The Data Stack Show - 199: How To Use Data Analytics and AI To Increase Profitability With Smarter Procurement, Featuring Cameron Jagoe of ProcureVue

Episode Date: July 24, 2024

Highlights from this week’s conversation include:Cameron's Background and Journey in Data (1:49)Running a Bakery (3:03)Applying Analytics to Bakery Operations (7:07)Reevaluating Business Operations (18:08)Optimizing for Profitability (19:09)Working at Newell Rubbermaid (20:11)Value Engineering Projects (22:11)Starting a Center of Excellence (24:53)Productizing the Approach (29:48)Tech Stack for Data Analysis (31:40)Data Cleaning and Classification (35:16)Market Build and Pricing Accuracy (37:13)The AI Tool as a Pointy Stick (38:20)Sourcing and Sales as Two Sides of the Coin (41:04)Challenges with Parsing Data (44:06)Personal Journey and Company Success (46:44)Final thoughts and takeaways (47:45)The Data Stack Show is a weekly podcast powered by RudderStack, the CDP for developers. Each week we’ll talk to data engineers, analysts, and data scientists about their experience around building and maintaining data infrastructure, delivering data and data products, and driving better outcomes across their businesses with data.RudderStack helps businesses make the most out of their customer data while ensuring data privacy and security. To learn more about RudderStack visit rudderstack.com.

Hi, I'm Eric Dotz. And I'm John Wessel. Welcome to the Data Stack Show. The Data Stack Show is a podcast where we talk about the technical, business, and human challenges involved in data work. Join our casual conversations with innovators and data professionals to learn about new data technologies and how data teams are run at top companies. Welcome back to the show. We're here with Cameron Jago.

Cameron, we're so excited to chat with you. Awesome. I'm glad to be here. All right. Well, tell us a little bit about who you are and what ProcureView does. Sure. So I'm the original founder and current CEO of ProcureView. And it's been my passion project for basically 13 years. And what we're focused on is improving, they say improving profitability for businesses by leveraging their strategic sourcing. You know, one of the axioms that we go off of is most companies are about 10% operating income, which means for every $1 we help them save on the purchasing side, it's worth $10 of sales. And it's right to that bottom line. So that's our focus. And sort of our claim to fame is just our really in-depth and really deep granular

analytics that we provide. Very cool. Well, I can't wait to hear more about how you do it. Yeah. So Cameron, this area is very near and dear to my heart. I actually ran a procurement team for a short amount of time. So I'm really excited to dive into that. I love data as well. So I'm excited to dive into that. What are some topics you want to cover? That's a good question. i didn't expect that one right out of left field yeah that is covering off guard uh i i can sort of chat about it you know my my hobbies until kind of the past couple years have been around this stuff you know this was what i was doing for my day job and then i'd go home at night and yeah i'd code up algorithms and projects at home

and then sell them on the side and you know once it became the full-time job i was like okay i gotta find something else to do i gotta find another hobby yeah right yeah yeah so i don't know i guess what are in your guys mind yeah sounds good all right well let's dig in okay cameron i i want to tell i want to start actually this is rare but i want to start by telling you a brief story that relates to something we talked about just before we hit record on the show okay so there's this somewhat new donut shop in town and they make these really exotic donuts. Okay. And I can't remember what the ingredients were, but I mean, it was, you know, it's along the line of, you know, pistachio frosting, you know, with chunks in it. Or there was one where it was like, Ooh, is this like fennel in this donut? And so anyways,

my wife and I are eating them. They're super tasty. But of course, you know, as you're enjoying a really tasty donut, your average person starts thinking about margins and how many of these they have to sell and how much it costs. Based on how many staff are there. I'm just thinking it's like okay these are like real pistachio chunks here like how many of these do they have to sell to like

you know those are not cheap you know are they shelled already like the reason i bring that up is uh you actually solved that exact problem for a real donut shop. So can you tell us a little bit about that story? And then I want to transition to the lessons you took from there into a large multi-billion dollar company. Sure. Yeah. So yeah, when I was back when I was in college, I, at the time, was driving cars for the Bay Driving Experience. And then anyway, the CEO of RPDE, he decided he wanted to start a bakery. I guess really it's probably his wife wanted to start it. I don't know the details. Someone wants a bakery.

Right. For whatever reason, they asked me if I would, if I would basically run it day to day for them. And yeah, that's a 21 year old. You're you're like, yeah, why not? Right. You know, like, are you still able to drive race cars though? Cause like, so the Richard Petter driving experience, so like, because

if someone was like, you know what your day job right now is driving people around and race cars, like going way over the legal speed limit for all the average people. Now, can you figure out how to make a bakery work? Seems like a little change of speed. Yeah, no, that's fair. Yeah. I still drove cars for a while. And I think part of it was I traveled a ton with them. So I'd be in class during the week and on and then at the shop at night. And then we'd fly out to wherever, Iowa, Kansas, Texas, what have you. On a Thursday, you'd work 18 hour day Friday, setting everything up, prepping everything. And then you'd spend about six hours in the car both days. And then you fly home.

