Alex Ramirez: 0:03

Why not give those individuals the time and the liberty to think to create a plan a whiteboard versus literally playing the whack a mole game 24 seven managing operations and so the unleash unleashing ingenuity, I think plays a huge role in warehouses today and well into the future.

Tom Raftery: 0:24

Good morning, good afternoon or good evening wherever you are in the world. This is the digital supply chain podcast, the number one podcast focusing on the digitization of supply chain. And I'm your host, global vice president of SAP. Tom Raftery. Hi, everyone. Welcome to the digital supply chain podcast. My name is Tom Raftery with SAP and do with me on the show. Today I have my special guest, Alex. Alex, would you like to introduce yourself?

Alex Ramirez: 0:50

Yeah. Good afternoon, Tom. My name is Alex Ramirez, co founder and CEO of CognitOps.

Tom Raftery: 0:54

Okay, and what's CognitOps.

Alex Ramirez: 0:57

So CognitOps is a three year old startup founded i Austin, Texas, Tom, that a plies machine learning to help w rehouse operators manage their w rehouses better whether it's a f lly automated building, or a c eap and cheerful manual o eration. Too many of these o erators depend on tribal k owledge in Microsoft Excel. A d so we're bringing advanced t chniques to a bit of a d nosaur age of warehouse m nagement.

Tom Raftery: 1:24

Okay, and how can machine learning help?

Alex Ramirez: 1:26

So the warehouse is rich with structured and semi structured data, Tom, and what normally happens with warehouse operators is they will create Excel spreadsheets and tap into their existing applications through heck standard ODBC connections, to harmonize data into pivot tables and create a pretzel have a pivot table to just help them make better decisions, but never really mining previous versions of the Excel spreadsheet to find patterns, right. And so if ml is really about finding and being able to mined for patterns to either run a prediction, or to just give you a better insight into what happened in the past, then it's a perfect application in this instance, where this is what the operators do. They look for patterns, but they don't have the luxury of historical information. And if they do, it's, it's mostly in their brains, and it's tribal. And that creates lack of resiliency and a lot of these operations.

Tom Raftery: 2:27

How does one come up with an application like this?

Alex Ramirez: 2:30

So I'm a bit of an arsonist putting out my own fire as as my co founder, Reese Mack and we've been in the space for a very long time, I started my career at Andersen consulting now Accenture and my first project Tom was to build a warehouse management system for a little company at the time called walmart.com. Think all of your listeners have heard of them. And from there, I caught the bug of building these applications. And that was the vector of my career. And whether it was warehouse management or derivative forms, which are kind of ubiquitous now with warehouse control systems, warehouse execution, etc. What we found was that the, there was always a boundary to our optimization, it was that other application we were integrating into. And often I would find myself in warehouses many, many years of my lives in warehouses in the middle of nowhere, helping these individuals not just make sense of what I delivered with my application, but also the context that they were missing from other applications that we were trying to guess as to how they worked. And so feeling this pain being very empathetic to the whack a mole game led us to believe that there was a there had to be a better way. And so when recent I got together in 2018, and played a game of Pictionary on a whiteboard to explore the the expansive whitespace still left in the warehouse. We thought, first let's let's help this this persona, this constituency, but how do we help them? And we had dabbled in operations research in our previous life in our in our company. And so we thought, let's apply something a little different that isn't so static and backward looking. And that led us to data science and machine learning.

Tom Raftery: 4:18

Okay. And you mentioned that you can with machine learning go back through troves of data which haven't been tapped to date. What kind of things are you finding? What kind of problems are you helping to solve for warehouse operators?

Alex Ramirez: 4:34

What we're finding is a pattern for how certain order profiles flow through the building. And so Case in point, a building right a warehouse gets designed from, you know, either a system integrator or consultancy that comes in and performs an analysis and says, Hey, client, you're gonna need 10 miles of conveyor and this is the type of automation you need. And this is how It should flow left to right. But that that design is predicated on some assumptions. And yes, sensitivities are added to the assumptions. But what happens when the pandemic hits and all of a sudden your assumptions are in the rearview mirror? How do you adapt the the operation and so what we're doing is we're saying certain orders that used to go through these flow paths are now a lot less than they were before these other orders that were designed to go through this smaller channel in the warehouse are now more frequent. What's the best way to adapt them based on the capacities of the of the equipment, the labor that you have? And so that takes an incredible amount of analysis from historical data to say, if and when there was ever a surge, how did we handle that surge? Right? So it's traditional kind of forecasting where in the lack of, of kind of clear understanding of the situation, look to the past to benchmark and that's the best way that we thought of leveraging ml to look for those historical patterns to give us the best sense of which direction to take when chaos ensues?

