Sustainable Supply Chain

AI, ML, And Data - A Chat With Sisu Data's Joel McKelvey

June 13, 2022 Tom Raftery / Joel McKelvey Season 1 Episode 233
Sustainable Supply Chain
AI, ML, And Data - A Chat With Sisu Data's Joel McKelvey
Digital Supply Chain +
Become a supporter of the show!
Starting at $3/month
Support
Show Notes Transcript

Our machines and processes are creating increasingly larger volumes of data - how can we handle this, what insights can we glean from it, and what are some lessons learned?

These are some of the questions I put to Sisu Data's Joel McKelvey in this episode of the Digital Supply Chain podcast.

We had a great conversation spanning why AI ML solutions are increasingly necessary for supply chain data insights, Sisu Data's recent Harris Poll showing the implications of a poor supply chain experience, and how best to solve these problems. 

I learned loads, I hope you do too...

If you have any comments/suggestions or questions for the podcast - feel free to leave me a voice message over on my SpeakPipe page or just send it to me as a direct message on Twitter/LinkedIn. Audio messages will get played (unless you specifically ask me not to).

If you want to learn more about how to juggle sustainability and efficiency mandates while recovering from pandemic-induced disruptions, meeting growth targets, and preparing for an uncertain future, check out our Oxford Economics research report here.

And if you want to read up on our Industry 4.0 blueprint repost, head on over to https://www.sap.com/cmp/dg/intro-industry40/index.html, and if you liked this show, please don't forget to rate and/or review it. It makes a big difference to help new people discover it. Thanks.

And remember, stay healthy, stay safe, stay sane! 

Elevate your brand with the ‘Sustainable Supply Chain’ podcast, the voice of supply chain sustainability.

Last year, this podcast's episodes were downloaded over 113,000 times by senior supply chain executives around the world.

Become a sponsor. Lead the conversation.

Contact me for sponsorship opportunities and turn downloads into dialogues.

Act today. Influence the future.



Support the show


Podcast supporters
I'd like to sincerely thank this podcast's generous supporters:

  • Lorcan Sheehan
  • Krishna Kumar
  • Olivier Brusle
  • Alicia Farag
  • Joël VANDI
  • Luis Olavarria
  • Alvaro Aguilar

And remember you too can Support the Podcast - it is really easy and hugely important as it will enable me to continue to create more excellent Digital Supply Chain episodes like this one.

Podcast Sponsorship Opportunities:
If you/your organisation is interested in sponsoring this podcast - I have several options available. Let's talk!

Finally
If you have any comments/suggestions or questions for the podcast - feel free to just send me a direct message on Twitter/LinkedIn.

If you liked this show, please don't forget to rate and/or review it. It makes a big difference to help new people discover it.

Thanks for listening.

Joel McKelvey:

So you have to use tools like machine learning to do it. And with machine learning, you can remove the requirement for a lot of very expensive, staff. And help people do what they do best rather than have everybody have to be a data engineer or everybody have to be a machine learning scientist

Tom Raftery:

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 at SAP Tom Raftery. Hi everyone. Welcome to the digital supply chain podcast. My name is Tom Raftery with SAP and with me on the show today, I have my special guest Joel, Joel, welcome to the podcast. Would you like to introduce yourself?

Joel McKelvey:

Sure. Thank you so much, Tom. My name is Joel McKelvey. I'm the vice president of product marketing at Sisu data in San Francisco and, at Sisu we build AI ML powered analytics, that our customers are using to help tackle some of the biggest supply chain issues that they have. So I'm super excited to be here. Talk a little bit about that and, and learn from you and share what we're doing with the audience.

Tom Raftery:

Cool. Thank you. Thank you. And tell me a little bit about the AI ML analytics solutions that you have for your customers. What problems are your customers having that they need your analytics to solve?

