Breaking Down Data Silos and Building Meaningful Connections Across Datasets
We kicked off Season 2 of our manufacturing podcast, Data In Depth, with Clark Richey from FactGem. Clark shared insights into how manufacturers can tear down data silos to gain a clearer picture of their business. He also discussed how to connect the dots to understand causes and effects, gain context, and predict likely outcomes. From managing product recalls to reducing waste in the supply chain, data integration and relationships are the keys to building a better business.
Check out this episode below and head over to Data in Depth to catch up on our first season! Don't miss an episode!
Follow Data In Depth on Facebook, Instagram & Twitter and listen and subscribe with any of the following:
Have a topic you'd love to hear more about? Interested in being our next guest? Let us know in the comments below!Transcript
Announcer: Hi, and welcome to Data in Depth, a podcast where we delve into advanced analytics, business intelligence, and machine learning, and how they're revolutionizing the manufacturing sector. Each episode we share new ideas and best practices to help you put your business data to work. From the shop floor to the back office, from optimizing supply chains to customer experience, the factory of the future runs on data.
Andrew Reiser: Welcome and thanks for joining us for season two of Data in Depth, the podcast exploring data and its role in the manufacturing industry. I'm your host Andrew Riser. Today, we are joined by Clark Richey, co-founder and CTO of FactGem. Welcome, Clark.
Clark Richey: Thanks, great to be on the show.
Andrew: So Clark, before we dive into things. Perhaps you can just share a little bit about your journey, and your passion for data, and what ultimately led you to co-found FactGem.
Clark: Absolutely. So, I co-founded FactGem about seven and a half years ago now with my co-founder and our CEO Megan Kwame. Prior to that, I've spent the majority of my career working in the United States Government defense and intelligence sectors actually as a contractor. So that really gave me an opportunity to work on some of the largest and hardest data problems that really exist today. As part of that, I became increasingly interested in two things. One, I'm interested in the connections between data points. So, and not just the information itself, but how is that information related to other relevant facts and data that forms a complete picture about what we're trying to understand. And two, I've always been interested in, how can we take these tools for data exploration that are being created on a regular basis, and maybe through a little more effort, make them available to non-engineers. Let the end user be a, whether that's a business analyst, a data scientist, or a VP of sales, let them have some agency to actually interact with, and manipulate, and analyze the data themselves. So, those two things kinda combine together and led us to where we are today at FactGem.
Andrew: Fantastic. I think those are two good pillars that we can form the basis of our dialog here today. So, with respect to the first item that you hit on, essentially making sense of data and deducting the information that comes out of these systems, oftentimes when we deal with companies or agencies, usually data is all over the place, right? So, it's in various systems, and what ultimately has been coined data silos. So maybe you can expand a little bit about that, your point of view on data silos, and how you start framing up those conversations with customers to educate them on this journey.
Clark: Absolutely. So, data silos is a fairly old term in the industry. Grew up many, many years ago. And really, it refers to the idea that traditionally, as we build systems, we build specialized systems. So we build them to answer a specific question, and then a we build of course, a data system to handle the data to answer that specific question for that specific business purpose. So this could be a system that manages your inventory. Right, it could be a marketing system. It could be your shipping system. But they tend to have just the information that they need to do the job. So if I'm the shipping system, I know probably a tiny little bit about inventory only in regards to like, what do I have to ship, maybe where is it going, how is it getting there, what is the status of it on route, and so forth? And that's fine because that's how businesses really used to operate, especially 15-20 years ago. And as consumers, that's kind of what we expected, right? But now the world's changed, and there's so much data and so many systems, things are highly connected. So no one is surprised. You may not like it. But no one is surprised, for example, when you do a search for something online, and then five minutes later, you go onto an app on your phone to play a little game, and you get an ad for something related to what you just searched for on your computer even though you're now playing a game on your iPhone.
Andrew: Yep.
Clark: It's not surprising anyone. Again, you might not like it, but it's not surprising. Right.
Clark: That information is now being connected, and that's what we've come to expect. And successful businesses are doing that really well. Whether it's in the manufacturing space, supply chain, retail, they're connecting all the dots. And that's where those silos lead to problems. We're separating those dots, and we have hard walls between them. So we need to figure a way to reconnect those so we can create that larger context to allow the business to make decisions between those silos. So to understand, for example, how does my inventory affect my ability to deliver and manage my supply chain and not to see them as separate pieces of the puzzle?
