Speaker 1 (00:05.614)
Welcome to the Hard Tech Podcast.

And everybody, welcome back to the Hard Tech podcast. I'm your host Deandre Hericus with my co-host Grant Chapman. And today we have a super exciting guest. I'm super excited about it. He's in the world of sports tech, Alex Krause from TraceUp or Trace. How are you?

How's it going, everybody?

Speaker 1 (00:24.172)
Hey, hey, doing well today.

Awesome. Yeah, so Alex, you've got a really unique background. mean, you live in California, if that's right, and you've worked at some of the largest companies in the world on the products that we use every single day. And then you've made the transition over to something that not everyone uses every day, but it's super cool in the world of sports tech. So we'd love to get an update and a background just who you are.

Yeah, yeah, yeah. you mentioned it, you know, I actually got, I did the whole PhD thing at Caltech and then I worked at Apple for three years on the iPhone. So we were working on the Face ID hardware, the thing that actually sends out 30,000 laser spots, you know, to make a 3D map and do Face ID unlock. Ramped that hundreds of millions of units for, you know, a few years. I did a brief stint making Elon's boring company.

They're doing tunneling equipment, totally different.

Yeah, a bigger scale and different hardware, right? Like instead of going from micron size, you know, worrying about tolerances, things like that, now you're at the macro scale.

Speaker 1 (01:26.242)
Yeah, exactly. Doing 300 ton crane lifts, you know, and making sure those are handled safely. Totally different. But we're only building one a year.

Yeah, yeah quantity and physical size of unit are totally different on those two they have platforms

Yeah. Yeah. And then, and that was great. Still love that company. And then now, you know, at Trace for almost three years now, it's been pretty nice. We're a small startup doing these soccer AI cameras. it's a soccer camera. You chuck it up on a tripod, two cameras record 180 degree field of view in high resolution. And then an AI backend basically processes all that footage.

recognizes all the players so that you get not just highlights or just like the game footage, every player gets their own highlight reel.

That's so neat. So you can basically like a coach can have it for a team not just like one dad for their kid This is like coach can do it for the team and after it's all been post-processed each team member can see their plays each member can see their team highlights and kind like the mix and match thereof

Speaker 1 (02:33.494)
Yeah, exactly. we actually find because we record the whole match and we will send you the video that's basically following the action. We can post process and send you a version of the game that's that's just you as if you had hired your own personal videographer where you're the center of the action. Right. All game, you know, or for all the highlights.

That's so crazy that you can get that automation to that scale that you don't have a human in the loop to do that.

Exactly. Yeah. So everyone gets basically their own cam version.

Yeah, and all with two cameras of hardware.

Yeah, exactly.

Speaker 3 (03:07.374)
What drew you to Trace? So how did you end up at Trace three years ago?

Yeah, I mean, love the chance. Obviously, I'd gone from Apple, where you're on these huge teams, to, you know, at Elon, you need lot more ownership. And I really liked that. So like at Trace, have the director of hardware here. And I like that I kind of make most of the hardware decisions, obviously, in partnership with the product team, you know, the CEO, like it's, it's always still a group effort, you know, to build anything good. But I like, you know, owning a lot of that stack.

and really being involved in the whole stack of the hardware piece.

That's awesome. I think that's why a lot of people work at Glassboard or places like Glassboard because we get to do that on more than one project at once, right? You know, our engineers that are leading hardware development for company A and B at the same time and you know, they're kind of riding that roller coaster all the time.

Yeah, yeah, exactly. New devices, new hardware design.

Speaker 2 (04:02.67)
always the thickness they say you know there's no rest for the weekend

Yeah.

