This technical information has been contributed by
It's Not the Data—It's What You Do with it that Counts
Shannon Precision Fastener, a user of the Sight Machine data platform, performs in-house heat treating.
Image courtesy of Shannon Precision Fastener/Sight Machine.
Many manufacturers are already generating and collecting large amounts of data that's either erased or forgotten. Now, a young company with roots in the U.S. manufacturing heartland and Silicon Valley has a plan for translating that data into operational insights that can lead to big improvements in quality and overall equipment effectiveness (OEE).
Sight Machine, a venture-backed startup with offices in Ann Arbor, Michigan and San Francisco, wants to bring manufacturers onto the Industrial Internet with a web-based platform that does more than store and retrieve mountains of data. The company's Internet-enabled software-as-a-service (SaaS) platform, also named Sight Machine, allows users to collect and store their vision data, and then goes a step further by transforming that data into insights that can be used to improve operations and quality. "It's a radical leap in vision technology," the company states on its website.
Sight Machine uses flexible algorithms to analyze data from industrial cameras, sensors, and factory software systems and, with an eye toward helping factory personnel get "the right information at the right time," enables them to securely, remotely access the information that they want.
"The art in doing this well, and the benefit to the customer, is putting the one or two things they really want to know on a screen, so that an operator working on a line knows what to do," said Sight Machine President and CEO Jon Sobel, in a recent phone interview. "The big data part is cool, but it's just an enabling technology. The only thing that matters is 'can you improve your production?"
"The art in doing this well, and the benefit to the customer, is putting the one or two things they really want to know on a screen, so that an operator working on a line knows what to do. The big data part is cool, but it's just an enabling technology. The only thing that matters is 'Can you improve your production?'"
—Jon Sobel, president and CEO, Sight Machine
The cloud-based service allows users to inspect raw materials, check parts in process, and share data with customers. Data can be accessed and analyzed in real time, stored for as long as needed, and then re-analyzed and retrieved at any time. Manufacturers can run correlations over time, track trends, and identify root causes. They can also retrospectively analyze previously acquired data as new variables of interest arise.
"Even if manufacturers are collecting data, they usually are not doing so in efficient ways," said Jackie Scherer, lead interaction designer for Sight Machine, in an interview at O'Reilly Media's 2014 Solid Conference in San Francisco earlier this year. "The next step is really making that data very usable on all levels of manufacturing operations—from operators on the plant floor who need that data in real time, to executives who need that data, and everyone in between: quality managers and technicians, et cetera. So we're not only working to unearth that data; we're also working to make it very useful and usable. You can't just get the data; you have to actually figure out what the needs are for people in manufacturing, and reach them. So we're designing a very usable system for them."
Scherer, who also gave a presentation entitled "The Industrial Internet and Manufacturing's Digital Revolution" at the conference, said that Sight Machine's service is well-aligned with a number of trends that she sees emerging in the manufacturing industry.
"Changeability of process will become a norm," she said. "The imperative has been stability thus far, and the amount of variability will just continue to increase. So this idea of being agile and nimble is going to be equally as important as the stability and security that manufacturers are accustomed to in repeatability of process. That's a trend that will be important.
"Another trend that's going to be in our favor is the control of machines moving away from the plant floor. Mobility is going to be penetrating manufacturing moreso, which helps us as well because one of our big benefits is 'data accessible anywhere.' So an executive could be on an airplane and be able to see how his operations are doing. It's definitely a huge benefit of what we offer. Just the fact that that control of operations is going to be more mobile is definitely a trend that's helping us."
One of the company's customers is Shannon Precision Fastener, a Madison Heights, Michigan-based manufacturer of cold headed fasteners for Ford, General Motors, and Chrysler, as well as some of the larger tier automotive suppliers. Since its beginning in 2004, Shannon Precision Fastener's aim has been to "fundamentally reduce variation in the fastener production process."
Shannon's quality manager, Bob Allison, says in a video on YouTube that Sight Machine has helped Shannon remove variation from its process—specifically, measurement variation. "The key is giving the people that run our equipment and manage our process the best possible data to make a decision," Allison said. "So the ability to assign a measurement, assign a value to something that previously was just a judgment call, is huge."
Ed Lum, president and CEO of Shannon Precision Fastener, also says in the video that quality is one of the differentiators that Shannon has chosen to pursue, and has been the key to the company's success. "We have a trusting relationship with our stakeholders," he said, adding that they don't do any receiving inspection. "They trust that our parts meet their requirements that they've given us in the print."