Which is not easy in a race car i mean i don't have a ton of experience but like that's pretty brutal yeah not a relaxing drive it's about you know the car's about 140 degrees inside lee you know so you just lose a lot of weight you know it kept me it kept me in shape at the time i thought i had fast metabolism i think it was just that but anyway they were to be honest i was i'd ask them if i could basically stop traveling as much yeah and to be honest i think that might have been part of it another part their brother-in-law who also drove a paddy him and i dj'd weddings on the side had a djing couple we're busy wow i've always had at least like two or three things going on for better or for worse and so anyway they asked me that and you know i went into it with to be honest a lot of naivete you

know i've worked restaurants here and there but obviously never run one much less a bakery i don't know how to bake i don't know it i didn't know any of the things and anyway we created a pretty i thought the problem would be cells and so we ran promotions all these things but cells wasn't the problem yeah there was where we were in harrisburg north carolina or by the racetrack at the time there wasn't a whole lot of similar options. There weren't no Perkins or things like that. So that happened really quick and really organically. And then we, you know, sales kind of plateaued and had more of that organic slow growth.

Yep. The problem was, is we were, we're still losing a crap ton of money. You know, we're talking 15, $20,000 a month losses. Yeah. I mean, for a small bakery, that's... Yeah, that's a problem. That means that you're not going to be baking for much longer.

No, and that was a, you know, that was a problem they kind of put on my shoulders, which at the time seemed right. I mean, since you were running it, were there any conversations where it was like a sit down of like, hey, listen, we're losing 15 to 20 grand. Like, tell us about that. yeah there were so i mean i i i came to them really first on it i'm with i have to believe they were aware of it to some extent because yeah money is leaving accounts yeah right

right but i don't think they were of the the extent or the regularity and then and so there was like look for you know is basically for every you know dollar we were making in revenue it was costing us was it was like about dollar 40 on or so right and i was like it's just this is sustainable and i thought it was going to be like i saw my role at the time as i'm just the essentially front of house manager you know i'm keeping schedules gone i'm keeping the operations gone and they're running the other side but then flipped it on me and i said it was what are you gonna what are you gonna do to fix this all right um you know to be honest i remember going home that night and just kind of racking my brain like how do you how do you go about this and

then uh at the time i took a uh to this route through college i graduated uh undergrad with a two was it like 208 credit hours of the 120 I needed? Overachiever. I would call that, not the best GPA all the way through, but I had three years of mechanical engineering. I had three quarters of the way to a systems engineering degree slash operations research. Most of the math degree and most of the physics degree by what would be my penultimate year and i was like you know i've got

all this information and all these projects we've done like this has to be applicable and you know the first thing i looked at was just okay most companies are about i used what i learned in operations research and manufacturing which isn't going to be true for for the bakery and didn't as well true but i started started off the assumption of most manufacturing companies or product companies, you can assume roughly half of their revenue is going into their direct costs to just make and sell their products. So I was like, okay, if it's taking us $1.40, we're selling for 95 cents per donut. We had other things, but those were our big sellers. I was like, all right, it's got us $1.40, we're selling for 95 cents per donut. We had other things, but those were our big sellers. I was like, all right, it's got to be in those product costs is what I first went to.

So as we talked earlier before the recording started, I was like, all right, how much does it take to make a donut? And I'm sure I annoyed the heck out of our bakers because I followed them around with the stopwatch and clipboard. Yeah. You notice I kept on like, don't rush. It's like, do this. Pretend I'm not here. Normal speed. Right.

Normal speed. I need good data here. And anyway, we did that over about a week period. And I collated the data and averaged everything out and so forth. And when you include your raw materials raw materials your labor time your machine cost which is pretty low in the donut but because it they have to proof for 16 hours you're running a really high humidity high heat box for all night that eats live electricity so include that and we found a fried but otherwise plain donut was 12 and a half cents to us roughly

now like okay cool like we're good 12 and a half cents 95 like and i was like okay well let's just how much wood is it when it's decorated because what we were doing at the time was you know we take whatever we had left the donuts about 10 a.m in the morning and we'd have the hourly staff just go and decorate it yep so that you could get the inventory because the next day you can't sell them because they're dried out yeah you might it's like hey they're already made might as well put them out there right and that way when people come in we have full cases it looks nice looks nice yeah yep and the the first one i did though was our chocolate donut because I had a hunch that it was probably going to be more expensive than we thought because one of the things we'd done instead of using a pre-made chocolate frosting or sauce, we were hand-making one.