And Alex when it comes to like the actual tech stack that you're manning so can you gotta walk us through it with what you can like the actual tech stack of these cameras just because I mean I think back to my time I started in the sports tech world back in 2020 and around that time you get started there's like I was in the team management side as a software service platform I played football in college and huddle

Spidey trace, yes, know this is the time like people are raising 50 million 60 million dollar rounds all in the world of cameras because I think huddle started it out and that was sort of a moat that you could create to make sure that a school system just couldn't download another software for their teams now you have an actual physical camera which Sort of was interesting to the school market So it's kind of curious on the tech stack and and I got to see from afar that how that was changing But never actually have conversations who were actually building those cameras

Yeah, yeah, so huddles, yeah, one of our competitors, Veo is another one you may have heard of in the, you know, in the soccer space. And one thing that's most different about Trace, you know, at the hardware organization level is we actually don't sell the cameras. So, you know, Veo, huddle, the competitors, we're actually hardware as a service. We're leasing the camera to you. One thing, like as someone who runs the hardware team, I love about that is I actually

Speaker 1 (05:28.52)
own all of the hardware. So if I need to push a software update, we just push a software update. It's my hardware. Right. You know, I'm not waiting for the user input. And in a lot of ways, it's a more honest model because yes, you can buy a HUTL camera or buy a competitor's camera. But if you're not paying for the service, you're actually purchasing the annual license. It's a paperweight. You can't actually do anything with it. You don't actually own something of

of value because it doesn't work by itself. And so yeah, we like this model that we lease it. also means, you know, I don't think of it as how can I sell more cameras, right? We're trying to enroll customers, keep them active. And then it's more like, you know, Hertz, Renaissance, you know, I'm maintaining a fleet of hardware. I need to keep it refurbished. I need to, you know, check health metrics while it's deployed in the field, things like that.

And so when was that hardware as a service business model decided on? Was that decided on from the beginning? Was that always the competitive difference between?

from the beginning. Yeah, yeah, even before before I joined it, you know, so I can't take credit for it, even though I really like it. I think it's wise. Because, again, because we're not really selling you something because it doesn't even work without the service on top of it. And so it makes sense to just just loop it in kind of into this this recurring revenue model. You know, it's also nice because when customers churn, we get the hardware back.

Yeah, it's not a lot of people rather a little loss

Speaker 1 (07:04.366)
Exactly, exactly. We get the hardware back and you know, sometimes you're chasing customers hard, most of the time we get that hardware back and I can, we have a rigorous QA testing procedure to make sure it's good to go back out in the field. But we send it out again.

That's awesome again the the place where I wish someone would have used hard was a service I don't have it was just a great case study. Have you ever heard of the Spotify car thing? Because they built up bespoke piece of hardware that was Spotify in your car is plug in your ox core Bluetooth No phone needed and then they charged you for it and they dropped support like within a year or some crazy short time All these angry customers that bought this hardware at you know an expensive price point and yeah

It's for the front.

Yeah, yeah, yeah crazy case study and like customer rights consumer rights of like how is this fair or legal or whatnot?

Yeah, yeah, yeah, because you don't really own it if they can be bricked. Yeah, they can really rely on it.

Speaker 2 (08:02.574)
Yeah, whereas you guys don't have that fallacy up front that I own this camera like no no I am paying trace for this service and I love the service and I'm a customer or the service wasn't for me and I shipped the hardware back and it was never mind to own

Exactly. Yeah. And to talk about the actual hardware stack, how we built it and how we thought of it as well, the main brains, we kept it very simple, which is how we built it is very modular and easy to set up rather than a lot of custom dev boards and things like that. So the main brain is actually a Raspberry Pi, just an off the shelf Raspberry Pi 4, kind of runs the system.

we have our programming application on that, right? The Pi 4, obviously it's an incredible platform. It's so cheap. You know, it has Wi-Fi certifications, you know, it's pretty stable, all of these things, just an incredible platform. So we just built it off the Pi 4. And then the camera modules themselves are also essentially not actually GoPro, but essentially just separate camera modules. And you you tell them to start recording, they record, and then...

They're insane.