Joel McKelvey:

Yeah, I think it's more interesting the problems they have than what we're doing. Although I'm certainly proud of our products, but what we've seen is. A real wave of data, hitting supply chain companies, first tier data, the data they're collecting from, um, all of their supply chain operations, second tier from their suppliers or their logistics partners, even third tier part suppliers to their suppliers, to the legit. So there's tiers and tiers of data. And it's really important to, to note that all that data together in one place is overwhelming current systems for data analysis. This, this data is, we refer to it as machine scale or cloud scale, but it's really, a lot of it is machine generated. It's sensor data. You'll see it in, um, Environmental sensors on, uh, shipping containers, crossing an ocean for example, heat, and whether they're upright and how much jarring occurs. You'll see it in a supply chain, from the manufacturing side. So, just fully instrumented, manufacturing, product lines,

Tom Raftery:

Mm-hmm

Joel McKelvey:

things are coming off the end of the manufacturing line. it's a lot of data and where the industry.

Tom Raftery:

I'm sure is like saying it's 37 degrees. It's 37 degrees. It's 37 degrees. It's 37 degrees and you're like,

Joel McKelvey:

Which is really useful. Information, but it's, it's a machine giving you the information and the problem is humans get overwhelmed by that volume of information, Right. We are, we're cognitively best capable with maybe around 20 things, but we're really good at intuiting from the correct data, a good outcome for our businesses. And so we, we have a, we have a bit of a challenge of this supply of data is too big for human beings to really analyze alone. Now, Smart supply chain analysts have been coping with this pretty, pretty heroically to date and the way that they do it is they prioritize.

Tom Raftery:

Mm-hmm . There's

Joel McKelvey:

look at those items, which might have the biggest impact on the bottom line. They might be looking places they're familiar with. If they've been working in supply chain for a, while, they're smart enough to know I should look here and here and here and here and here, um, and dig through that data.

Tom Raftery:

there's a journalist in, a professor of journalism from NYU. I think he is, his name is Clay Shirkey and he says there's no such thing as too much data. All there is is filter failure.

Joel McKelvey:

Filter failure is a really good way to describe it. It's a, it's a signal from noise problem, right? And when you have machine scale data, the best way we have found to sift through it is by using machine learning. So you can use a machine to help conquer this problem of machine data. And it's not. Because humans are bad at doing things from an analytics perspective. It's that we need to take it from cloud scale to human scale so that human analysts, these super intelligent supply chain people can use their brains effectively rather than spending all their time just trying to sift a few nuggets from, uh, this huge pile of sand of data that we have. Right. If you're looking for an individual grain, it's better to have a machine, find the grain for you and then analyze the grain.

Tom Raftery:

Right.

Joel McKelvey:

Yeah. Grain might not be the, Tom might not be the right word to use in the context of data, but , but, uh, it's sifting a signal from noise and it's a, it's a big challenge. It's one of the reasons why businesses are challenged to make data driven decisions. And we've seen some analysts firms saying that, only about half of all decisions are actually made with data in a business. And that, that is reflective of supply chain as well because, because we're humans, we're actually pretty smart. I, they say no data. I would say based on experience or feel smart people, but still, um, that's surprisingly low. And as the amount of data goes up, the amount of data that gets used for a decision that uses data goes down, the percentage is going down because there's so much, there's so much more signal. There's so much more noise and not as much signal in that noise from a percentage perspective.

Tom Raftery:

So your machine learning solutions are helping fix that....

Joel McKelvey:

Yeah, you can use math to solve some of these problems. and those types of math are, relatively well understood areas of things like, trend and anomaly detection in data, correlative analysis, root cause analysis, um, some, decision graphing and charting capabilities that, we've been thinking about for a long time, but we haven't really applied in the data space. Really. I think this huge wave of data. We've been focusing as, and I'm a data person, not a supply chain person. So that's where my background lies. The, the data industry has been focusing on how do we just handle it and put it somewhere. We've been trying to figure out how to ingest it and store it. And what we have lost is this last mile of how do I put the right data in front of the right people? And, if you've been in data for the last five, six years, like I have, we, we, at the early part of that period, we saw a big movement toward what was called the democratization of data, getting all the data to all the people so that they can make all the right decisions. And, uh, it's a bit provocative, but I think that was a faulty approach. Because again, if I give all the data that we collect to my business, Raw