And anyway, when you added the chocolate to it, that donut on its own was right around 75 cents now. Just to clarify then, the price was the same between the two donuts? Okay, yeah. Yeah, so all donuts we sold for 95 cents. You see a similar model like Krispy Kremes and Dunkin's. All donuts outside of some specialty are all the same price, right? A glazed donut I think came out to, those weren't too bad. I think they were mid 40s i don't remember exactly because the chocolate was one i'm like all right we're not we brought out chocolate

donuts we're not making decorating any plain donuts to be chocolate unless one right yeah right because we were getting to the end of the day and we were throwing away a couple hundred in these you know and it's easy to do the math on. They're about almost a dollar a piece. Yeah. Burn a couple hundred bucks a day and it adds up, right? I mean, it's something that's a little bit counterintuitive, I think, where it's like, we have these donuts. Let's decorate them. They'll look nice in the case. People like chocolate donuts.

We don't want to throw them away. But there's actually a point here where it's like, actually, we should throw these plain donuts away versus decorate them to be chocolate donuts. And that would not intuitively work in most people's mind. No, and to be honest, it's probably one of the first times in my life that I'd used analytics in a way that did really invalidate an intuitive assumption. Growing up racing, we used analysts all the time and i got really comfortable with you know reading like trace lines and like okay this is where i'm losing time and making appearances and all these things right but those were always pretty close to intuitive it was just figuring out how much do something

or where yeah right yeah right on and in this case, it wasn't. And I mean, I remember I was taken to them like, hey, I'm because I didn't want to make the decision without their input. They tended to come into the shop in the afternoons. Right. And so their view of like having an empty case, they're like, well, you're not running this place well. Right. And like it looks unkempt and all these things right and so i brought them and they you know it took a little

bit to get them convinced but you know we we basically just went through the model yeah like look these are the things are going into what do we take out here what are what is the biggest change and you know one of the problems that is implied here, too, is setting correct PAR. How many donuts do we make per day? Because it takes 18 hours. So we'd have to start the day before. And on days we were closed, we'd have to come in actually on our off days to make them. But I'm like, look, even if forecasting methods at the time, and even now forecasting methods, a good forecasting method,

a mean absolute percent error of plus 25 30 percent's good i was like our shifting cost is so much that we can't just rely on a correct part yeah we need to be more strategic about our whole production process and so that was our first change and it had had obviously a major step change in the business. Was not decorating. And then we followed on with that. I was like, okay, now let me go work on the park. And at the time, not being, let's say, maybe the brightest Googler, I didn't even think about the fact that I could probably find packages for forecasting models so i sat down on a sino notebook and i i wrote out a

a a taylor series method for converging wow that's hardcore yeah that is hardcore oh it's one of those things like looking back i'm like man i'm idiot like if i just it's been so easy but i was like you know i was like well if i set it up this way, I know I can solve this analytically and I don't need to write a bunch of code or whatever. So we started that and we did it based on hourly sales. And that was another sort of not as intuitive thing because we've been approaching everything as daily sales. But when we went and looked at it hourly, we found like on weekdays, it was what? It was like over half our cells were in an hour and a half window of like 6.30 to 9 o'clock. And there'd be a light lull until probably mid-morning, 10.30, 11.

And we'd get some more cells in and then essentially it dies. And there'd be this just slow trickle till we closed at 6 p.m. And they wanted us to stay open at 6 because where we were, where we're located, we get a bunch of car traffic coming from Charlotte. So like, hey, we'll get people on the way home to get desserts or what have you. So doing the hourly forecast, sales forecasting, before we got it accurate enough to go down to product level and fix car, first time I was like, hey, we need to shift our hours. One,

we get like no sales. We don't get enough sales on Monday to cover our labor overhead, much less the product. Right. Yup. I, and at that time I was like, Hey,

if we close Monday, we're going to just lose ourselves. But you can't make money when you're losing it. Yeah. We did that. We closed on mondays and then we also shifted our close time on most weekdays to well it ended up being two o'clock camera went right to two o'clock but we cut that down and you know the other thing where we now have better hourly sell rates we can see all right we need one person like front of house person to come in earlier in the morning. And then we can bring in a second.

Instead of bringing them in at the same time like we were, we can stagger them. And now we were not paying double labor at times when we don't need it. And that opener can leave earlier and the closer obviously closes. And doing that change made up, and it was something like a little over half of our losses.