Speaker 1 (09:11.214)
After we record the footage, download that, the Pi downloads it and we upload it to S3. And then all of the processing is basically in the cloud.

That's interesting so your cameras have their own storage like this isn't like the pie is buffering the footage and logging it to an internal memory like those camera modules are doing that as a you more or less like off-the-shelf product you guys are just Interfacing with that cameras digital interface to go grab that data and use the pie as the shuttling agent to get that data up That's right servers. that's so fascinating

Exactly. And so it lets us get started. know, before, you know, I think next year we're going to be in you know, tens of thousands.

Just gonna have to you can share volumes, yeah. But this is a point of like, where is that cliff of like, pies get hard to source of that volume and not, like, one of those choices is gonna happen.

Yeah

Speaker 1 (10:02.232)
Yes. Exactly. So before we were at that volume, this modular architecture was great because it meant it was pretty easy when we run out of one camera sensor was starting to get low. I just swapped in a whole new modular unit because they're just communicating over a standard interface. Is it USB? I didn't have to redo a ton of firmware. It's just over USB.

So yeah, you're literally writing a Linux USB driver to go dial in on that camera, say, I'd like the footage now, or hey, I'd like you to start recording now, and that's all you're doing? Yes.

Exactly. And actually our partner who sells us the camera modules, they actually wrote the basic interface, like here's the API call over USB to start recording, to get files, to whatever, right? That's so great. And so it means on our side, especially as we're hardware as a service, I'm maintaining this fleet. Backwards compatibility is critical so that if I get a return from a customer two years ago, I can

just upgrade and put in the latest camera modules. So our entire fleet, even over lot of hardware revs, is actually fully intercompatible.

So you have like a bunch of different products that can all differently combinationally get into a trench coat together and be a trace unit. Yes. That way your supply chain other than the pie is incredibly robust. And when you guys launched the hardware the first time because I pie is right after the pandemic where

Speaker 1 (11:28.622)
Great question. was. Yeah, yeah, yeah. was it in it would have been late 2021. Yeah, it was was first launching. That's right. So right.

don't know if we're going back online.

And we went through some tough times with the pie. We made the decision on the pie. was like, it's awesome, super available. And in COVID, we were paying wild prices, two, three X, the nominal price of a pie, just to make sure we had them.

Were you buying the Amazon kits that had like the case and the memory card and everything included with it? We had to do that for you.

Anything we could get our hands on. any of you get our hands on buying on Amazon, finding random sellers, you know, we had a good distributor partner who sourced some but yeah, yeah, the price is just, I think at their peak went up to around 3x the normal price of a pie if you really needed them.

Speaker 2 (12:18.242)
But the beauty of your model is you didn't have to make margin on that piece of hardware that day. As long as that Pi lived long enough as a hardware service flight, it paid itself off and you grew the customer base and grew your fleet and that's the thing you can keep investing in. And the Pis are the robust, whereas the cameras are like the thing you're probably gonna wanna upgrade.

Yeah, yeah, because the pies honestly almost never break. Like if things break in the camera, because what's amazing is, you know, we do have some cameras that break and get damaged. have higher failure rates than I personally would like. I'm always trying to get it better. But one of the reasons they break is it's on a 15 foot tripod. You'll see it in video. It gets hit by a soccer ball. Someone runs into it. It just windy and get, you know, and fall over.

15 feet drop is really hard to survive for anything.

Exactly, exactly. the iPhone, yeah, you know, a lot of iPhones wouldn't survive that, especially repeatedly. Like I can see in the returns evidence that like, this one survived, but it hit the ground. You can see some internal comments are busted or yeah, yeah. So there's that. Yeah, yeah, exactly.

There's a lot of fun.

Speaker 2 (13:28.214)
so much fun. the great part is if the Pi survives, you have that hardware, right? You have all these disparate parts that you might have a 40 % broken unit. That means you still have 60 % of the cogs of the next unit.

