Deeper Insights into Manufacturing Operations


Scott King

Managing Partner, Brilliant Data

Kevin Sturgeon

Solutions Engineer, MapR Technologies

Having operational metrics available at a glance improves decision-making. Join us as we welcome Brilliant Data’s Randal Scott King, who will provide a demo of their new Solution for Manufacturing Operations. SMO collects data from multiple systems within the organization and applies predictive analytics, giving both a breadth and depth of information missing from other solutions. Get faster, more accurate information about:

  • Inventory (including Days on Hand)
  • Production and Scrap
  • Complaints and Claims
  • Safety
  • Predictive KPIs like expected production levels


David: Hello, and thank you for joining us today for our webinar 'Deeper Insights Into Manufacturing Operations' featuring Brilliant Data. Our speakers today are Scott King, Managing Partner at Brilliant Data, and Kevin Sturgeon, Solutions Engineer at MapR Technologies. Our event today will run, in total, approximately one hour, with the last 10 to 15 minutes dedicated to addressing any questions you may have. You can submit a question at any time throughout the presentation via the chat box in the lower left-hand corner of your browser. With that, I'd like to pass the ball over to Scott to get us started. Scott, it's all yours.

Scott King: Awesome, thanks. Hi, everybody, and thanks for coming. We got a lot to show you today; I don't like to bore people with a lot of "fluff", so this webinar is probably going to be shorter than others you've been to as a result. We're going to talk briefly about why you would want an analytic solution like this; what kinds of problems it can help you solve. I'm going to spend just a few minutes showing you how this is set up and what the parts are in it. And, frankly, some people told me I'm crazy for doing that, because I'm literally showing you how you could build this yourself, if you were so inclined.

Scott King: But, I figured, if you got the time and the resources and the talent at your disposal to do that, more power to you. After that, we're going to do a live demo, so you can see how it works, and what you can do with it. I'm sure that's the part that everybody's most interested in. Again, we're not going to waste a whole lot of time getting there.

Scott King: So, let's just jump right in. I'm Randall Scott King, I'm the Managing Partner of Brilliant Data, the company that created this solution you're about to see. We're joined today, of course, by our partners at MapR who make the data platform that serves as the foundation for the solution. I think they've just got an incredible product. It makes a lot of things easier; like manufacturing analytics, sales analytics, really, any kind of data analysis is easier with the data on MapR.

Scott King: Let me break Webinar Rule #1 already and go over who this will help, and who it probably won't. If you've already got some mega-expensive analytics solution in place, this probably isn't for you. Those systems are expensive to implement. Doing a rip and replace is going to be painful unless you've got a major support or license fee that you'd like to get rid of with those, then this probably isn't for you. If you want to ... if you're in that situation and you want to drop off, feel free. Nobody's going to ... get offended by that.

Scott King: Also, if you don't really believe in using a lot of this data in your decision making, or if your company doesn't historically use a lot of data for decisions, this might work for you, but you're going to have to account for the cultural changes that are going to have to happen. That's a whole other kind of project, and while we can help you with that while we do that, this ... that's not really the focus of this webinar.

Scott King: Now that we're all on the same page, the growth of data over the past decade has, just been, astounding. According to people who measure this kind of thing, probably somebody in a university somewhere, we generated as much data in 2015 as we did in all of recorded history before that. When you think about that for a minute, that's what 10,000 years verses one ... that's how much data we're generating these days, and these days, it's not just us, anymore. We've got machines and devices creating data, too, and it's just become somewhat of a tsunami, really.

Scott King: Companies like yours are realizing this, and they're realizing how they can put that data to use to help them run their businesses. It used to be ... if you needed to make some decisions and you wanted data that would help clarify what was going on and help you make that decision, you were kind of limited by the data available. There wasn't that much of it, and, frankly, a lot of it just wasn't really reliable. There was a lot of made up data, there was data that had no government process, so you weren't sure about the quality of it. Of course, there was just anecdotal evidence, which was hardly ever reliable.

Scott King: The environment today is a lot different. It's exactly because we have so much data available. Business leaders have come to expect that there's an answer to almost any question, and it can be gleaned from some kind of data available somewhere in the company. They may not know what the data is, or where it's located, but there's this notion in the back of their head that it's available somewhere. When they're told that it isn't actually available, a lot of times they don't understand that and they get frustrated. I'm sure that some of the people on this call have experienced that.