Tom Raftery: 6:15

And how easy is it to get because it has to be individual local data per warehouse, how easy is it to get the data for that warehouse to feed the machine learning algorithm? So it learns the patterns?

Alex Ramirez: 6:27

The luxury we have with warehouse management systems, including, like SAP Wm is W. Ms is are there well oiled machines that depend on quality data coming in from any RP, a host system or order management system? They were built with the assumption that if the if the data isn't sound in its content isn't perfect in its formation, then it's going to get spit back, right? And so that's the luxury we have is that the W ms data starts pretty pretty well in terms of quality. And then it just kind of lives throughout the process. Sure, there's data quality issues, but some of those are more physical manifestations of, for example, the warehouse management saying, this location has this item. But it's not really there. Right? That's not the W MS is fault that's more operational and process driven that it is the system. Sure, death, taxes and bugs are the three certainties in life. But let's say that most of the W messes are soundly built, the data that we're tapping into is relational in nature, it's a snapshot in time. And it's why data warehousing, Business Intelligence projects have not really taken off in the W ms space, because how the data is stored is very, it's very, it's very snapshot in time type data set. And so the way we're tapping into it is the same way these Excel spreadsheets would tap into it using, you know, Native Connections like ODBC, or JDBC. But what we're doing is instead of just saying what's happening right, now, we've tapped into periodically certain objects, and we're taking this data, ingesting it into the cloud and building a time series data set, that then turns that very vertical, right snapshot in time data into the horizontal one that allows us to understand the certain velocities of the equipment of how orders are flowing, how people are moving, to then apply some of these rich techniques. And so the data is fairly easy to get to. We created cognitive apps to be very low friction and low drag. And that we we don't want to go into an IT organization and ask them to create new interfaces. Sure, we have API's and we can be an endpoint. But most of our implementations we've been data Wranglers where we can go in and say, I know where the data is, we have 100 plus years of collective experience building these systems allow us access in a secure fashion, making sure that it's all private and securely stored. And then allow us to create that derivative data set that then applies machine learning to great efficacy. Okay,

Tom Raftery: 9:16

okay. You came up on our on our prep call with a new acronym w o. s. Can you talk to me about that?

Alex Ramirez: 9:25

Yes, I I wish I had a better better acronym, Tom, because I think the industry needs a new acronym that starts with W and ends with s like, you know, it needs another as 400. But the operating system is apropos for what we're trying to do and that warehouses to date have been optimized in chunks with various applications, whether it's hard tech with robots and their control systems, or inventory management with warehouse management systems. And so these islands of applications While they have been tethered together through interfaces, you know, sockets and Jason and XML and flat files, etc. What they lack is that context that then this Operations Manager struggles to understand. And so like an OS that provides resources, whether it's ram or disk space, right to your machine, we want to be that operating system that provides the same resources to a warehouse instead of RAM, we're providing the right labor allocation to the various areas, right, instead of disk space, we're providing the right positioning of inventory and equipment that it's predicated on slow movers, for example, well, what's a slow mover today, you should maybe use some forecasting techniques to say this is what I predict a slow mover to be, etc. And so we see labor equipment configuration, the types of orders that you should be looking at an inventory positioning as the valuable fuel or resources to help drive the most usage and effectiveness from the applications. And so it it felt right to call it an operating system. We'll see if it takes off or not. Maybe it does, maybe it doesn't the luxury I have is we're a startup and we get to change pretty quickly how we how we brand the business?

Tom Raftery: 11:15

And is the solution you're providing? Is it an on prem? Is it a cloud? Is it a hybrid? Is it up to whoever gets it? How does that work?

Alex Ramirez: 11:24

Yeah, no. So we were we've been very intentional in creating a single instance, multi tenant environment. So we deploy on Google Cloud. Early on, in our founding, we were meeting with certain retailers, and we were hypothesizing on which cloud provider we should go with. And of course, we mentioned AWS was an option. And this retailer, quite aggressively, my data's in AWS, you're not going to get our business. And so guess what I believe in maniacally serving my customers. And when my customers potential customers are saying, my data does not sit in AWS, well, then you look for the next big players. And so we landed on Google Cloud, our CTO was very familiar with it. And it just made sense for us. So it's all subscription based, our pricing models very easy because we don't want to complicate matters, with number of transactions or optimizations, users etc. It's a per building subscription, typically annually, or bi annual subscriptions and unlimited users for each warehouse.