Andrew: Yep, yeah, that makes perfect sense. So when you're engaging with a perspective or a new customer, is there a framework or approach that you take them along maybe to do an assessment first, and then make recommendations? Can you maybe share a little bit about what that often looks like and how you kinda boil that down into a process?
Clark: Yeah, absolutely. I'm a big believer in doing things simply and in small steps to provide value. So, along that path, when you have a conversation with the business leader, whether that's a CEO, or someone running the line of a business, once you start talking to them and getting to understand a little bit more about how the day to day business works, very often you'll find they will express pain points in terms, not necessarily of, I can't do X. Because one, early on in the conversation, no senior business leader wants to tell you they can't do some critical business function, because that would be poor form at best. But if you listen carefully, you will hear that certain things are painful. They take a long time to do. So, for example, I worked with a very large healthcare organization a while back, and they'd brought in a brand new CFO. And this healthcare organization, as many have been, grew up through merging acquisitions. So they bought hospitals and other hospitals, and so forth. And this CFO was asked by the CEO, how much business do we do with this particular customer? And she thought that should be a pretty easy question to answer. You go to the system, and you pull up the customer, and you make a query, and you figure it out. Well, because this was built up through merging acquisitions it wasn't one system, it was closer to 50 systems. Because--
Andrew: Right.
Clark: When you acquire someone you acquire all their IT along with everything else. So it took almost nine weeks to get the answer. And of course this, especially as a brand new CFO, you wanna make a great impression. This was not the impression she wanted to make. So, you'll hear stories like that. I can do this. I can get there, but man, it takes me a long time, or I have a whole team of people that are manually querying one system, and then another system, and then another in order to manually put together that data. So you'll hear things like that, and you'll also hear that there's certain things they just don't ask, certain questions they know they can't ask because by the time you get the answer, the business opportunity's gone.
Andrew: Right, yep. It's a very common scenario and story that we hear as well. So similar to the scenario you gave in healthcare, I think manufacturing falls into that bucket with acquisitions over time, and in come cases, there's 30-40 different ERP systems. And so, a lot of these major initiatives that we start to see are around master data management, and a lot of that is ERP consolidation or just building these warehouses to help build the business rules and source all this data from these different silos to then make sense of 'em. So before we pivot and talk a little bit about empowering the analyst, can you maybe expand a little bit more on what you're seeing in these scenarios like you just described as it relates to these painful processes of manually querying each of these systems and aggregating that data? 'Cause there may be either lessons learned or some recommendations that you might have as it relates to how to approach a problem like that.
Clark: Yeah, I mean, there are definitely new approaches and techniques out there. As a cautionary point, first off, I would be extremely leery of any product company, or consultant, or vendor who came in and told you, one, we're gonna solve all the problems. It's basically magic. Don't worry about it. 'Cause that tends to just not be reality. And two, anyone who tells you, we're gonna remove your data silos and replace them with something else, what we call rip and replace. You might eventually want to do that, but typically, the data silos you have are there for a reason, and they're functioning very well within their context. So, I would caution against that approach that is sometimes put forward by various companies and organizations. And then again, the small step additive approach, right. Let's find that one pain point, see if we can add a capability on top of those capabilities that you have now in your siloed data sources, and see if we can address that problem. As I mentioned, as a number of new technologies and some companies out there that doing this type of work, personally I think one of the foundational technologies that is really helping here is in the graph database space, which is again really designed to connect these pieces of information together.
Andrew: Can you expand upon that a little bit more just to share more about what graph database means and how these pieces fit together?
Clark: Yeah, more than happy to. I mean, that's a foundational technology that we're built upon, of course. So if you think about databases, typically what we think about is we think about rows and columns, right. We think about tables. So, something that looks like an Excel spreadsheet is essentially the structure of a traditional relational database. And it's more or less been the industry standard for over 50 years. Now, relational databases are really good at answering known questions about known data within a fairly narrow viewpoint.
Andrew: Sure.
Clark: Graph database, on the other hand, look completely different. So if you would go to talk to any manufacturing company and have a conversation with them and try to understand their business process and their business model, at some point, probably pretty early on in the conversation, someone's gonna go to the white board and then start drawing some circles and some lines to explain, they have a product, and maybe they have some different manufacturing lines, and different components that go in, and all of the different important key business elements, and how they're related. And you're gonna have a bunch of circles and lines with probably some arrows and things on 'em that's gonna express that. And for us as people, that's very natural for us to think about information that way and become very clear for us to understand a picture that looks like that. That is essentially what a graph database structure looks like. It's a bunch of data points connected via relationships. And the really important thing there is that relationship aspect. It's really designed to let you understand how things are related. So again, if you think about manufacturing, typically, most manufactures don't make every single component that goes into the end product, right. We buy sub-components or outsource the manufacturing of sub-components that we then manufacture into larger components that of course we sell as a completed product.