Scott King: On a side note for a minute ... A few years ago, I was the director of business development for an IT teller and my organization had a hundred million quotas to meet that first year that I was in the position. We were nowhere near it. We were just getting started, and that hundred million was going to be a big jump from the previous year. To get a better idea of where we were as an organization and try to get a handle on the size of the problem, I started asking a lot of questions of different people, mostly in IT. I really wasn't getting very many answers. It was frustrating for everybody. They're getting tired of not knowing what I was asking, too, because the guys in IT especially felt like they should know; they should have that data.

Scott King: It turns out they did, and they didn't know it. They have the data because we had one of the better ERP systems out there, but they didn't have the tools or the process to turn that data into answers. It was just kind of sitting there for the financial guys to use at the end of the month or the end of the fiscal for their reports.

Scott King: We started to change that, and it wasn't very long ... it was a pretty short period of time before we started to see success. We wound up hitting that hundred million mark that year, which was ... that was a big deal, because that was double digit growth over the previous year. We couldn't have done that if we didn't know how to turn the data that we had into answers. On a personal note, what was even better, really ... what was the best part of it, was the frustration everybody was feeling with it, with each other, just kind of went away. It was great. The guys in IT didn't dive under the desks when they saw me coming anymore.

Scott King: This is what's been happening in business the past decade or so. Businesses are relying on data more and they're getting the benefits of that. When you hear that your competitor just gained some percentage of market share, and then you see their name in some use case, an analytics company somewhere, it doesn't take much to put two and two together.

Scott King: So that's data growth and data use, generally. Let's talk about manufacturing, specifically, because obviously, that's what you care about and that's why you're here.

Scott King: In some areas, like sales and marketing, it's the same; almost everybody has those functions. Specific to manufacturing, you got things like inventory levels, and those typically divide out into raw material, work-in-progress, and finished goods. You want to know the status of all those because it helps you in several areas. You want to know how much raw material you need to bring in. If you can predict that out several months in the future, that's even better.

Scott King: It's the same with production. If you can know how much you're producing and see variations in that day-to-day or week-to-week, or whatever, you can plan better. Again, if you can predict that, that takes it up a notch, doesn't it? And then there's scrap. There's always going to be a certain amount of loss in scrap, due to the manufacturing process, itself. That's okay. We want to minimize that and that's tough to do if you're not measuring it and you're not sure that the data you've got is good, right? Even more importantly, you want to know if there's a change in scrap. If a certain machine or work area in a plant suddenly goes from a typical two percent scrap up to five percent or ten percent, or whatever, that needs to be checked out.

Scott King: As an example, if you've got a press, and that press is going along at that two percent per month, and then, one day, it just jumps up. Somebody sees that on the dashboard, so they go check it out. It turns out, there was a new guy that had set up a fan near the press and debris got in it and nobody noticed. Nobody noticed on the floor, but it showed up on the dashboard. You get it fixed that day rather than getting it fixed three days later and there's a cost savings associated.

Scott King: Now, that example may or may not be based on a true story; I can neither confirm nor deny that. Of course, you've got manufacturing-specific metrics, like OTIFs, or some people call it 'on-time delivery'. Over in safety, they've got OSHA stats they have to keep track of. Not only is data growing, but recording requirements are growing, too. Sometimes those are from upper management, but sometimes it's from, what I call, upper-upper management ... the government. That's especially true if you're republican.

Scott King: Brilliant Data's been watching how firms use data for a few years now. We've helped some of them the systems in the processes in place, too. What we've noticed over that time is, they tend to go through the same stages. We can spot, almost immediately, what stage they're in. We call that task 'Data Maturity'. All it really is, is just ... where they are in collecting and using data. It doesn't even ... it almost doesn't even matter what kind of data they have. You can buy data; that's pretty useful when you're further along in the process, because you can start to incorporate external data to enhance your own.

Scott King: If we look at the 'Data Maturity Index' here, if you look over here on the left, there's not a lot of people in the 'Data Dark Ages' these days. They tend to go out of business or they're a small, family-owned plumbing shop, or whatever. Most business are collecting, at least, some kind of data, even if it's just sales. They're sharing that data around the company with spreadsheets. At least, they're discussing the data and what it means at that next stage in 'Analytic Infancy'.

Scott King: The problem is, at this stage, nobody's sure where the other guy's data came from, but they're sure theirs is correct. So, you get disagreements about whose data is better, or you have departments that have data that's critical to the business and they don't let that out ... they don't let other people use it.