Tom Raftery: 12:35

Okay. And talk to me about the Enterprise's, you know, nailing down operational visibility that before no bad practices.

Alex Ramirez: 12:46

Yeah, so some of these I can mention my name others, I'll just refer by by industry and vertical market. So for was our first client. And they've been an outstanding partner of ours really pushing the envelope in terms of how do you create this operating system application in an environment that has many different applications, best of breed, warehouse management systems, best of breed warehouse control systems, lots of automation and interesting flows in their e commerce network. We started with our first implementation at four back in 2019, actually, so we found the business 2018 did a bit of vision market fit in 2019, after we raised our first round of capital, we partnered with Sephora. And so we went live with them had some really interesting learnings around how to not be so aggressive with the prescription side of machine learning our original vision was, let's just be kind of this whisper in the ear of the Operations Manager saying hey, Tom, do this at this time, or do that at this time. And believe it or not, these team members wanted to trust but verify. And they believe that we were smart, but they didn't think we were that smart. And so we came in hot with a lot of recommendations. And instead we wound back a little bit to really provide a platform that through visualization, they could start to feel trust towards the evil AI in the in the background. And in 2020 we quickly started to grow the the client base. Medline who's on the website was our second customer and for the for their operation. It's a fully automated operation with different pick engines from different providers, a legacy warehouse management system, similar pain points of multiple applications, but each application while being integrated are basically sibs for that context and how to manage a grocery company. Super center company out of the Midwest came in At the end of 2020, presenting the same problem but completely different building profile where it was forklifts and it was fact fresh product. But similar pain in labor positioning, cycle time, visibility etc. And 2021 has been an absolute explosion in, in customers. But similar pain points, Tom, that our thesis back in 2018 was irrespective of whether you are fully automated or manual, if you're a warehouse in North America or in the globe, or soon on Mars, if Ilan musk gets his way, there's going to be a human being that has to make sense of all the data that's being streamed through some system, right. And when you look at the constituency that normally serves in this role, you know that I was a math major in college and I couldn't perform this job. It's incredibly difficult. And so we want to make this an easier role and an easier job for these operators to fill to create that, that resiliency. And so the types of pains that we solve for clients, like Sephora and Medline are uniform in that they are telling me how many people I need? Where do I need them in all the various areas of the warehouse, irrespective of the applications that I have inside my building? Right. So imagine two different applications are bifurcating a warehouse in half, and you have system a on one side and system B, standard, integrations aren't going to say, Well, hey, I'm running dry here, you do something to make me better or I'm overloaded on one side, right? You need to slow down the decision making on the other. Instead, they're sending standard interfaces that simply say, Hey, here's an order. Here's an item, right? This is a truck that's coming, you better get your act together. Until that that sieve is what we're trying to fill in with context with data with prediction and prescription. It's uniform across our very varied client base.

Tom Raftery: 17:39

Yeah, that was gonna be my next question. In fact, you I mean, you talked to Medline and you talked retail, so is it that is it, you know, everything from, I don't know, fresh produce through to technology stacks through to pharmaceuticals through to you name it.

Alex Ramirez: 17:55

Yes, absolutely. So some of our customers are, as I mentioned, Medline, with, you know, the b2b, wholesale, but their profile and the speed in which they fulfill, they might as well be an e commerce business, they serve their customers, with the same maniacal focus that, you know, we serve ours, and speed is of the essence for them. Same thing with for being in their e commerce network, and we're across all three of their e commerce operations in North America, another company in the West Coast, a shoe distributor, fully automated building, same thing. It's about service level and speed through the through the facility, one that we just signed as a fortune, I think fortune 10 beverage company, that same thing, labor positioning, service level attainment. And so you name the vertical market, the industry, we're seeing everybody really say I have data, I have plenty of control towers in my warehouses with, you know, 18 different screens and somebody's playing, you know, like they're in some sort of game in the matrix. And I have sub optimal results. And so instead of driving more automation into the building, because they're either vertically or horizontally tapped out of real estate to throw any more bots into the operation, what do you do? Well, then you focus on the people that are driving the operation, right. And to date, they've just been under invested in, I think, in the last five years, if you add up kind of the market growth of warehouse management, there's been about $10 billion of investment in the W ms space. But yet, we have never walked a warehouse and seen the operations management be perfectly nailed down, because there's just way too many things. And these systems were built more for workflow execution than they were for planning and management. And so we're really enthused about the uniformity in application and value proposition of I just need to manage my building faster, cheaper, and with higher quality. And if I can improve the human beings decision making, then I think everything else will follow then we may want to rationalize robots and you know, drones and robotic dogs, etc.