Andrew: Sure.
Clark: And so, now, this challenge of understanding how everything is related can become very complicated and extremely critical. So we think about product recall, for example. If some component product, three degrees down the chain from you, so I get some widget from a manufacturer to help me produce my product, but really that's composed of 50 different things, and they're subcontracting that out to somebody else, and one of those pieces has a recall. Well how do I know which of my end products actually have that piece from that supplier in there? That's a really hard problem, and it's one that is really plaguing not just manufacturing to some degree, but really anyone who is shipping things. So you see this in the food industry all the time. You see these massive recalls of--
Andrew: Sure.
Clark: Food products. And again, the reason is, it's really hard in traditional systems to understand and query those relationships across those many degrees of separation, and that's where graph databases come in. In which you ask those complex questions about how things are related to one another even across many, many gaps in real time.
Andrew: So I think that's a good segue into to talk a little bit more about FactGem and the products that you all are creating, and maybe you can share a little bit more about that but then how that transitions and translates over into this new world of empowering end users and analyst to have this low code, no code approach of building these relationship and making sense of that data.
Clark: So what I'm really interested in here in this context is seeing how much power we can give to that end user to do as much of the analysis, as much of the data management as possible, again, without having to write any code, without having to engage with expensive software engineers, or consultancies, or any of that. How far can we push that? And so, we start with the concept that I just mentioned, which is that white boarding model. And, again this is all about giving agency to the business. One of the challenges now that's common across not just manufacturing but really all fields, is that right now if you want to, let's say, create a new application that helps monitor your supply chain so that you know you have enough pieces coming in to get into, to create your product and so forth. Well you have to describe that business process to your IT department who then models it, and they come back to you, and they show you some picture. And they say, we understand that this is kind of how the data works. And the unfortunate part there is most of the people in that room look at that picture and have no idea if that matches what the business actually said.
Andrew: Sure.
Clark: 'Cause it looks nothing like. Again, someone probably went to the board, drew some circles and lines, and that looks nothing like that. So that immediately leads to a disconnect and causes potential problems. Even if the model's right, it is maintained in a way that is different than the business thinks about it, and that leads to this cognitive disconnect that can cause friction when the business is trying to do analytics. And the IT department is saying, well, that doesn't really make sense. You can't ask the question that way, or you get a slightly different answer than you would expect. So we wanna remove that gap. The way you draw that picture on the white board, as someone who deeply understands the business, that's the way the data's stored. And you can draw that yourself. You don't need an engineer to come in and do that. Reduces time, reduces friction and room for error.
Andrew: We can come back to the pieces of empowering the user, but I'm curious, in your point of view, what's the hardest part about all this that we've been talking about and driving this change within these organizations?
Clark: I think you hit the nail in the head right there. Change is the hardest part. Again, if you think about manufacturing in particular, there's a whole bunch of challenges facing the manufacturing industry, and if you talk to a leader, very, very few of them will list software or IT in the top 10. They're gonna talk about the competitive landscape, write the cost of labor in the U.S. versus overseas, tariffs now these days, and so on. IT's probably not making the top 10. So that's challenge number one is helping them understand that there really is not just probably costs to be saved, but market opportunity to be gained by better connecting this data together. Then beyond that, you're presenting something that is a fairly new concept. Most of the software platforms and things that are being used in manufacturing are at least conceptually, fairly old. Again, it's not a high risk area, right? You have people saying, let's take a risk and do some really brand new cutting edge thing that I heard about in Silicon Valley. It's typically not happening in the manufacturing. That tends to be more of a low risk area. So you're saying, this is fairly new. It's about 10 years old, the underlying technology. So it's not like brand new, and we have good used cases and case studies, but you are asking to at least take a risk. And again, that goes back to the approach of saying, let's do something really small. We can immediately measure a return on investment so you can limit that risk.