Tom Raftery:

Mm. Yep. It

Joel McKelvey:

will blow their minds cuz there's just too much of it. What I have to do is a much more curative mindset. I have to do that filtration as an analyst and put the meaningful items in front of my business. And the way I can do that is using AI ML to sift what is most meaningful and in supply chain, we're talking about metrics that we can measure relatively easily; things like, delayed orders, missing orders, late arrival to customers, lost components in the system damage rates. Right. We can come up with all of these types of metrics and we can run if we had the compute power, against all our data, a process that determines what in that data actually impacts that metric. What are the things stack ranked in that data that, that gets us to understanding why that metric changes, analytics to date has really been around what's changing and we've relied on this human magic of figuring out if I know what changed, I can probably intuit why. And it's hard to intuit why with all the data now. And so you can use AI ML to do that. Now, there is one really interesting, hidden piece of, success when you do that in bulk with math. So I just mentioned that you can find those areas of the data, those population subpopulations in the data that have the biggest impact on a given metric like manufacturing, faults or delayed shipments. But when you use the math to process all the data, you find two sides of sort of the same measurement can be valuable. One is what has the biggest impact. The other is maybe several million or several billion things that do not have a big impact. And that for a business that is trying to control risk is very, very important because smart humans can pretty quickly get to maybe two or three things that might have a big impact, but they're not gonna be able to check everything else to see whether or not there's what we would call a black swan or a surprise somewhere in that data that might be driving a changed metric or a decline in the successful deliveries or whatever. And so, thinking not just about what has the greatest impact, but all the things that don't really have an impact can be super valuable in the supply chain space,

Tom Raftery:

So it's allowing you to rule out probabilities.

Joel McKelvey:

Rule out massive probabilities. So if you look at you know, a, a data set with maybe a hundred columns and a many millions of rows of data. And you look in there and you look at all the possible factors and combinations and factors in there that might impact a metric, a data set that big would have, depending on the data set. Of course. A billion possible things that might impact your damage rate and a human being is going to be able to test to see whether or not one of these components actually impacts a metric like damage rate. Probably one every few minutes, right? Humans are not slow, but you can use cloud computing to test all of them in 5 6, 10 seconds, right? Like that's a lot of testing and now you've ruled out all of the things that you really were never going to get to as an analyst, in addition to really knowing which ones you should get to

Tom Raftery:

And potentially finding some surprising outcomes.

Joel McKelvey:

potentially finding some surprising outcomes now, as data people we had heard, last year, around the holiday time, a lot of we've seen a lot of press, just like everyone has around supply chain, supply chain, disruption, and particular from that sort of brand to consumer fulfillment component of the supply chain.

Tom Raftery:

Mm-hmm

Joel McKelvey:

we've heard, we've heard all about all the delays we've heard about all the concerns everyone had with supply chain as a result of the pandemic and, other factors and one of the things we wanted to do because we have such a, a wealth of customers who are in that space was find out more. So we did commission a poll and, um, we had some really interesting findings. A Harris poll, in the United States, we had some really interesting findings about what it looks like actually in supply chain, out in the world, and the ramifications of that, which was super interesting as well. So cuz we're data people, we love data so we gathered some data on that as well. And we have some really interesting findings we found there too. So when we look across the data of how people are buying and consuming and how they're impacted by supply chain, we had a couple of really interesting findings. So, um, let me, let me run through that super quickly here. Yeah, so we found, of our sample set, which is reflective essentially of all Americans in north America was our sample set. Um, we found nearly all of them had done some type of online purchase, which should not surprise anybody, particularly in the time of coronavirus, but 89% of our respondents had purchased gifts online in the last year. And that's reflective of our own experience and probably the experience of a lot of people listening to this. Um, you know, almost 90% of people. What we hadn't expected was how many supply chain issues had impacted those people. So of those 89%, 71% of people had experienced a supply chain issue as we go forward. And what's interesting from us as data people, because we, we work with people who are collecting data and analyzing data all the time. We think of the supply chain as perhaps more sophisticated from an instrumentation standpoint, as we think it was, we hear a lot about this, volume of data, this massive data that, people are receiving. One of the findings of those, those 71% who had experienced supply chain issues, almost a third, had a missing order where they'd ordered something and it just went missing. And that was a really big surprise for me in particular, where I assumed if somebody's tracking all those things, they're getting scanned in and out of every place they're going. And yet a third of them are going missing. there's.