Referring to quality as Shannon's "trust capital," Allison added that when the company entertains new customers, they want the customers to know beyond a shadow of a doubt that a Shannon Precision fastener is the highest quality product that they could place in their vehicle.
"When we realized that Sight Machine and the technology that they brought could potentially improve our process dramatically, we'd ignore the opportunity at our own peril," said Allison. "The folks from Sight Machine have been great to work with. They've done a great job of taking our requirements and our desires and turning that into a functional product. If you have any opportunity to harness technology to improve your process, to make you more efficient, to improve the measurement processes you have in your plant, take advantage of the opportunity."
Sight Machine's investors include IA Ventures, an early-stage venture capital firm that invests in companies seeking to create "competitive advantage through data," and O'Reilly AlphaTech Ventures, a seed-stage investment firm.
Sight Machine President and CEO Jon Sobel took time out to talk with D2P recently about the company's data platform for manufacturers. Following is an edited transcript of our conversation.
D2P: Can you tell us a little bit about how Sight Machine got started, and what was the unmet market need that you and your co-founders identified?
Jon Sobel: Sight Machine was founded by several people who had both long experience with Internet technologies and what we now call "Big Data," and manufacturing. The original idea was to try to find good uses for modern data technologies in manufacturing, and the first founder of the company, Nate Oostendorp, spent months speaking with factories in Michigan about pain points. It was a very unusual thing for an engineer to do—to try to figure out the unmet need before building anything. And what he heard repeatedly was that there are a lot of cameras in factories that generate a lot of image data that is used to make a pass/fail call in inspection, and thrown away.
He set about recruiting several people and developing some modern software frameworks for processing massive amounts of image data. And several others of us joined the company with the thought that we would focus heavily on the collection and analysis of image data because that is data that is very rich, and the value of quality to manufacturers is well understood.
Within about six months of first commercializing our technology, we got pulled by several early clients into collecting a much larger variety of data and combining that data with image data. So, for example, we now concurrently analyze images, sensor data, and information from other factory software systems and are able to give much more comprehensive insight than we originally set out to do.
The main difference between what we're offering and what has been available in the past is a result of the underlying software architecture. For decades, factories have had very sophisticated systems for collecting information, but these are large, expensive, difficult-to-implement pieces of enterprise software. And the way we're approaching this is using what people in the Internet industry would refer to as SaaS—software as a service, or cloud technology, which is much more suitable for fast, quick, point installations in a factory, and then can be expanded as needed.
Another significant difference is because we work with unstructured data technologies, there is much more opportunity for working with large amounts of data from different sources and giving real time insight than has been possible before.
D2P: You mentioned that previously, machine vision systems didn't retain the data—it was thrown away. It sounds like your data platform, or your service, can archive a lot of this data and make it available at any time; is that correct?
JS: Correct. So, amazing but true, factories have invested significant sums in generating very useful data. Taking pictures of everything is great data to have; having sensors all up and down your line is great. The tragedy, from a computing point of view, has been that the data has been either erased—as in the case of cameras, it's often just erased—or put in places where it's very difficult to access and use in any practical way. We have come across countless situations where people are using thumb drives to shuttle data over to a server where they'll never be able to view it again, or are using FTP servers to store a limited amount of information in software that was never set up to do anything like that.
So, yes, a fundamental difference between this type of platform and what has come before is that this really is intended to collect and use data that's already being generated.
D2P: How would you say Sight Machine's data platform contributes to a smart manufacturing operation?
JS: We help companies gain real-time insight into their quality. We help them improve operations, and we allow for people to exchange and analyze data across operations, so that, for example, a business analyst in another part of the world can see onto your factory floor and know how well your process is working. That's all a function of modern data technologies and Internet architecture. I don't need to have your software installed in my operation to see what you're doing; I just need a Web browser.
D2P: So what would you say is the biggest benefit of connecting a factory's machines to the Internet?
JS: Real time insight, delivered in a very simple, straightforward way. There are a lot of data companies that say, "Here's all this data we've collected; what do you want to know?" The art in doing this well, and the benefit to the customer, is putting the one or two things they really want to know on a screen, so that an operator working on a line knows what to do. The big data part is cool, but it's just an enabling technology. The only thing that matters is "Can you improve your production?"