Yeah. So I'm curious about this. One of my burning questions here that I've always wondered is, do you think companies, so I would imagine a lot of companies determine, do some kind of probably less sophisticated version of that when they first start. Like, all right, we'll be open these days. Like, we don't want to be open. And some of it could just be personal convenience

or whatever. Who knows what the reasons are. But I often have the theory that most people don't reevaluate that. Because say that's what you started with and you're five years in. And there's probably a spot where like, hey, if we opened later here,

earlier here, whatever. Did you end up reevaluating some of these things? So we, we did it once a quarter going forward. One of the things that we thought we'd see would be a lot of seasonality. And we had some, but not, not enough at the time with the length of data that we had to really tease it out. So we'd check it quarterly and it did lead to some shifts. So one of the things we started doing was actually, we started opening later on Saturdays and Sundays at first, and then we ended up shifting Saturday back down to early opening.

Just blowing my mind that people are showing up at the locked door at 6 a.m. on a Saturday. Yeah. But- Review the security camera footage, right? Oh, no, we're already in there. We're already working. You know, you had, if you're opening- Oh, you were, that's right. You're there. Yeah. But. Review the security camera footage, right? Oh, no. We're already in there. We're already working. You know, you had, if you're opening.

Oh, you were. That's right. You're there. Yeah. Yeah. You're there. People like peeking through. Yeah.

That's hilarious. So we reevaluated there. But to your point, I think part of it, and this was the hardest thing to get through to the owners was there's this fear. If I'm not going to be open essentially all the time, what am I going to lose? Yeah. What am I going to miss? Yeah, yeah, yeah.

Totally. Yeah, yeah, yeah. And it really just came around to asking them and, you know, obviously at this point, I can't remember exactly, but the gist of it was, you know, are we optimizing for number of sales? Top line, are we optimizing for profitability? Yep. So this has been number one. for number of sales? Are we optimizing for profitability? So this has been, number one, one thing I love about

it, there's a couple things here, Cameron, that are amazing. So one, these people having you solve this problem using really advanced data technique for a donut shop outside of Charlotte, North Carolina is like wow they like

they absolutely struck gold but now okay let's fast forward so yeah donut shop uh you did all this incredible work to help them optimize it's still running by the way now they end up closing about two and a half years after it left. Okay. That's not surprising. However, they closed with full cases. That's what we know. Full cases of chocolate donuts.

Probably did. Okay. So you would think that all of those problems that you solved would be sort of fully solved problems at a really large company. Later in your career, you went to work for Newell Rubbermaid, which is an enormous... I mean, I don't know how many brands...

I think people are familiar with Rubbermaid and the plastic products, but they own... Yeah, Sharpie. Yeah, I mean, the Max, when I was there after after we merged at Jordan, I made us a $16 billion company. We had 200 something brands. Yeah. And what did you do there? What did you do at Newell Rubbermaid? So I got the, I actually got the position at Newell because of the bakery.

Which is amazing. I was at UNCC's campus where i was in my undergrad and my my now wife then and then fiance was getting her master finishing her master's there in architecture and she went to a job fair so i was like hey you know i'll go i've got a little bit of time i'm good at this and i just started talking to one of the Newell recruiters. And honestly, I can't remember how it started or how we got on the topic. But then she was like, I have to place a call to someone. And she was calling a VP in their strategic sourcing for writing and plastic components.

Got connected and they ended up making me an internship role. And then seven, eight months later, I had to pitch a full-time job to the chief purchasing officer, chief procurement officer. Wow. At a company that big, that's a serious, that's a heavy title. Yeah. Yeah. I'm still not sure why I had to do that part. Maybe an intern pitch to the- If he can do this, he can definitely do the job. Right. Yeah.

I mean, it was basically like, well, what value have you brought us in these seven months? And in that time period, what I had brought over to them was that, that clean sheeting that should cost some money. And you know, just doing a bunch of our different projects that essentially we're looking into for what they call value added value engineering, where you cut costs out products or you redesign them, things like that. And you know, I was happy to say, Hey, you know, you guys paid me $15 an hour in the last seven months.

I've been worth, that year I was worth like one and a quarter million to them. That's called leverage and salary negotiation. Exactly. So I think you should make me a role. And pay me more than 15 an hour and pay me at least 1585 yeah yeah but yeah i got the role there and then that next year i was one of the lead analysts in our plastic components area which is we spent right around a half a billion dollars in plastic components at that time and like on the sort on the supply side

like yeah like like like plastic resins and yeah yeah it's like that oh yeah so actual plastic resins we're about a billion in just the raw resins the plastic components are like oh these are posts okay so these are yeah so we'd be buying from supplier like just this backing for the remote this is one component got it wow things like that and we had a good year in our group new themselves had a kind of rough year so it helped leverage wise as well we beat our goal by like 3x and a lot of it was just tied to making it fact-based what do these products actually cost what you know what our actual patterns, what has been the history. How do we come to a in a lot

of ways, the decks I would end up making for negotiations, I'd get our vendors to say yes, agree to everything the whole way, and then we'd pop out the completed model, and I'm like, cool, so you agree with the whole thing. I'm glad we had this conversation, right?