Scott King: The next step up from that is when you become aware of business intelligence and how that can help you. Usually, the company gathers data from several different systems, they organize it, and put it into a data warehouse. Upper management says, "Okay, if you're going to use data, it needs to come from here, because we're sure about this data". Usually somebody, usually a BI consultant, gathers requirements and creates these various dashboards, which shows different people in different jobs, the data that they need that's relevant to them in their job.

Scott King: Each one of these stages on this slide is, just, a huge difference from the one before. When you go from 'Analytic Infancy' ... let's just call a spade a spade ... I call this phase 'Excel Hell', because spreadsheets just ... they multiply like rabbits, and they're a pain the neck to try to maintain. Some of you know what I'm talking about, I'm sure.

Scott King: Marketing people that put together this said I shouldn't call it that, and they came up with 'Analytic Infancy', and I guess that does sound nicer, but really, you don't want to stay in this stage, because it's frustrating. Moving from that into 'Business Intelligence', it's like night and day.

Scott King: Is it all sunshine and roses in the 'Business Intelligence' phase? No. You're going to have some things come up or you realize that a dashboard doesn't really show what you need, and you have to redo it. The issues here are nothing like 'Excel Hell' ... I mean ... 'Analytic Infancy'.

Scott King: When you get all the way to the right, you've gone through 'Analytic Infancy', you've gone through 'Business Intelligence', you get to 'Advanced Analytics'. I wanted to say using data it's kind of like ... using drugs. I'm sure I'm just killing the guys at MapR right now by saying that, but they're on the other side of the phone, they're just dying that I said that. But, it's true. Using data is addictive. You start solving problems with analytics and then you start looking for other problems to solve. Then, that new one requires some kind of different data, so you got to go out and find that and bring it in.

Scott King: When you're in the 'Advanced Analytics' stage, knowing where you stand today or knowing where you've been and what has gone on in the company in the past, that's no longer on us. When you start with predictive analytics and you start being able to predict what a customer is going to buy, or what your production level is probably going to be like in five months from now, that's addictive. When you get to the 'Business Intelligence' phase, you usually don't stay there, because you realize how much that helped you and you want to do more.

Scott King: These are the stages we see firms going through and not just our clients, but the whole industry. Just as a quick exercise ... just do it in your head ... What category, do you think, your company falls into right now? I mean, you don't have to answer that here, but it's useful to look at these criteria and kind of place where you are.

Scott King: When it comes to manufacturing, we went out and we talked with clients and others in the industry to see what kind of information and what kind of insights they were looking for. I'll be honest, I didn't know much about manufacturing a few years ago. But a senior director at a client of ours kind of took me under his wing and explained a lot of things about ... I didn't know what OEE was, I didn't know what OTIF was. When he showed me how they wanted OTIFs calculated and why that was important, it just made a lot of sense. Of course, you want to know that.

Scott King: As we're building this system for them, we're getting a real education in manufacturing processes and how the business runs. These are, what I think, the minimum, the absolute minimum question that an analytic system needs to answer, and even this isn't all. You got to measure scrap for reasons we already talked about. You got to have a handle on what products your customers are complaining about or returning. You need to know what your day's on-hand for inventory are. It's got to be accurate, all of it. These are the kinds of questions that we built this solution to answer for you.

Scott King: Let's go ahead and talk about the solutions for just a minute. This is what I was talking about before when I said a few people said I was nuts, because I'm literally showing you how to build this yourself if you wanted to. But, you know ... that's okay.

Scott King: This is an example of what the dashboard resolved looked like. These are the areas that we covered right now, and we're actively working on more. If your curious to know about what comes down the price, just email me and ask, I'll be happy to share that with you. As we stand right now, we've got complaints, claims, safety, production and scrap, and inventory. Already, it covers a wide range of functions in the business and keeps track of them. You can use that information you need. You got systems to make ... the information in the trends are right there in front of you and it's easy to read in a graph-able format. You know exactly what's going on in those areas. Besides having up-to-date information, you can look into the future of it with inventory forecasting and production forecasting. Those are based on a number of things, including historical data, which I'll show you during the demo. You also ... they adjust for seasonality. It's not just a straight line, which isn't very useful. If you typically have a dip in March, or summer, or whatever, that's going to show up.

Scott King: How do we get this data? Here's the blueprint. All the way over to the left here are the typical systems in a manufacturing operation, and it should all sound familiar. We get a lot of data from the ERP, whether you've got SAT or Oracle, or something else. When we first created the solution, we did it with an SAT system, but we've expanded that out now. There's also the Environmental Health and Safety System, we pull data from that. We can pull data from your HR system to track staffing, so if you want to keep track of full-time hours versus part-time versus seasonal, we can do that. And I forgot to mention that earlier but, yeah, we can do that. That's something plant managers told us they needed, was to be able to track staffing hours.