And Alex, in addition to that, whenever we were first chatting in our pre-podcast conversation, you were talking a little bit about the Nvidia chips and the trade-offs you had to make between keeping the pie or the Nvidia chips. what does it look like for you as balancing the bomb, plus 10,000 units going into next year? How are you sort of thinking about that?

Yeah, it's the classic decision. think hitting a lot of hardware makers, AI people focused on this is like, should I try to move more compute to the edge? Is like the key question. Should I move to the edge? And this was a few years ago, we were working on it and designing actually, I have a functional unit of like a version with an NVIDIA inside that would have done a lot of AI compute at the edge.

So hard.

Speaker 1 (14:30.07)
It's not clear if we could have done all of our AI compute, like you still need something in cloud, but a pretty powerful, awesome edge unit. And we looked into it and one of the problems was, especially those Nvidia units at the time, it would have doubled at least our bomb. So it would have been, let's say, a thousand dollar camera for our cost.

And to dig into that, it's one of those things that you're doubling the bomb. And the question in a product world is, what does that do for the user? There are certain cases where having real-time edge compute is worth money. It helps the use case. for you guys, whether I get it right after the game or on the drive home, it's not as important. And tertiary to that, is also the, often is that hardware computing? If you have a case where your edge computers use 70 or 80 % of the time, it probably makes sense to edge compute it.

But if these get used every weekend and five days a week they're not computing, it's way easier to rent that time on a server somewhere.

Yeah, exactly. we kind of went through this exact thing of like, does it make sense to basically, you're paying upfront, basically all your capex, you're basically purchasing a huge amount of compute and just putting it on the edge. And you when does it make sense? If you need really low latency, it makes sense. Right. So in this case, if we needed to recognize the goal, wanted to recognize the goal. for scorekeeping. Right now, essentially.

you know, for scorekeeping, for things like that, If you do those features where you went live, essentially, then you need on the edge, you know, and it's just that simple.

Speaker 2 (16:12.906)
Broadcast as well.

Yeah, exactly. So then you say, okay, well, I want to live stream the game. And actually we do provide live stream now. We rolled it out recently with our current hardware, which was really cool. That basically in software with firmware, we figured it out that a hundred percent of our cameras for free just overnight got live streaming capability. And the trade off is the live stream is simple in that it's just live streaming the raw feed from the cameras.

Very cool.

Speaker 1 (16:45.752)
But we're not, because we don't have the built-in AI to follow the action. But, you know, humans are pretty good at that. Like if you're watching a live stream, you don't mind that much the cameras. there are benefits. Like I'd love if we could zoom in, use the bandwidth better. Yeah. Yeah, exactly. But we realized that we didn't need to basically massively increase the bomb cost to deliver the core features for the customer.

which is highlights, game analysis, the match video, and actually that we had a path to, for zero dollars of hardware cost, deliver a live stream. Yes, it's not following the action as well as you'd like, but it is live stream. People use it. People use it every

Well and the question that begs me is can you guys pull another rabbit out of the proverbial hat and Just whatever you were going to compute on the edge go put on a server and yeah It might have a two minute delay or 30 second delay That's another soft change that the existing hardware if we can stream to the cloud The AI can get it in the cloud then go rip it apart and track the action and do the highlights and like you could I mean Generally with enough compute horsepower and enough you know time to make the code you might be able to make an individualized stream per player

Yes, exactly. I, you know, try to think about it. I made a presentation once when we were making this decision about, you know, what should we put on the edge? Should we go this way for like a high end edge computer or just keep kind of how we're doing, which we decided. And like the first title slide was like, when should you do hardware? And the answer is like, don't. The answer is don't do hardware. As a first step, if you can find a way

not to, don't use the hardware. Don't rely on a hardware solution because you're baking in all of this cost, all of these assumptions, et cetera. If you want to improve something, you don't roll a firmware update. You build a new one and it's like slowly. And so obviously I love hardware. I've been building it for years and years in lot of places, but you should be really intentional about.