Yeah. You know, and the kind of I'm glad we had this conversation, right? Yeah. Yeah. And the kind of little side thing to add the time we had, we had McKinsey and as consultants and you know, what they were doing was just taking forever, make spin cubes and then just talking. In my opinion, they were just talking, you know, I was like, well, we think you should do this. And if you say you're going to combine spins like yeah but what are the hard figures on this what

are we going to deliver out of it and uh i can be a bit of a competitive person so at night i was redoing i was doing all the their projects in parallel to basically swallow ego and just show yeah show that you're a glutton for punishment. Yeah. Especially looking back, I'm like, why would I do that? But kind of out of nowhere, one of the junior partners there came friendly with me. We struck up a good conversation and a good working relationship. And yeah, probably about six months into that, got announced that they were going to start a McKinsey and talk the C-suite into starting a center of excellence for analytics. And that they were going to put me as the senior manager for it, which is quite elite from a,

from an, you know, bottom tier of the full analyst. It was like a, it's like seven, seven rungs higher or something like that. Right. We started the department and then hr knocked down a bit but good i was not ready to run an entire department by means but really what it came down to especially once i got a plastic components and started working on we did projects on finished good tools to help negotiate everything from a you know screwdrivers and that to 60 foot tall saw blades to use in manufacturing facilities. Holy crap. 400 piece printers that I would clean sheet every piece and all the products that go into it.

A whole bunch of things. And really what I found is they had all the exact same problems that the bakery had. And I kept thinking, I kept waiting for this like aha moment. Like, oh, this is the magic, right? Right. This is, you know and instead it was the i mean it was the same thing it was bad data it was there's too much for people to go through manually so they'd be like oh let's just look at the top couple and that'll work it

all out you know we'll use the parade or we'll work it all out and like which actually really does seem crazy when you talk about buying half a billion dollars of like finished plastic parts and it's like let's just look at the top couple and it's like that is larger than most companies in the world like your cost for this yeah yeah oh yeah and i mean so like we so one of the negotiation decks that we had from mckinsey they clean sheeted a single pin cap and like we see this much gap on what you're selling to him and this one and we're gonna ask for the whole and i did across the whole thing and when we did across from all the skis over time and probably for using those mark builds i alluded to earlier once you waited for you know weighted those gaps per, you know, annual spend, it was 50% higher than

what McKinsey thought it was. It's just, they chose a bad single one to go on because of how they do it. They weren't able to expand it out. And that's nothing against them at all. They're super smart people there, everything else. Right. But they have a playbook and they follow the playbook. Yeah.

And, but it blew my mind because we'd be in convert literally we'd be in these negotiations across the table and like well you know maybe the suppliers out chins in china and they're like you know labor in guandong has gone up you know 10 every year since 2009 like yeah and like yeah but residence down mountain. And these negotiations would be just these picking back and forth. Really similar to if you're horse trading at a flea market, essentially. Yeah. Sure. Yeah. I bought my first new car that same year and I spent four and a half months collating pricing data across the country on it. And so I narrowed it down to two dealers I thought I could leverage.

And then that day I drove 10 grand off the cost of it. But I came in with a stack of papers and I had them lead me through. I was like, show me where I'm wrong. I do it nicely. We build a relationship. But if I'm doing that for this stupid, I bought a car for $23,500. Why are we doing this for half a billion? I mean, put in perspective, our total spend at Newell was, when I left, was $10 billion.

And that's how almost all of it was running. And that really coalesced to me, okay, I need to take these tools and these methodologies, iterate on them, improve on them, but really abstract them out. Make it into a, basically just a go-to-market strategy for the stuff. You know, same way that nowadays, if you're going to buy a house, you'll look at this estimate. Yeah. That, I mean, a lot of ways, that's what ProcureView is for our clients. So you productize. So you're doing this at Newell Rubbermaid.

So you're taking, I mean, this is a crazy story. From the race car to the donut shop, you take the lessons from the donut shop into Newell Rubbermaid, and then you productize that. And so now you offer that as a software platform that you can run or that your clients run. Yeah. Yeah, we've got 200 users now. You know, we're...

Companies, like customers. End users. So we take a... One of the things that we do do differently than a lot of other players in the SaaS space of this is we have a hybrid approach. So we pair basically consulting, almost like an outsource analytics team with the software. Partly because I'm not a UI designer and a common refrain is it can be overwhelming. And there's a lot in there.