D2P: Your website mentions that Sight Machine uses "networked cameras, vision algorithms, and cloud computing to control quality and improve OEE." Can you take us through each of those and give us the nuts and bolts of what Sight Machine offers to manufacturers?
JS: Sure, and I should be very upfront. We're a young, venture-backed company. I'll give you a couple of use cases, but we probably look a little bigger than we are. We are working with companies like Chrysler, so we're doing this for some very significant companies, but they generally prefer that we not get too specific about their problems. So I'll give you some examples.
At one factory, we are combining insight from 30 or more sources of data in an automated process to identify bottlenecks in that automation equipment's functioning. Those data sources include cameras, temperature and pressure sensors, PLC, and other types of automation equipment. That hopefully addresses the networking aspect. We provide simple graphics that show where the bottleneck is and why.
In another factory, we are taking image data from different types of cameras and helping the manufacturer ensure that there are no defects in the produced part, and, if there are, understanding where they're coming from. That particular manufacturer thought they had a couple of defects a week. We've seen situations where they have as many as 20 or 25 a day that they didn't previously know they were having.
D2P: How are your customers so far finding the data platform, and what kind of feedback are they giving you about what they like or may not like about it?
JS: The feedback has been good. Our early customers are expanding their use of our technology, so, of course, that's the best kind of commentary—when they ask you to do more. So that's been very promising. What they tell us they like is that they can associate production information with each item. So now they have what some people would call a 'birth certificate' for what they make. They like being able to share production data with their customers. In complicated supply chains, the way you generate confidence with your customer is you prove you have control over quality, and they like that. And they like being able to see, through a combination of pictures and simple graphics, what's working and what isn't. They really like the feeling of control that this technology provides them.
D2P: Your website mentions that operational intelligence is called the foundation for modern manufacturing. Can you talk a little bit about why that's the case these days?
JS: Sure. Manufacturers make massive investments in their facilities. It is a highly capital-intensive industry, where the money gets spent upfront and the way companies succeed is to manage assets well. Good factory operators can tell you what a minute or hour of downtime costs their operation. So if you can move the needle even a point or two on metrics like OEE (overall equipment effectiveness), they know that that translates immediately into value.
We were in a plant last week, and we asked the operator, "How's your OEE?" They had moved their OEE up 10 percentage points in the last year, and he gave us a dollar figure for what exactly that was worth to him. Most people running factories do this math in their head before they even meet us.
D2P: So what do you typically tell them as far as why they should be using Sight Machine's data platform? Why do they need to monitor their operations in real time, for instance?
JS: What we typically say is "You've already got the information you need to answer some important questions, but nobody has given you a way to use this information the way you want. What if we could help you do that? Would it be worth working together?"
We typically start small. Manufacturers are practical, level-headed people who, thankfully, like to see it to believe it. So we generally work with companies by picking a discrete project, letting them see how this works, and then, if they like it, expanding the effort together. The idea is new, and talking about it in abstraction isn't the same as seeing it on a computer screen. Once they see what they can have, the idea generates momentum on its own.
D2P: What would you say is your biggest challenge in convincing manufacturers that they should adopt your platform?
JS: They don't believe that this can work. They've been sold a lot of software over the years, and a lot of promises have been made, and the better something is, the harder it is to believe that it actually works. So there's an appropriate, understandable amount of skepticism around high technology, which is why we say, "Let's try something small first, and if you like it, we can go from there."
D2P: Do you sell them a modular type of arrangement?
JS: Absolutely. The whole way we approach it is, instead of saying "Let's do a $5 million implementation that's going to take two years, and you're going to have to shut down your factory to do it," we say, "Let's try one thing in one part of your factory," and you can do that with this technology. It's cost effective for us; it's cost effective for them. If they have an area of their factory that has a couple of PLCs, or a couple of cameras, and we can get some data from them and give them an example of what we can do with that data for them, then we're no longer talking about hypotheticals—they can see it. They can test it, they can have operators play with the software, and once that's in, everybody starts to communicate very clearly about what they would like to use it for.
Our first customer, in Michigan, helped us develop an initial product. They immediately began speaking with us about putting it in a second facility as soon as they saw it. The same dynamic is happening at other companies, such as Chrysler. Once it's on the floor, people can see the benefits instead of it being just a conversation.
D2P: Are you at liberty to mention the name of that first customer?