Speaker 1 (19:02.976)
Do you absolutely need a hardware solution? Is that really the best option here? And the other thing about putting compute at the edge versus in the cloud. If the compute's at the edge at the time I record the video, let's say my AI models, I don't know, had an issue. It picked the wrong soccer ball, whatever. Well, I've already processed it. I've basically lost the original video. I can't reprocess it. Or if I want to release a new model,

I have to wait until the customer turns on their camera, uploads it. Hopefully they turn on their camera and update before their next soccer game. Whereas like, if I need a hot fix in the cloud, I just push it. And I can even like retroactively say to the last 12 out. Yeah, exactly. The last two days of games just reprocess. You have all this flexibility. Yeah.

I thought that was a super good point there on if you don't have to do hardware, don't do it. And I love that because this is the hard tech podcast. This is literally what we talk about. But I think it's a great point. with your background, the companies you've been at, what is that criteria for you? With the next couple slides in the deck, did you explain to your team, don't do hardware unless you ultimately have to. And this is basically the decision making criteria as to how we decide if hardware is necessary.

Yes, yes, exactly. It's, you know, in a nutshell, like when you need hardware is because you run out of all the other options. it's, it's the necessary evil. and the trade-off is, obviously like the biggest companies in the world build hardware and do hardware and use it. And there's a couple of reasons for that. One is it enables you to do things you can't do otherwise. It, but it's also very sticky, which is like, we, could have trace.

Yeah, it's the necessary evil.

Speaker 1 (20:45.87)
talked about and could have gone down the avenue of like, trying to just release an app for your iPhone. Right, that records it. But of course, that's always really hard. You have to rely on the customer to give up their iPhone, place it correctly. You don't own any of stack, all those things. So the real point is consider all your other options first and then do it when you have to.

Yeah, and we always take a look at that with our, know, we're looking at any new product for clients. It's like, all right, what off the shelf can I go steal and put into your product before having to develop from the customer ground up? And we go find a, know, it's called an IoT gateway and we can just plug in an off the shelf sensor and call it their product. Like, here's a 3D printed box, go buy these three things and put them together.

and launch with this and only when your volumes cross the point that buying these gets expensive and you can ROI my engineering time to build that into a custom solution, should you pay me to do that, but I'm here to help you crawl, run. Right? I think this is thing that's hard. We are that necessary evil, just like talking to me in a meeting, you have to to get to the hard tech team, but here we are. But no, it's that tough decision of how much and when. And it's really good, I'm going to call it diligence, to keep the ooh nice feature at bat.

That's right.

Speaker 2 (22:03.406)
Get you to crawl walk around and not get the you know the leadership that's like oh, no We want everything today like can you afford everything today does you know? Yeah, a hundred clients or five hundred clients today need everything

And it's not even, you know, just it's like, can you afford everything today? You might not have had the money or the team or all the investment needed to, do a custom design in 10,000 units. The other, the other thing is to kind of realize how little, know, like before you enter a new space or new market, new product, you actually don't know that well exactly what all the key features are going to be, how it's going to work, how you want to do it. And so there's so much value. If you can cobble together.

buy an ESP32, 3D print your own box, click in a few USB, Bluetooth, Wi-Fi connector, right? Do that version of your product. You can definitely sell the first hundreds, the first thousand of your units, and you're learn so much from that. But then when you go to do the real design, all the money aside, just you learn so much. Your product is gonna be so much better.

And you will have gotten to market sooner, like almost certainly, with the cobbled together thing that isn't as polished as you would have wanted.

You get the validation and everything like that.

