append R&D for new product innovation.
Summary
Manufacturing business processes incur a profound change when product utilization is viewed in real-time. For example, product replacement becomes predictable, inventory control becomes continuous and product forecasting becomes more accurate.
Sensory-based data flags deviations, identifies patterns and permits simulations that facilitate timely and low cost experimentation to improve product performance, quality and longevity. However, this is only the tip of the ice berg as when this machine generated big data is thoughtfully linked with customer attributes and behaviors it becomes a valuable source to deliver real-time information about consumer preferences, customer intent and market demand.
According to Gartner, there are now over 26 billion devices on the Internet of Things. It's a relative certainly that all viable manufacturers will eventually leverage sensory-based and machine generated big data. What's not so certain is which manufacturers will have the foresight to tap into this technology while it can still deliver advantage over their competitors who choose to ignore what may be the most valuable product and customer information source available.