Tom Raftery:

experience myself, not so long.

Joel McKelvey:

Yeah. And, you know, I experienced delays, um, a couple of months ago when I was ordering some furniture as well. So, you know, half of people are seeing processing delays. 64% of people are seeing late deliveries, right. It's it was really pervasive and it remains pervasive. And so, this is a really interesting and impactful space for most people, even not data people, even not analysts, even not supply chain people. Right? This is, maybe it's the golden age of supply chain. Tom, I don't know where everyone knows what supply chain is now, but maybe it's not golden age cuz it's not everyone knows it in a particularly good way.

Tom Raftery:

What can be done, I guess, to, to, to find that one third of items that went missing and , or to stop it from happening in, in future.

Joel McKelvey:

Yeah. You know, we work with big online marketplaces and online retailers a lot, and this is a very expensive problem on the orders of tens or hundreds of millions of dollars per year for these companies. It's a big issue and we know it's a big issue, digging into the data around it. And just having talked to a lot of these companies, we've gathered some really interesting information about how, how big a problem it is. Right, three quarters the people we've talked to are changing how they buy. They might be shifting from brick and mortar to online or online, back to brick and mortar because the online is perceived as maybe having more delays. If you can see it and buy it, you certainly get it on time. Right. but, they're starting to shop earlier. They're finishing shopping earlier in the year for the holiday season. They're more concerned about delays. They're more aware of it. And so what this means for a brand or an online vendor is these delays are going to be impacting your business. There's more awareness out there. And we saw a lot of consumers saying that boy, we're not happy with companies that, that, give us a delayed, shipment. You know, the supply chain is, is not a chain. I think it's a misnomer. It's more of a web right. There's a lot of things, feeding a lot of things. And, something like a late delivery has a real impact on how the person who sold it to you is viewed.

Tom Raftery:

Yep.

Joel McKelvey:

And, uh, that brand or that vendor that has sold you, that product that comes delayed may not be responsible for the delay. right. It could be your logistics vendor, your warehousing vendor could be the manufacturing delay. It could be a shipping delay across, across an ocean again. Right. We saw issues with the port of Los Angeles here in the US. Right. So there's many things that could in that chain. That's that supply chain that could impact something like a late delivery, but the brand is on the hook, no matter what, right. It's it doesn't really matter. You're the buck stops there is, is really how that works. Half of people think that the company's gonna be re should be responsible for lost orders. Half the people think that the company should be responsible for processing delays even, these are not their fault. And. 15% of the people from our poll said, they'd never order from a company again, if they experienced that

Tom Raftery:

Wow.

Joel McKelvey:

that's one, 5%, 15%, right. That is a huge number. Now, anecdotally, I don't know whether or not that number is simply a respondent trying to send a message, like stop, delay my orders. Right, but you know, you do feel powerless and, and frustrated when that sort of thing happens and it makes a ton of sense. So we, actually, it was part of the poll we commissioned, we asked what can a brand do about this? What can a, vendor do about this? And again, digging into data as a data person. I think it's really interesting how recovery is not hard. It simply takes a bit of attention on the part of the company. So things to take a disappointed customer and help them reengage with the brand and feel better, you know, just over a third of them said an apology would help, right? So that's an apology is relatively free. It's right. It doesn't, I mean, you have to pay someone to make the apology, but right. This is not, this is not hard. This is not, you know, costly, but it is, a personal response to a personal concern on the part of the customer. And that personalization drives increased brand loyalty in fact, rather than decreased brand loyalty. There's lots of things that also, you can reengage with customers on at least according to the poll data. And I agree, I feel the same way personally. You know, if you give me a discount code for my next order, as a, as a form of apology, not only will I get the apology benefit, but also you're going to invite me back and you're gonna have a second chance to make it right. And to regain that loyalty, same with like a coupon code or a discount code or a, an invitation to a future sale. And almost half of the respondent said, Yeah, they'd be really interested in that. And so, and 93% said you could do something that would help So, so. It's not the end of the world when you have some type of supply chain disruption, but it is something that companies need to focus on. Now we have, a company we've been working with that has experienced, really four types of concern when delivering to end customer. And again, I'm talking about sort of that online purchase through the fulfillment channels to delivery at someone's home. And there's, there's really four problems that they had seen that they were trying to tackle, delayed deliveries, damaged deliveries. And that could be damaged either at point of manufacturer or during the shipment process, lost orders and processing delays. Right. So these sort of, we're not sure what happened to your order. One of those are more mechanical and more data driven. The others are more sort of getting it to you on time driven. Right, and you know, looking across their rich data set, they had a bunch of data from a bunch of sources, including their logistics partners, both warehousing and transportation, from their manufacturers, they were able to come up with, several hundred columns of data and, almost a billion rows I believe of transactions that they could analyze. And they were able to find some really interesting things. They found. One of their shipping partners was more likely to damage large products than smaller ones. One of theirs was much cheaper when shipping smaller, but not larger, but they were better at keeping the larger ones intact during the shipping process and were less likely to lose a small package, right? So you, you have these sort of trade offs. So as a company, digging into the data, using a tool like, or another one can be really valuable when you are renegotiating your shipment contract, you can say, Hey, we're not shipping any of those big SKUs with you anymore. You damage them too much. And that shipping partner can say, well, maybe we'll give you a discount, or maybe we'll insulate you in some way from damage and the costs associated with damage. so you can renegotiate that contract. You can send SKUs against one shipping partner versus a different shipping partner. You can just choose who you ship with based on size or package type or source or destination, right? When you, when you can correlate those things to increased damage rate, then you can save a ton of money doing something relatively simple, not just save money on your shipping contracts, but also. Less damage, less impact on the customer, greater customer retention, less churn repurchase rates go up. A average order value goes up, right? All those things that are good, get better, all those things that are bad get smaller and that's the power in the data that we're seeing as you, as you analyze these wide data sets all this data associated with supply chain.

Tom Raftery:

And how many people in supply chain are literate?

Joel McKelvey:

Boy, that's a hard one, right? I was recently in a conversation with a, a bunch of, German executives from Northern Germany, and they were mostly concerned. These were, uh, CIOs, CEO level people. They're mostly concerned that they don't have enough data. more than anything else. Uh, I think the quote was, you know, we do a bunch of manufacturing, but we're not Tesla. Right. We don't have a sensor in every single thing everywhere, in every part of our, manufacturing line and our supply chain, because. We've been around for a hundred years. And so we have, we have some legacy there that, that's really important. So, I think the first thing for getting these, uh, manufacturing or supply chain oriented companies on board is really taking a look at what you need to instrument and what you don't and getting the data in place. The people problem, which you mentioned is a real challenge of getting folks data literate and.

Tom Raftery:

Sure.

Joel McKelvey:

I see data literacy as a bunch of stages. Our CEO here at Sisu data recently wrote a blog and he said, you know, it's hard to get enough analysts. He said, if you were able to buy to hire enough analysts to do all the things you want to do with data very soon they would be the largest organization in your business, and that's not financially tenable, right? Like that's not, that's not. Okay. So you have to use tools like machine learning to do it. And with machine learning, you can remove the requirement for a lot of very expensive, staff. And help people do what they do best rather than have everybody have to be a data engineer or everybody have to be a machine learning scientist. There's really two types of AIML out there in the world right now. And it's the stuff that you do yourself as a company and you have to hire AI ML specialists. There's tools like Sisu, which are operationalized. They've already built in the way to run the data through them and you get the results out. And that can save you a ton of, of literacy challenges as you go forward. But you still have to be able to tell as an analyst or as a data team, a real story to your business about, around three specific dimensions. You have to be able to tell them what happened, you know, did you, did your metric go up or down? Are you having more damage today or more delays today than you did yesterday? You have to tell them why, which is what we've been talking about a little bit. How do you dig into the data and get to the causes of that change? And then you have to be able to tell them what next, how do you prioritize how you tackle that problem. And as a data team, tools like ours and others can really help you get to those three parts. More technically it would be a, a, a descriptive diagnostic and predictive type of an analysis against the data. But that's what you have to tell the business. Just now giving them a, your metric went up is no longer really seen as state of the art. Sure your C staff, your less data literate, people would love to know the metric has gone up or gone down, but really what they need to know is why and what next.