Tom Raftery: 20:23

And speaking of are those kind of technologies actually starting to make it an inroad into warehouse management. Yes. Robots you mean? Yeah. Yeah, absolutely. Or drones or robot dogs are?

Alex Ramirez: 20:37

Yeah, I've yet to see a robotic dog in the warehouse. But you know, I won't be surprised. So the way the way my golden doodle pulls me around, I can see how they'd be pretty good goods to person station if you had the right treats. So I think yes, absolutely. We're seeing explosion in robotics and autonomous mobile robots, you know, the likes of six river systems and locusts and fetch, of course, goods to person automation, right, fixed assets, anchored to concrete, but keeping the the person you know, position in the same area. The the issue, in the non contrary and belief we have with robots is while it is no longer about cost, justification, and ROI on labor reduction, right there, there is a resiliency thing and a cost of business, if your competitors are going faster, and instead of purchasing or re buying stock, they're reinvesting in the business through these capital expenditures, automation is is going to continue to be a thing. So that's, we're not going to stop it with cognate ops. But you find yourself now thinking about 400,000 warehouses in North America. And is everybody going to turn to robots as the solution and even when you do, the robot isn't going to come in and take over the entire warehouse, right from inbound to outbound, you're going to introduce the robot in a certain area, so therefore becomes another island of automation or system. And guess what, you've now created another integration point that is a sin for contacts that now needs even more cognitive apps help. Because if you don't make the right decisions, instead of somebody that you're paying 1516 $22 an hour, right to sit idle. Now you've got an expensive piece of equipment, they're sitting idle, because you've made the bad decision. So we're bullish, whether it's a roboticized world or a manual, you know, Westworld, or wild west Westworld. I guess it's an automated one, right. But the the pre Westworld one, the human being orchestrating things in the warehouse, I think is going to be a constant for the next few decades. I think our existential crisis, Tom, if I can be frank, is the fully automated, you know, machine learned operation? Maybe it's because I run hot, I may not live to see you know, the old age, but maybe not in my lifetime, are we going to see that warehouse where you just press a button and everything goes up? Until then, I think there's plenty of room for robotics and automation and cognitive ops where we kind of think about the human in the loop aspect of machine learning and operations management.

Tom Raftery: 23:28

Okay, so you're saying, Hey, I will augment people rather than replace them?

Alex Ramirez: 23:33

Absolutely. I think the, the, Our vision is to and we went through a process with our team members where we talked about vision, mission, purpose and values. And I love our vision in that what we're trying to do is unleash ingenuity to give rise to more resilient businesses. And it's the first two words of unleashing ingenuity that we focus on when we talk about cognate ops being a human in the loop like a decision, tribal knowledge augmentation tool versus a replacer. Because there's so much creativity that gets lost through attrition, or through the constant barrage of stress and pain and confusion in warehouses when they're going fast. That How do you adapt that building when our team members do not have any oxygen to feed the creative side of their businesses and so I've watched some warehouses where it is a perfect example of human ingenuity you think wow, that's there's no way I would have ever thought of handling You know, this type of of problem this way. And so why not give those individuals the time and the liberty to think to create a plan a whiteboard versus literally playing the whack a mole game 24 seven managing operations and so the unleash unleashing ingenuity, I think plays a huge role in warehouses today and well into the future.

Tom Raftery: 25:05

That's all fascinating, Alex but where to from here? what's what's kind of in your three to five year plan? what's what's next on the line for a company of ops.