JS: Sure. Shannon Precision Fastener. They were a great first client for us. Fasteners is a tough, precision engineering business. They're banging steel and making 2 million fasteners a day at that factory. You can't get any further from Silicon Valley than a fastener plant in Michigan. They were fantastic—they are rigorous statistical thinkers about quality, they tested our stuff hard, and they pushed us to give them something that they can rely upon for their quality. And as soon as it was up and running, we started talking about doing it a second time.
D2P: Is that second use of it already in operation?
JS: No, we're discussing implementing it right now.
D2P: A lot of people are talking about "the next Industrial Revolution." How would you describe this next Industrial Revolution, and what we can expect from it?
JS: That is a very big idea. I'm not sure I'm qualified to comment on it, but I'll share some personal beliefs with all the caveats.
We've come very far by optimizing materials, physical processes, and standardized ways of doing things. The next chapter in industrial innovation is about combining what we can do with information and data with the physical world. Organizations that are good at working with software and information have an opportunity to innovate traditional industrial processes, and those who can glean insight from massive operations that are generating large amounts of data will have an edge in the future.
"The next chapter in industrial innovation is about combining what we can do with information and data with the physical world. Organizations that are good at working with software and information have an opportunity to innovate traditional industrial processes, and those who can glean insight from massive operations that are generating large amounts of data will have an edge in the future."
—Jon Sobel, president and CEO, Sight Machine
D2P: It appears that there's a lot of demand now, and will be going forward, for people who are able to work with data and provide advanced analytics. It sounds as though you have that type of need at your company.
JS: Yes. The DNA of our company is a great mix of people who understand and enjoy the disciplines required in traditional manufacturing, and cutting-edge data science types. We believe strongly that success in the Industrial Internet, or Smart Manufacturing, will come from blending the two worlds.
This is not a situation where the high tech community can walk up to factories and say "We know what you need; here's what you should do." And it's not a situation where factories, by themselves, necessarily have all the tools to do this. It's not an area that they've focused on, traditionally. The combination—the bridging of the two worlds—is where the opportunity is. We firmly believe that; it's why we have people in Michigan and California, and, soon, in Asia, and it's as much a cultural challenge and opportunity as it is an opportunity for technology. People in manufacturing think differently than people in high tech, but if you use data and logic and good clear thinking to solve problems together, there are great opportunities here.
D2P: Where do you see Sight Machine's data platform heading in the future in terms of its manufacturing applications?
JS: Of course, you can never predict the future. But let me put it this way: Our next year or two will involve extending our platform across a variety of manufacturing verticals. We began in automotive, we've moved into a couple of other verticals, we see lots of opportunity with electronics, medical devices, pharmaceutical, food and beverage. There are regulatory requirements in some of these industries; all share an imperative for quality and all are thinking quite aggressively about how to use technology to improve their prospects. We hope to identify solutions that have common benefits across all these industries.
That's a big goal, and we have respect for the challenge. But what we've been struck by is the commonality of needs across these industries, and in the past, there's been a high degree of specialization because the technology has been so specialized. We believe a company like ours could be quite successful as a horizontal supplier of technology to multiple industries. Data is data.
D2P: There's been a lot of discussion about the need to reboot our training and education of skilled professionals for manufacturing. So growing our workforce with more tech savvy people who are educated in the ways of so called 'New Manufacturing,' might be a positive thing for businesses such as yours, making it easier for manufacturing companies to see the benefit in your data platform. Is that something you would agree with?
JS: Absolutely. I might put it slightly differently. You know, there's been some fascinating press about the unexpected results of automation. You might have seen an article some time ago in the Wall Street Journal about Harley Davidson, and how the mix of skills changed. The difference is that in a world of Smart Manufacturing, companies are necessarily going to place a lot of decision making and judgment in the hands of their employees, who will need to make judgments and decisions about flexible operations. And we feel that those kinds of operations are going to benefit from having tools that enable people to have the information that they need to make the decisions.
D2P: And would you say that your platform would enable manufacturers to respond more quickly to changes in customer demand or market demand for their products?
JS: Absolutely. That's a deep theme in all of this, is the breadth and speed of customization. Mass customization is here. We work in a factory that has, literally, tens of thousands of permutations of what it makes. And companies who have to rapidly change or modify operations highly value the ability to do that well and then quickly optimize their operations to get them back up to scale. Having the ability to know if things are working right when you're making a lot of changes is valuable.
|Home | About Us | Back To Technical Library | Contact Us|
Copyright © 1996-2010 JobShop.com. All Rights Reserved.
General or Technical Questions? E-mail support@JobShop.com