Speaker 2 (23:24.846)
I couldn't agree more. It's just this thing that the hardware guys need to steal from the software team. You know, all the app developers and software startups, you know what their first version is? Because it's not a website or an app. It's like an MS Paint, PowerPoint, or maybe a Figma version of their website. They go show their first five friends or 10 friends. And then the very first version of app, they code it together. And today you just go vibe code it. But back in the day, you'd go build on a platform and put your logo on it and fake it. And in the background, even if you didn't have the right API calls or stuff,

you would literally have humans go do the thing. Like someone would upload the video and you go fake edit it. And they'd hey, did you like the service? And then you have to figure out how to automate that thing that you used to do with human scale. And think doing that with hardware is such a great parallel to say, like, what can you fake until you make?

Yeah, yeah. the real thing about doing it custom, unless, I don't know, weight or exact minimum size is a requirement for your product, you can almost always get a bunch of cobbled together off the shelf stuff. It's going to be a little bit bulky or a little bit heavy. And then it's really just a question of cost. So as Trace scales, we're getting up into those 10k units, things like that, is where I think it does start making sense.

It's worth the custom design and the custom design for all these things. But even that, it's only, you know, just because I think it'll be cheaper. It'll be worth it. It'll save at that scale.

Yeah, it's the interesting part is what do you customize first? Because it might not be that you you're not going to build your own camera sensors You're go buy camera sensors. Yeah, and you're gonna replace the pie But do you use a CM compute module instead of a full raspberry pi 4? Right like there are crawl walk approaches even in the middle of it's still custom But which shoulders of Giants do you stand on because all of our clients on IOT is like do you go use a particle?

Speaker 2 (25:17.964)
right, all-in-one wrapped up solution that gets your backend and your hardware, or do you go use a module that's been pre-approved, or do you go chip in, right? these are the trade-offs you need to make, volume scale, risk, you know, and all those stuff.

Yeah, yeah. I, you know, to evaluate that, kind of put together, we're doing a redesign, you know, some of the internals of the TraceCam, basically to improve it. And one thing that's nice, the way I'm always doing various hardware revs, but of course, externally, it doesn't look like it. You can't see it looks like the same, you know, TraceCam, we're always like improving the internals for reliability and things like that. And now we're doing a big rev and, you know, we looked at

I basically considered over the first year, amortizing all of our NRE expense. And then like thinking of that as the, what the cost of the module would be. And you can get, I think a little too cute with finance terms. You say, well that's NRE, but we're going to do this for years and years and et cetera. And I think if it doesn't like obviously pencil out in the first year,

It's probably not worth going to really hardcore solution. It depends what you're doing, like certainly at least consumer tech. If it's not obvious, you know, don't wait on this three to five year, know, some really long amortization window with a discount rate.

You still have to make money. That's my favorite thing. was like, no, go pretend you to get a loan for your NRE and what the interest rate is. And what is the efficient payback schedule for your product's lifetime? Some med devices are 10 year lifetime from a design perspective. like, you can really afford to do big NRE and get out of your skis because you could afford the finance charge if you had to externally finance. Yeah. I love the way you're approaching that. Like, no, if you're going to sell finances, it has to be like a one year ROI because the owner is either not going to make money for a year and bet it all on farm the next two. Right.

Speaker 1 (27:13.954)
Yeah, Yeah, exactly.

I love a couple of the points you guys made a moment ago. I remember whenever we were running all in.

back in the day. And I was just in my college dorm room, and we were on Twitter. My co-founder and I and all the coaches, I'm sure you're pretty aware, they do a lot of recruiting on Twitter. And so we just jumped on there. We started shooting a lot of DMs, because that's how we're selling. We don't know anything. We're in college. And we're just getting meetings. And one of my co-founders was a graphic designer. he had just created a mock-up. Basically, as long as I was going through it, I could walk a coach through a demo of our product, even though obviously there was no line code. It was just Adobe XD, and we programmed it.

And we got our LOIs, because there was an investor who like, we're interested in your company, you guys are clearly taking this seriously, is anyone going to pay for your product? And so, was like, we have to go prove that. We got LOIs, we ended up raising capital, then the rest is history kind of thing. I think with hardware, I think in general, B2C consumer SaaS is really, really difficult. Arguably, there's probably not

That's awesome.