Scott King: So, all our data is sitting out there, and we call these 'Source Systems'. We get that data out of them using an ETL tool like Talend. ETL stands for: Extract, Transform, and Load. And that's exactly what it does. It extracts the data from these source systems, it changes it into the format that you specify that you need--that's the transform part--and then it loads that data into MapR. And now I'm really telling you something I shouldn't cause you know what? There's a free version of Talend. I mean, you don't have to pay a dime for it. You can just download it. And it does a lot. I mean, the commercial version obviously does more. But still, the free version is pretty powerful; I use it all the time. So there, you just got some free IT consulting.

Scott King: So, we put the data into MapR with Talend, and that's where our product, SMO, goes to get that data. SMO never touches the production systems. They're just chugging along, doing what they do, and there's no chance the data's going to get lost or damaged because Talend works with a read-only account that you create in those systems. It doesn't change a thing in them, it doesn't bog them down asking them to calculate things for the dashboard. And that's important; it's a bigger deal than most people realize, I think. And a lot of BI vendors gloss over to you that, and they'll tell you how you can connect their product straight to those source systems. But, that's a really bad idea. You don't want to do that. You don't want your BI tool bogging down a critical production system, you know? Just let that system do what it does, ask it for the data, and then leave it alone.

Scott King: Our dashboards run on an instance of Tableau that talks to MapR. You can set it up so you have different instances. One for development, another for production. Or, if you want to control what people can see and what data they can see, which you really should care about that, we can set it up so that MapR and Tableau work with your existing active directory system to control that. To control who sees what. We think Tableau is the most flexible product for making dashboards so that's what we went with. And none of us here on the call, not me, not any of the guys from MapR, none of us are getting a dime for telling you this. Tableau's not even here. I don't think they even know we're talking about them. But, yeah, it's a great tool. That's why we went with it.

Scott King: So there you go. If you want to build a solution like this yourself, that's how you do it. And you cover all the bases if you follow that blueprint: (a) You're not interfering with your production databases, (b) you're building in security and you're bringing the critical data into a single place, and you've got the best tools to build your dashboards to do it.

Scott King: All right, now, my ex-wife would say that nobody wants to listen to me for this long and she's probably right about that. So I'm going to stop for a few and I'm going to let MapR take over. They're better at this anyway. They don't say things like "Excel hell". And, in just a few minutes, we're going to switch gears and show you the solution in action, which is the main point. So, MapR? Kevin?

Kevin Sturgeon: [crosstalk 00:21:48] Thanks Scott. And I may steal "Excel hell", but I'm certainly going to steal "Data addiction". So thank you. And folks, I know that I stand between you and the actual demonstration, which is what you're here for. This is something that is extraordinarily exciting for me as well. We work quite a bit with Brilliant Data in a lot of solutions, and this is a fantastic one. And it really illustrates the power of what you can do with data in this new world of advanced analytics and predictive analytics and moving beyond that maturity slope of standard BI.

Kevin Sturgeon: And I wanted to take a couple of minutes to talk about this, what we call the MapR Converge to Data platform which is where the data is stored and analyzed in place. And, real quick here without going too far into this, I know because I stand between you and the demo, this is all stuff that Scott had covered previously. Value and business value really comes from combining enriching data sets across departments' source systems, line of business, human and machine generated sensors, all sorts of different things, as well as first- and third-party data sources. So there's this explosion of data generation out in space that needs to be combined.

Kevin Sturgeon: This data also exists in a myriad of formats. We talk a lot in this world about data as structured and unstructured is sort of a common word nowadays, but source data exists as event streams, it exists as relational database tables, it exists as log files, or traditional files, or legacy systems. All sorts of different things.

Kevin Sturgeon: What MapR provides, again, what we call a 'converged data platform' which consolidates and supports storage for not just single-purpose file types, but relational data and non-relational data, event streaming data, structured and unstructured into a single place. And what's different about this is that most people will start to construct repositories based upon the specific data type that they're looking at. Sort of a single purpose, whether that's a no sequel database or an HDFS pad only system or something like that. And that part can converge all of those different types of file systems together into a single cluster.

Kevin Sturgeon: We also support, as you would expect, and as the diagram showed, different analytical engines. Sequel processing, advanced machine learning and AI through things like Spark, but also, we're seeing more and more of general purpose workloads through container orchestration like Kubernetes. And now Spark is supported on some of those more modern workload processing engines.