Tom Raftery:

Makes sense. Makes sense. Joel, we're coming towards the end of the podcast now. Is there any question I have not asked that you wish I had or any aspect of this we haven't touched on that you think it's important for people to be aware of?

Joel McKelvey:

I think there are two big dimensions when you're talking about data in supply chain that, as a data person, or even as a supply chain person, we tap dance around a little bit and the first one is where do I get all this data? how do I get all this data? And I mentioned those as companies I was speaking with, who said, you know, we, we don't have that data on our manufacturing line. We don't have that data in our, in our supply chain. And that's not an easy answer, I don't think anyone can answer, how do I get all that data from all my vendors, but increasingly people are preferring supply chain partners, logistics partners, who are willing to share that data who are willing to, provide it to you for analysis as you go forward. And I would say that, you know my advice to people who say, how do I find what's going on in my logistics partner? Ask, put it in your contracts, cuz it's going to be important to you. And what we've found in that Harris poll data is doesn't matter if your logistics partner is the one who's messing up, it's still on you. So you're going to want that data. And that's something for a discussion. The second problem, I think that we don't talk about is when I have it kind of, what do I do with it? Where do I store it? How do I manage it? And this is a, a problem that, the data vendors are still working on, which is data engineering, normalization, cleaning, joining. This is a tough problem. The industry continues to iterate on it and work on it. But, so I don't have an easy answer there at all. There's a lot of tools out there that can help. But none, I think solves this problem of, uh, we are getting more data from more places all the time. How do I reconcile it and make it usable in one place? But you know with, AI ML, we're seeing improvements there too. And once we have that data in a place for analysis, we can absolutely save a ton of time on getting it in front of the right people. So there, there is hope there. And I, I see that we have a, a pretty rosy future when it comes to getting and using the data. But, there's no easy answers to boy, I have more data than I know how to handle, how do I handle it? And the answer is, you know, data teams are in high demand, but we see the value. So the takeaway is we can see the value. When you look into the data, you can find things that can help your business. That's why we're here talking. That's why people are digging into supply chain data and, we'll continue to do it as a vendor. You'll continue to do it as, as supply chain experts and, we see that there's a lot of benefit to be had here.

Tom Raftery:

Cool. Cool. Cool. Joel, that's been really interesting if people want to know more about yourself, Joel McKelvie, or about Sisu data or the Harris poll that you mentioned, or any of the things we talked about on the podcast today, where would you have me direct them?

Joel McKelvey:

Yeah, well, so of course it's super easy to go to sisu data.com and look at that. But, um, I would recommend for people who wanna look specifically supply chain stuff, just Google up, Sisu that's S I S U data and then supply chain, and we will pop right up at the top of your search and you'll be able to find, an infographic and a blog and a white paper and a bunch of other stuff that talks about, some of what we're we've learned and what we've been finding in, in supply chain space with data.

Tom Raftery:

Excellent. Joel has been fascinating. Thanks a million for coming on the podcast today

Joel McKelvey:

Oh, thank you for having me, and, looking forward to, to working with you and your listeners. It's been great.

Tom Raftery:

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 sap.com/digital supply chain, or, or simply drop me an email to Tom dot Raftery @sap.com. If you like the show, please, don't forget to subscribe to it in your podcast application at choice to get new episodes, as soon as they are 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.

Podcasts we love