Andrew: What you just shared there is exactly the right approach. I think that all these initiatives become too overwhelming, and the business is looking for the business value and the ease of doing business and getting the information that they're looking for, and IT is wanting to provide that, but it's kinda bridging that gap of expectations, and I think the best way to do that is like you just shared. You've gotta break it down into bite-sized chunks and create opportunity that has return on investment and can be demonstrated in a relatively short timeframe that doesn't become too overwhelming for either side of the coin there.
Clark: Yeah, absolutely. You have to recognize too I think that there are a number of stakeholders that are trying to get access to the same capital you are.
Andrew: Yep.
Clark: All right. You've probably got people say, I really want to get the newest version of this machine on the manufacturing floor. And they'll have good stats that show you it's 15% faster, it makes 2% fewer errors, etc., and they can very quickly draw a line from that to revenue. So, yeah, you've gotta start small and draw that line directly to how this impacts the bottom line for the business, or you'll never be able to capture their attention.
Andrew: Right. So one of the other questions we typically like to ask is, where do you see the future of this going? So how do you see tools like what you're creating getting introduced into the business and evolving? What's the outlook over the next couple years of where some of these priorities and initiatives where manufacturers might be looking to spend money on this stuff?
Clark: Well I think what's happening in the marketplace with some of these really large companies that are around now that do have the money and the expertise to take a very high tech approach, while that of course is creating a lot of market pressure and competitiveness, it is going to force other companies to look at, how can I become more efficient, and understand that efficiency, even though you're a manufacturer of a physical good, is often now about the data. It's not just the manufactured good. Everything is really data driven these days. And the truly successful companies are capitalizing on that. So, I think you're gonna see more and more innovators start to emerge. Again, I realize that I don't have to spend $20 million or whatever it might be to do a major overhaul of my physical plant in order to achieve some kind of a growth or a competitive edge. If I'm just smarter about how I manage my data, I can spend a lot less money and get ahead faster. And you're starting to see that to some degree I think, but that's really going to gain momentum pretty quickly here in the beginning of this decade.
Andrew: Yeah, that makes sense. Any additional closing thoughts or topics you'd like to dive a little bit deeper into before we wrap up the show?
Clark: Just as I think about manufacturing, so while it's not my area of expertise, I've always been fascinated by robotics, and I think there's been some fasting developments in the field of robotics, particularly manufacturing. And when you combine that with some of the artificial intelligence systems you get some really interesting opportunities to connect even more data to those systems for potentially some really interesting and unexpected developments in terms of productivity and just general innovation.
Andrew: Yeah, and I agree. We've had some previous guest share their perspective on introduction of smart devices on the shop floor and robotics. And to your point, I think that that just creates even more noise, right, as it relates to these streams of data. So not only do you have your systems of record and these data silos that already exist, but now you're introducing so many more devices, and widgets, and robots that are also internet connected and have capabilities of lots of data generated off of them. So finding that signal versus the noise of all this information, I think is gonna get more and more complex, but for those innovators that are on top of it and embracing that, then I absolutely agree. It's gonna be a differentiator for 'em.
Clark: Yeah, I mean, like who'd of thought a couple years a that someone would of come up with the idea to keep people at one spot, connect my order of systems to my fulfillment systems directly to my inventory systems, and move shelves--
Andrew: Right.
Clark: To people to pack boxes. Like, that sounded crazy--
Andrew: Yep.
Clark: If you say that five years ago. It turns out, whatever you may think of their business practices, it was a genus idea, and it's a really interesting way of combining some fairly elegant but sophisticated robotics with a lot of data to create a tremendous amount of efficiency. So yeah, it'll be interesting to see what happens.
Andrew: Very cool. Well, Clark, I appreciate the chance to chat with us today, and thank you for joining the show.
Clark: Thanks for having me. It's been great.
Andrew: So, for those listening, if you'd like to learn more about FactGem and their solutions, I'd encourage you to visit FactGem.com. That's F-A-C-T-G-E-M.com. And if you'd like to connect with Clark, we'll be sure to provide the relevant links to his online profiles in the show notes. If you enjoyed this episode, please take a moment to rate the episode and subscribe to Data in Depth, available on iTunes, Google, Spotify, Stitcher, and pretty much anywhere you might consume your podcasts. Thanks again for joining us today.
Announcer: Data in Depth is produced by Mountain Point, a digital transformation consulting firm focusing on the manufacturing sector. You can find show notes, additional episodes, and more by visiting dataindepth.com. Thanks for listening and be sure to subscribe wherever you get your podcasts.