Kevin Sturgeon: And what's important about what we provide here at MapR is a unified framework. It is our software that can run and provide interfaces for all these different types of data, all these different formats, with a single management interface, single governance scheme, single securities scheme, versus having to stitch all of this together on your own.

Kevin Sturgeon: The second point I'll make before we get into the demo here is that this data, this data engine, this data fabric, can be distributed globally. Especially in manufacturing where you have a myriad of locations internal and external. Shop floor, data center, cloud provider. Our software and our clusters can be installed in any and all of those locations all the way out to the edge, and can distribute data and processing consistently across those locations in a single namespace. So you truly have the ability to consolidate and store and act upon all of the different data sources in your environment.

Kevin Sturgeon: So that's just the quick plug for the MapR Converge Data platform; building this notion of a global data fabric for storage and processing across edged cloud data center and what have you. So I'll turn it back over to Scott to go into the demonstration.

Scott King: Let me just share my screen real quick here. We're going to start with Production/Scrap because that's what people always usually ask about first. So, just real quick David, can you see that?

David: Yes.

Scott King: Okay, awesome. So, there's a lot going on with this page, but it's all centered around answering just a few basic questions. And those are: How much did we produce? During what time period? And how much scrap was there?

Scott King: So everything you see on this page right now is a variation on those questions. So, for example, here in the middle: here in the middle of the graph that shows Production versus Scrap on a daily basis. The solid bars and the numbers above them, they represent production, while the orange line is scrap. So if I hover over this bar like I am, we can see that the production that day was 125,508 in terms of units each, and we see that 1.1% of that was scrap. That's measured in base unit of measure, so whether that's units each or pounds or whatever the material's unit of measure is, we add all that up to get this number.

Scott King: Now, that's pretty useful, especially when you can look at it on a daily granularity instead of weekly or, even worse, monthly. The first client we worked with on this, they were getting it weekly, and sometimes they didn't even get that. I mean, the data was there, they just didn't know how to get it into the hands of the people that needed it.

Scott King: We could also look at a different way of measuring production on that same graph. Like dollar values, for example. So, if we come up to here to the top of the screen, you'll see we've got a selector named 'Measure' and if you click that, you see that we can change our criteria from 'Units of Measure' to 'Weight' or to 'U.S. Dollars'. So, let me click 'U.S. Dollars' and you can see that all the values on the graphs just changed to reflect that we're looking at dollar values now instead of quantities.

Scott King: So now if I hover over the bar, we can see that it shows what we were looking at before, but this time, now, we're looking at dollar value. So $124,161 in production that day, versus $2,712 in scrap. So we can look at quantities, we can look at base units of measure, we can look at dollars, and all of that is shown on a daily basis here on the graph.

Scott King: Now that's great. What if you want to control the number of days you're looking at on the screen? Well, we can do that. Off to the right here is a selector that lets you specify the timeframe that you want to look at. So, if we want to look at just the past several days, we can do that. Look at the last nine days, for example, that's what I've got here. Or we can look at the past few weeks, again, if I want to look at just the past two weeks, for example, I can come over here and set that and the graph will change to show me just that time range.

Scott King: Now, what we're looking at so far is from one individual plant. It's Charlotte. And I know that by looking here under 'Facility Name'. So, like we mentioned earlier in the demo, this particular fictitious company has a little over a dozen plants scattered across the U.S., but, if I want to look at all of them, I just come up here, select 'All' rather than just 'Charlotte', and I hit 'Apply', and the data updates to show us what's going on across the company.

Scott King: And then, again, I can come back and just show 'Base Units of Measure'. Anyway, right next to 'Facility Name' is 'Division', and this company has two different divisions. They've got 'Commercial' and 'Consumer'. And, by the way, if, as we go throughout this demo, sometimes some of the data is not going to make a whole lot of sense, and that's just because it's completely made up.

Scott King: So, just between these two connectors, between 'Division' and 'Facility Name', we're covering what almost everybody wants to look at, from the CEO down to the Division VP, down to the Plant Manager. But actually, we can get even deeper than that. So, let me come here and change this real quick so that it shows, in order, this is the data for individual work centers in a plant. So, for example, this punch machine. We can see that it has produced a little over seven million in production, 1.6% of that was scrap. That's an individual work center within a plant. So we're getting down below the granularity level of just the plant. We're seeing what's going on actually in an individual work center.