It's very detailed. It's very detailed. It's very granular. Upside being we've now had our outputs, our data audited by the top four accounting firms for a major equity event at one of our clients. It was a one half billion dollar equity event. And they found us to be what was called the single source of truth they made all the decisions and put all based on what our data was right wow um which is pretty funny because you have accounting systems and everything else that are tied to your actual bank accounts that should be yeah all right that wasn't accurate enough it's like that is awesome but also it's huh that's a little yeah oh same i went out i didn't know that was happening i got told after the fact hey did you notice that i'm like i wish i'd known that like yeah yeah you know well john you had some questions like what's happening under the hood i mean

yeah you worked a ton with natural language process if you had a couple questions about like what's happening under the hood what's the magic i? I mean, or not magic, you know, going on the data. Yeah, we talked a little bit before the show. You've got two sides to this, right? So we've been talking about the procurement side. There's also a pricing side. They're very much tied together. Because obviously, you procure for X and you sell for Y if you're in any sort of, you know, business where you're selling goods. So I guess starting with the procurement side, I think it'd be really interesting to walk through some, like for our listeners, data people,

like walk us through some of the tech, maybe even walk us through some iterations. Like we started with this tech, we moved to this tech. I think that'd be really interesting to talk about. Yeah, sure. So I'd say there's probably, there's two separate tech stats that I kind of developed in parallel and then brought them together for this. So the first is going to be related to that, you know, the pricing in your sort of standard analytics off of transactional data. We built a number of proprietary algorithms,

but the two major ones are what we call a purchase index, which is an algorithm that bootstraps a market index for all of your transactions, all your purchases. So you can see real-time, we update monthly, but quote-unquote real-time, basically where is your pricing moving? Just like you've watched the Dow Jones go up and go down. And everything tied to that is going to be on more of your sort of standard,

like I said, standard economic and econometric sort of stack. And I say stack, and I don't actually mean that in our true like infrastructure stack. I mean, more just like concepts, right? And then the other side of it is our, for lack of a better term,

what we call right now our data cleaning side, which is probably the more interesting side when it comes down. It doesn't sound like it, but when it comes down to it for most people, because that's where we leverage a number of internally developed AIs to speed that up.

So we have a foundational embedding network that vectorizes all of our products. Every product we get in, we then use it. Like from an ERP. So like, okay, so you create, so you like would connect to an ERP and then you create a bunch of embeddings in a vector database. Yeah. So we actually don't generally connect directly to the ERP because of security reasons.

I personally against it. That's the number one vector for ransomware attacks into manufacturing companies. I don't believe that. Yeah, that makes total sense. Yeah, as about a $1 million a year company, I don't want to

risk that. So we're talking SFTP is what I'm hearing. SFTP and flat files. And we thought, I can write an API into it. Sure. Please don't make me. Yeah. So we take that.

We embed all the products into a common space. And then for every client, we stand up usually at least five, sometimes more, very rarely less, but usually about five classification and record linkage AIs. Partly, I'm a big fan of the distillation method. So, you know, train up five model or let's say five models you use them as teachers along with like a temperature constraint on the on their outputs for a fourth that's small or things like that love those kind of models because you they're cheap to run once you train them yeah yeah yeah really expressive and then quite a bit on the record linkage side, because again, when you're at these large companies, it becomes really easy to get redundant items or to lose history of items. So by linking together green t-shirt medium with medium green t-shirt or t-shirt green or what have you we can now see are there differences in

their costs or did they change you know any you know choose your own adventure kind of thing so but we leverage those for creating what we call an item master that's where we can put you know cleaned group linkages cleaned descriptions especially if they drop them all off or we have one client that buys a lot from Uline and all their Uline products, all they do is put the item code and not the actual description. So get all that filled in so that we know what the things are. And then we run just ComNet classifiers on it for the categorizations. It's usually three to five levels deep of a hierarchy. And our target for that is within the first month,

we should be at least 94% accurate on unseen data. And our best for a current client is we're at 98.62%. And that's actually in six different languages coming in. Oh, wow. So like, languages coming in. Oh, wow. So like, okay, wow. International, yeah. Yeah. Which actually is such a good use for that type of technology. Oh, yeah.

I can't spend the time doing that. I mean, like I said, we're a small group. Basically for every company we go into, we're going to do the equivalent of about 30 to 50 people's worth of work. Wow. But generally speaking, they don't have those people yet. The companies are going to get the ones who are looking at making this investment. They're like, hey, before you do that.

And I try to tell them to hire people because I'm like, I want it to be SaaS long-term. I want to train you how to fish and then push the boat out. Yeah, for sure. So those are the two main stacks where we probably won a lot of business initially is on the data cleaning side, especially in what's called indirect spend. So it's the stuff that you put on like your purchase cards, your credit cards, flights, travel, all those things. Everyone's like, no one, you can't do anything with it because there's no item codes. We don't know how to put them together.