Alex Ramirez: 25:14

So I think the the vision we have for cognate ops is we talked about becoming a collective memory for supply chains. And it starts with helping one warehouse right with one operations manager and slowly growing that way or quickly growing in our case, but when I think about extending In the application up, and performing supply chain visibility, not the you know, plenty of parties perform supply chain visibility, its transport its, you know, track and trace its visibility to all the various nodes. But no one has really created this really interesting data asset of true warehouse performance that I think we can tap into to connect many warehouses together, not just within one enterprise, say so for us three buildings and give supply chain leadership visibility to who's performing better, who's performing worse, and provide predictions around like, this is a thing that you can adopt across all three to become that rising tide that raises all ships. But imagine being able to tie one enterprise with another on a non competitive basis to say, hey, the way that you do shift planning the way that you do labor allocation, the way that you leverage your equipment, you can learn from each other, and, and become this kind of collective memory for for supply chains, we were really excited about that part of our vision, we don't want to contend with broader supply chain visibility tools, we think we can be very complimentary, and that we're going to get, we're going to own the warehouse data asset, and can be leveraged for many, many visibility use cases. But I think just giving someone understanding and comprehension of who's performing better, and why I think is going to be pretty important.

Tom Raftery: 27:05

Super Saver. Alex, we're coming towards the end of the podcast. Now, is there any question I haven't asked you that you wish I had, or any topic we've not touched on that you think it's important for people to be aware of?

Alex Ramirez: 27:18

I think the question that we haven't talked about is, and maybe this is a discussion between you and me, because you've you've seen a lot of these podcasts, and you've seen a lot more than than I have is, how do you feel? Or maybe I'll give you my take, and then we would love to hear yours is with the availability of data? How are warehouse managers kind of evolving into being data literate, and data analysts? Right? Is it possible to push this data analyst position down into the organization? Have companies found a way to train to reinvest? And if so, how have they done that, because that's a struggle that we have candidly with, you know, introducing this very optimized application is one of the learnings we had back in 2019 was, instead of starting first with the prediction and prescription, let's go back and kind of evolve through how to become very data and information literate through some descriptive analytics, then introduce some predictive ones then start to introduce prescriptive. Is this a challenge that you're seeing across the board, whether it's in supply chain or in finance? Because we certainly see it?

Tom Raftery: 28:36

Yeah, no, absolutely. data literacy is something that many, many organizations are struggling to come to terms with. It's, it's quite a new field. I mean, it's it hasn't been comparatively speaking that long since we've deployed the kind of sensors that are generating the data that are allowing us to, that are challenging us to to interpret that data and make meaningful decisions from it.

Alex Ramirez: 28:59

Yeah, no, indeed, We that's a school of hard knocks lesson for us has been to treat these individuals with kind of the, the courtesy of, yes, we want to introduce a new paradigm, if you will, of operations management. And just because I think our UX and UI is really intuitive, and it's really easy to train up doesn't mean that they're going to change behaviors. And so there was a book that I read a long time ago called the power of habits, I think about it often. And it was a story about World War Two and the mothers that would stay back and try to reintroduce or introduce kidney and liver and tongue because all the prime cuts were being sent to the frontlines and so they said just start to slowly introduce tongue kidney liver into like meatloaf with other ground beef and before you know it becomes everything, you know that that people want. And so I think about our evolution in warehouses that way, right? We were the kidney. At some point, we're going to taste delicious but right now, let's make sure that we're empathetic to to the the pain they have, which is I don't have much training, you're telling me that I need to use data, but I've never used data before. And so I'm glad that it's across the board and not just with with the warehouse.

Tom Raftery: 30:27

Alex. That's been fantastic. If people want to know more about yourself, Alex are up cognate ops our warehouse operating systems or any of the other things we discussed in the podcast today, where would you have me direct them?

Alex Ramirez: 30:40

Our website is a easy place www dot cognate ops comm c o G and it o p s comm You can also email us at info i n o at cognitive ops.com. Super, Alex. Fantastic.

Tom Raftery: 30:55

Thanks again for coming on the podcast today.

Alex Ramirez: 30:57

been a pleasure, Tom, thank you so much.

Tom Raftery: 30:58

Okay, we've come to the end of the show. Thanks everyone for listening. If you'd like to know more about digital supply chains, head on over to sa p.com slash digital supply chain or are simply drop me an email to Tom raftery@sap.com. If you'd like to show, please don't forget to subscribe to it and your podcast application of choice to get new episodes as soon as they're published. Also, please don't forget to rate and review the podcast. It really does help new people to find the show. Thanks. Catch you all next time.