Scott King: Let's go back to Charlotte, real quick. And everything changes to show just what's relevant to Charlotte. So I can see over here in the 'Work Center' graph, I can look over here and see the numbers for this FUTOBA machine. I told you, the data may not make sense sometimes. Okay, so we can look at all different levels of granularity from the whole company down to an individual work center and we can see what's going on in the past up till now.

Scott King: But you know what's really cool is when you can make educated predictions. And that's what this graph down here at the bottom is all about. The graph in the middle that we've been looking at is days of Production and Scrap. But down here, everything is measured in months. And what we've done is looked at historical production levels and we used an algorithm to say okay. That's what the lighter blue bars are to the right. It's a prediction of what production levels are going to be for the next six months.

Scott King: If you look at this data, you can see there's a pretty clear seasonality pattern to it. It's always down somewhat in the winter, and then it peaks in the summer. And the algorithm, if you look at the results, the algorithm takes that into account and knows how to adjust for it. This isn't just like somebody saying "Okay, during the summer it's always down so let's make the estimate a little lower", the way that a human being would do that. This knows how much lower each month ought to be. So it's a lot more accurate.

Scott King: Speaking of being more accurate, if you wanted to take the accuracy of this up even further, several more notches, you can use more data than just historical production. So, maybe your business is one of the ones that's affected by weather, you know, some are. So, we can look for that correlation between your data and weather data, and, if the correlation is there, we can account for it and use it. Or, if your sales people are real good about keeping up the company's CRM, and, let's face it, some are better about that than others, then you could use that data as well.

Scott King: So there's a few different things you can do to get just spooky accurate about this production forecast. Even though the simple model we're using here works pretty well, better than most humans. And, remember, what we're looking at right now is for an individual plant. So, if a plant manager has an accurate forecast to work with, he can use that to tell them what he's going to need over the next few months, and he can prepare for it. Or, procurement can use that to negotiate better deals with suppliers on an individual location basis.

Scott King: So having accurate forecasts opens up a lot of possibilities when you think about it. I mean, the sales team, it benefits them. Even marketing can benefit from it. All right. So that's Production/Scrap. Let's look at Claims and Complaints.

Scott King: One of the things that will jump out at you about this page, besides the fact that this fictitious company somehow has both Armani and Boeing as customers, is that (a) its year-over-year, and (b) it's all ranked. So if a customer submits a claim and sends something back to you, this page is going to show you: how many times that happened for each customer, how many claims were filed against each facility, how many claims for each kind of material or product, and for what reasons.

Scott King: So if we look at 'Material', and we look at what's getting returned and getting claims filed the most, well, it's pallets. We're sending orders out on returnable pallets. So the customers call up and they return those to us, no big deal. And if you look over here at the 'Reason' codes, we see the number one reason for claims is 'Returnable Packaging'. So, that makes sense. Now, third highest on the list is "incorrect price billed". So that tells us we should be looking at our processes around that. We should maybe be going and asking questions of the sales team, and maybe the shipping team, about how we can fix that and be more accurate with pricing on orders and get that number down.

Scott King: This screen doesn't just show claims. So if you track complaints, and gather that data from customers, then you can represent that here as well, like we did on this graph up here to the right. Complaints may or may not have turned into claims. It's good to track both of them so that you know.

Scott King: Now, if we look up here at the facility data, we can see that Tampa and Cleveland have more claims than every other facility by wide margin. But if you looked at the production data under those two facilities, you'd notice that they produce more than everyone else, too. So that doesn't necessarily indicate a problem. That may be okay.

Scott King: If we look at Tampa, you see the only thing showing in "facility"--let me come up here to "facility" and go to "Tampa". Get rid of "all". Select just "Tampa" and apply that.

Scott King: So now we're looking at just the data that's applicable to Tampa. Over here on the left, these are the customers that buy products that are actually made in the Tampa facility. And our reason codes over here change to just what's applicable to Tampa. Again, because of the measure, we can look at those in either in terms of "Numbers of Claims" or "Dollar Value of Claims".

Scott King: That's pretty much it for this screen. Claims and complaints are pretty straight forward; they don't get too complicated. You can pretty much show everything you want to know about them on a single screen.

Scott King: Let's move on to inventory then. The inventory page is great because in addition to showing your inventory quantity and the value of that inventory month-to-month, we also show you you're current days on hand for inventory.

Scott King: Down here on the left, we've got your historical inventory level values measured in currency, up to and including this month. When I hover over one of the bars like this here, we can see how much inventory we held on the first day of each month for comparison purposes. Then that's measured in dollars. Right next to that, we see the same thing, but it's in "base unit of measure" instead. Whereas the left is about the monetary value of the inventory, the right shows quantity. It goes a step further by also showing your days on hand with this orange line.