Like, well, best Western hotel Orlando is probably the same as Orlando Best Western. Yeah, sure. And if it says three nights and the other says four nights, maybe we just say, what's the average per night, right? Yeah. You have a good. It's a widget. We tend to get our foot in there and then we drive it out with product pricing, doing

those market builds. One of the things we pride ourselves, doing those market builds. One of the things we pride ourselves on is our market builds where we're clean-sheeting every single transaction and product category. And some of these categories might have 2,000, 3,000 SKUs. We're on average on any given month within 2.5% of a bid if they went out and bid it right then, or what you could get uh now i can tell you that someone and if someone said to me i wouldn't believe until i saw it so you know when we're

working with our clients we tell them hey validate test it put it out you know what if it's things are wrong and things are wrong and you know first iterations those suppliers are going to give you more information when you tell you. If you tell them, hey, I think a can of Coca-Cola is 90% cost as the serve. It's not, right? It's like 1% of it. That guy's going to tell you, you're an idiot. It's mainly water. Sweet. Yeah. Load that into the algorithm. Yeah, exactly. We've updated and we actually have a live tool where you can do it, not tied to every single item, but as a category level, you can do it live with them, which is some good interest on the convolutional neural networks and then deep learning since 2014.

The current thing around it, I don't think it's all hype. I think there's chunks of it that are hype, but the thing I try to tell all of our clients, everyone we meet with is, we see it as a tool. It's a pointy or stick. Yeah? And if it doesn't deliver significant value quickly on it,

it's not the right tool or it's not ready for you yet or, or, you know, what have you. And yeah, I tell them the proof's in the pudding for it. So anyway, did that,

that answer that question? Yeah. Yeah. That was awesome. That was awesome. John, did you have a question about the pricing side on like selling

because that was something you dealt with a ton which is sort of the inverse right so like because you obviously help people create significant leverage and what their the prices that they're paying to their suppliers but then that is a margin question for my business but I have to turn

around and go sell it right and you guys had a bunch of inventory and bought a bunch yeah so so it's really interesting and i think there's a lot of businesses that still operate this way they will take cost it will often not even be fully landed costs it'll be like cost of goods there'll be some like unknown gap that maybe they like categorizes overhead like all right cost is 30 bucks yeah add five dollars for overhead 35 bucks we'll look at the top couple and we're good yeah and we're gonna mark it up 37 you know yeah like it's surprising how you know how many businesses it's run that way right like they have like oh like oh yeah that seems about right and then they check some

competitors pricing like okay let's tweak it a little bit. That's it. Period. That's it. So there's a lot of sophistication you guys are doing on the procurement side that I think also applies on the pricing side. And there's a big, from my perception, a big gap in the market, especially when we did research on this, there was one firm that was just a little bit too big we're talking like you know quarter million dollars just to engage with them like it didn't make sense was it uh profit no it was a different one okay but anyways so there i think there's this gap in the market and it seems like a lot of this logic would you know translate into pricing So let's talk about that a little bit. Yeah. One, you're right.

I mean, sourcing, I wouldn't say procurement itself, but sourcing is just the other side of the coin of sales. Yeah. Yeah, the salesman's not the one pulling the product off the shelf in a box and shipping it. Same way that your sourcing lead isn't going to be the one placing the actual PO. They're going to be negotiating

what is it for how much and how long is this price good for, right? Eventually, it's this business side. And those costs or those inputs that are identical. And, you know, really what we do when we're doing those market builds and clean sheeting, we're making that pricing model for those suppliers. We build in the margin. We try to build the margin in a way that is fair and sustainable. I'm very big on that. I will fight tooth and nail with my clients on,

I'm not going to make you models, push someone to the bone. I've gotten a call Friday night. Well, Friday night in China. Friday morning, for me, when it happened that they were basically pushing this one supplier too far, and they were putting our half a million dollar of injection molding tools out onto the loading dock. And it's going to rain all weekend. So I go figure it out.

Oh my God. I was like, cool. And I was, at that, I was essentially just an internal consultant. I'm not owning this apartment. They're like, sweet. Now I have to take this up to this director who, you know, and, but that really struck with me because one coming from a small business background, I remember how it felt to get just bullied.

And that's what it is. Yeah, for sure. And finally decided it's not worth it. It get just bullied and that's what, and find that it's not worth it. It's just, it's not. And so we build those models to be fair on both sides of it. But what that really means, we've just made a model that is, that is a, basically extract abstracted version of what takes to produce their items.