Scott King: We can hover over either one of these and see our average days of inventory for each month. We can see if we're getting better or worse month-to-month when it comes to that. You can see during the summer we got a little out of hand with it, and then got it back under control towards the end of the year.

Scott King: Now, just like we saw with production, we can take either one of these graphs and we can drop it below the other and create a forecast that'll account for seasonality. In addition to being able to predict what we'll be producing, like we saw in the other screen, we can see what our inventory levels will be if we continue the way we're going and we don't change anything.

Scott King: We can do that with these by dropping one under the other. But what we put together for this demo--what I'm showing you here--is that we're doing that on year-over-year graph on the top right here. The blue line shows that's all of last year, and the orange line shows current. Now, you see the orange line turns green after the current month. That's to indicate that everything after that is predicted. And what it's showing us is that last year our inventory levels overall just kind of swung wildly. There was this big dip in the spring of last year, and we overcompensated for that during the summer. Then we ended up with an excess of inventory at the end of the year. So we didn't do such a good job of managing inventory last year, obviously.

Scott King: This year was better. The dip in spring wasn't nearly as pronounced, and there wasn't this huge jump from spring to summer. So, the prediction algorithm sees that, and its prediction for the rest of the year doesn't show the same increase in the last quarter as last year as a result. It knows this year is going to look different that last year. The green line here is what the prediction is, but the green shaded area is what could be based on probability. What I mean by that is it could go as low as 112 million at the end of the year. Could go as high as 124. And that's all based on probability. So it gives you a range in the prediction on a graph like this. Which is nice. It's all based on actions we may or may not take from now until the end of the year.

Scott King: Now, at the top of the page, we still have the division and facility selectors like we do on the other pages, but now we've got a new one to go along with those. For inventory, we added "material group", because that's especially relevant here in regards to inventory.

Scott King: You can go there, you can select a particular material group, and all the stats on the page will change to reflect the data for just that material there. If you look right before that, you see we have current inventory levels for the various materials groups. If I scroll over one of them, we see how much of that we've got measured in base quantity. We can change that; we can make that display in currency if you'd rather have it that way.

Scott King: Finally, on the top left here we have a breakout by percent of the inventory showing whether it's raw material, work-in-progress, or finished goods. And that's shown by facility.

Scott King: So look at Birmingham, for example. They do a lot of raw material. 89 percent of their inventory is raw material. So that may or may not indicate a problem that needs to be investigated.

Scott King: Now, down here in Cleveland, they've got the opposite situation. They've got very little raw material, but they're holding a lot of finished goods. Again, that may or may not indicate a problem that needs to be looked into.

Scott King: That's pretty much it with inventory. So let's move on to safety.

Scott King: Now, this page ties in mostly to your environmental health and safety system. It shows data relevant to injuries, illness, lost time as a result of those, that kind of data. This data is required by OSHA for reporting purposes, so it's good to have an understanding of that. Not just company-wide, but at each plant. Of course, we have the division and facility selectors up at the top just like we do for all the other pages. So, you can get down to those levels of detail when you need to.

Scott King: Similar to the other sheets, we break things out into categories. So on the left here, we've got the numbers for each classification of injury. Whatever you keep track of in your health and safety database, and whatever categories you spread those out into. These are just what we came up with. You may have different categories all-together, but whatever is in your system is what will show here.

Scott King: Just like some of the other sheets, we give you the year-over-year analysis. We see at this time last year, we had 19 lacerations. But this year at the same time we've got one less--18. So that's good.

Scott King: To the right, we've got the cumulative totals of all the categories shown year-over-year. We've got 44 for this year versus 36 last year. Now that's not so good, because it definitely indicates we need to be paying more attention to safety education at the plants. But whether there's more incidents or fewer, you can see that here and know whether you're improving or whether you're not.

Scott King: Finally, down at the bottom, we have the raw numbers broken out into things like "total recordable incidents rate", or "TRIR", "lost time incident rate", and those are some of the regulatory stats that I was referring to earlier. Again, those are shown as current year versus previous year so you can compare the two.

Scott King: So that's it for the demo. I'm going to turn it back over to MapR, and we're going to maybe do some questions. Do we have any questions, David?

David: Yeah, thanks, Scott. We do have a couple. Just a reminder, if you have any questions, to submit them through the chat window in the bottom left corner of your browser.

David: To get started, we have a couple that came in.