And when we've done like, so when we've done audits on it, we aim for within two and a half percent mean absolute percent error on it. And then on some products, especially ones that are high value or closer to like a commodity type stuff, things like stretch film or Corgi boxes, we tend to align within half a percent plus minus of all their actuals. And so to your point for pricing, if we, when we tie it and we're doing this in beta now,

the only step change we're making between doing the procurement side and sell side is now with sell side, we're concluding in the procurement costs as well. Yeah. To try to watch, you know,

what that margin is. And now you're not going to necessarily try to leverage suppliers on your costs and be like, hey, which customers are buying for the right price from us? And who do we want to give discounts to? Because we want to grow with and they're a strategic partner. And some of the things that we found is for one product group, the lowest price by vast margin

was to one of their smallest customers. And it's like- That's not uncommon hey you're making more margin when you're selling to walmart but not to you know whoever and i'm not telling you and i'm like hey go screw over this little guy but like yeah one when walmart finds that out you're gonna be really screwed because you're not making any money over there you're like oh we're giving you the product for cost well a lot of lot of people don't know this. So I think it's Walmart specifically. I'm sure there's others. You sign agreements where it's like, we get the absolute lowest cost. And if they find out, it's a problem. Yeah. I can't say aloud. Yes. Yeah. It is a major problem with that.

But it's one of the things that comes down to, can you parse the data? Can you make sense of it? And then once you're parsing it, we're talking companies with millions of billions of lines of transactions, and sourcing teams are small. Newell, even when we were 16 billion, our entire strategic sourcing, including our execution engineers, everything else on the ground in China, India, so forth, was about 200 people.

Oh, my gosh. Yeah. That's surprising. Right. So you just don't have the time to go through this. So everyone's just using these heuristics. And that's where you're losing on it. So one of the things that we've been crowing about from our marketing side is, and this wasn't something I knew would happen going into it.

It's been a happy output. Our worst customer's return on our cost is about 22 and a half times every year. Wow. They've been gaining on it. And it's not, again, it's not because they're bad, their jobs or anything like that. It's because he just can't make sense of it all.

And what we're really good at is finding those needles in the haystack where you can pop it up and ask them why. Yep. Yeah, that's great. Well, Cameron, I, this time has flown by.

Unfortunately, we're at the buzzer, but I have two, I have two questions for you. Okay. Uh, well, actually one, just to reiterate for the listeners where can they find out more about procure view what's your website just in case anyone's www.procure.com i will say do use the wwws we migrate our website last year and now it doesn't like it without them yeah and that's just for

those keeping score okay two questions one do you ever get out to drive cars really fast anywhere anymore to get that adrenaline rush i've got i've got two sports cars at home had a gr86 and a yamaha r1 powered lotus 7 okay but i actually i had a i had a bad skiing accident in 2014 oh man okay i killed off all my death reception and i lost the ability to to make sentences and a whole bunch of things and yeah i can't get a racing license anymore so i was summarily let go by ferrari north america portion north america it was a rough it happened in january and by march and it kind of filtered out and i was having to sell my race motor i had two race bikes so both the race bikes wow i was just getting like these emails that i'm like thank you for everything but like

see ya yeah that's tough wow so now i just do it virtually yeah okay love that okay and then last question is when you go into a donut shop are you able to order a donut without thinking about the cogs no i hate man no i've had i'm not a very social person in settings. I like to, like, I hate getting my hair cut because I hate when they talk to me. I just say I am. But like, I've been at line and donut shops and someone's complaining. There's a great donut shop in Atlanta, Georgia by the name Spline Donuts.

Okay. You know, giving them a little shout out. They opened in 09. They're amazing, but they're about $3 a donut. And a couple of years ago, we were down in Atlanta and we're in line for them again. So we used to go when my wife went to college at Georgia Tech. And anyway, a person was complaining about it.

And I couldn't hold my tongue. With this person, I'm like- Well, actually, it's 18 hours. I felt like they were tacky like well i love that donut shop like we've known the owner for a long time blah blah but i'm like like you gotta be an idiot like you see that's a broiled marshmallow on top you know how long that takes up like and then i'm realizing like oh my god i'm gonna yeah like it doesn't take ai for you to know that this is a very expensive donut to produce right okay i wrote an entire academic paper on this, right?

Right now. Oh, man. That is so great. Well, Cameron, this has been such an awesome show. So great to learn about your company. And just congrats on an amazing journey. And we wish you all the success in the world with ProcureView. I appreciate it. Thank you guys for having us. And we wish you all the success in the world with ProcureView. I appreciate it.

Thank you guys for having us. Love to catch more of the show. The Data Stack Show is brought to you by Rudderstack, the warehouse-native customer data platform. Rudderstack is purpose-built to help data teams turn customer data into competitive advantage. Learn more at ruddersack.com.