David: "Are you able to personalize the dashboards?"

Scott King: Yeah, absolutely. Because it's built on to Tableau, we can go in and change just about anything. We can do something as simple as adding the customer's logo and changing the colors to fit their standards for advertising and marketing and all that. Or we can even go in and if they say "okay, yeah, this particular metric is great for automotive manufacturing. What we need to see is this other metric and this is how you would calculate it." We can go in and change that as well.

Scott King: There's not a whole lot that you can't change. We really look at the solution as being a good starting point. You come in, you put this in, you get the data in, and start putting it out in this format. Then you look at it and go "you know what else would be great? If we could stream in some IoT data and improve our OE."

David: Cool.

David: Hey, Kevin, I'm going to bring you back in here and ask you to add a little bit of color--I saw you were answering some of the questions in the chat window--and see if you could add a little bit of color to the ones that you were replying all.

Kevin Sturgeon: Sure. I appreciate the questions. Apologies if I was unclear during my small piece. MapR is software that can be installed in the cloud, On-premises, or in the Edge. It will utilize commodity hardware and software. Commodity Lynax blades running standard disks, or even object storage. You get cost efficiencies as your data grows. Since it's able to support all of these different myriad data types and formats, you get much more utility versus traditional data lakes that are using, say, to handle to these file systems. So I can yield, store, log data as it arrives. I can do traditional Hadoop API operations. I can take transactional data from databases. I can take change-data captured at event streams or event streaming data to have a very economic and very, very high-scale exabyte scale place to store and operate on data.

David: So Kevin, while I have you on, I'll follow a question to that that came in. "Which cloud vendors can I use with MapR?"

Scott King: That's a great question, and really the answer is "any of them". Since it will work on commodity environments, we can deploy against the course we have partnerships and marketplace crises at the major cloud vendors. All the way to folks like Oracle for their bare metal cloud, and other boutique providers that can spin up standard infrastructure. Give me a Linux machine and a block device to attach to and we can install this.

David: Great, thanks Kevin.

Scott King: I'd like to jump in on that one if I could, too. One of the biggest reasons we chose to partner with MapR to do a solution like this is--there's a few things, actually--one was scale. MapR in the file system that you guys have can scale just incredibly. You want to store more data, okay, well you just add more nodes to the cluster.

Scott King: The other thing, too, is this idea of it almost doesn't even matter where it is. You can store data on Premises in a cluster that's on premises. You can store it if you spin up the cluster in Adger or AWS or any other cloud, you can store it there. But what's really cool is if you have some combination of those. So if you've got an on-prem cluster and you've got another cluster out in Azura, the form of back up, the data replicates. So it's in both systems. It almost doesn't even matter where you put it, it's going to show up on all of them.

David: Cool. Thanks, Scott.

David: So another question for you, Scott, that came in. "How long does a project like this typically take?"

Scott King: The first time we did this--and I mentioned during the presentation--a few years ago when we first started doing this, I really didn't know that much about manufacturing. That first project took us probably six months. The good thing about that was we had quoted a fixed price for it. So, it ended up taking longer than I thought it would, but the customer still ended up paying me the same amount for it.

Scott King: Now that we've got a lot more experience with it, we know what we're doing, the answer is, like most questions in IT, "it depends." It depends what your systems are. It depends how familiar you are with those systems and knowing where the data is that we need to pull in. But you're not looking at six months anymore. We've had instances where it was six weeks. We've had instances where it was two and a half, three months, I think. That one was that long because the customer just really--the data was scattered everywhere and they didn't know where to get some of it.

David: Great. One last one for you, Scott. "Is the solution and production really this fast?"

Scott King: No. In order to keep things flowing smoothly and to avoid the chance of anything going wrong during the webinar, I'm using Tableau Desktop with an extracted data. So everything is all local. It's all on the laptop.

Scott King: In production, again, it kind of depends. Mostly, it depends on how fast your WIN is, really. What happens in production is you go from your browser, you log into Tableau server, and then Tableau server talks to MapR and gets its data from MapR. So, whenever you change something on one of the dashboards in a production system, it'll take a couple seconds. But, again, a lot of that is dependent upon how fast is the connection between Tableau server and MapR. And how fast is the connection between your browser on your machine and Tableau server.

David: Great. Okay. That's all the questions we have. Want to thank you, Scott and Kevin, for your time today and thank everybody on the call for joining us. If you have any questions or would like to set up a personal demo, contact us at 1-855-NOW-MAPR. For additional information, please visit or

David: Thank you again and have a great rest of your day.