Speaker 3 (28:18.99)
50 really, really successful, truly B2C consumer companies. really think it's just a business model perspective. The hardware does make it sticky, which is unique. Because really, you guys are dealing with prosumers. It's not directly consumer. These are prosumers. People are paying for their child to go through travel soccer or what have you. I think it's just cool how, one, I think the business model is really awesome. But two, just your approach on the hardware development. And your guys' team is what?

yourself and then how many others are on the hardware team like

Yeah, on the hardware team, it's me and we have one firmware engineer, John, who's excellent. He's been with the company for 10 years now. But that's it right now. And then we have people on the hardware side doing refurbishment, managing the fleet. We have a shipping team here in California. But true hardware engineering, that's it. And then we've got our contract manufacturer overseas in Taiwan, a really close partner.

who helps us with a lot of things. But that's the other amazing thing, right? have well over 10,000 units in the field, shipping thousands of units managed by nominally me and John.

So incredible. mean, that is such a like rock star success story of hard hard tech company. Right. You you knew when to stand on the shoulders of giants to get the hardware done. Necessary evil to sell what I'm sure your software teams probably bicker and is what your customers you know what they appreciate over the hardware. that that back to Andre's point that stickiness of buying anything and getting a physical product like retail therapy. There's very few software that a retail therapy.

Speaker 2 (30:03.148)
But man, when you get a new shiny toy and you go set it up, even if you have to interact with the software to make it work, that is the core new Christmas present, right? This is the new toy.

Yeah, it's still fun. It's still great. And it's also nice. Just have a dedicated thing. So one thing I love about our camera and how it was designed is, yes, you can connect with your phone, whatever, send the game ID, information, whatever. But you don't have to. If it's not connected to power internet, if you push the back button to turn it on, it starts recording. That's it. You turn it on, it starts recording. It's a camera. That's what it should do.

That's so good.

Speaker 1 (30:42.974)
There's a single button if you can't connect to it on the field. It's recording. Don't worry about it No, and then you get at home you if it's plugged into power and internet it goes in upload mode It knows to do uploads the video But yeah, there's no there's no setup like and I love how Simple it was really the product team drove a lot of that how simple that concept was of like why are you turning on the camera? You know, if it's not plugged into basically power and you know connect to the internet you're recording a game just record

that.

Speaker 2 (31:13.614)
And I think it's such a great thing to say about UX, user experience design. Like, hey, we could put every bell and whistle on this and make sure that you need to dial in and set the focal length on the app and do all the stuff. yeah, let's auto-magical most of it. Sure, the settings might not be perfect, perfect as if a pro photographer set it up, but it's what everyone wants and it will work every time. And that's the magic. People don't use things that work 80 % of the time. There's this like magic moment that if there's too much friction, you just stop doing it. Right? In any experience.

down to our unit type, it was all time and sales. Like find the friction, because there's a certain amount, although it someone not want to talk to one of our clients if we're helping them with go to market. Or for us, like how do we engineer, you know, conversations that don't feel like there's friction in them to have an open and honest conversation about like what everyone's doing and where you're at help. And it's like removing that friction both from what we do as a service, what you guys do in a hardware piece, and what makes magic on Christmas morning. You can open up the new toy and it works. There's not an hour of downloading the right things and setting it all up right.

Set up and waiting exactly. Yeah. Yeah, it just works is the goal

That's incredible. So my next question is, what's the most exciting thing in the forefront for you guys? Like what's, what is the new shiny object inside that you're able or willing to talk about today? You know, is this that, you know, internet AI has gotten better and faster or where are you guys headed that's exciting?

Yeah, so I'd be remiss to not share. So Trace, obviously we run these soccer cameras, but also there's Trace Vision now, which is taking all of the same AI components where we built an AI that's able to track a kids youth soccer match where every single person is usually the same gender, about the same age, everyone on the same team, they're all wearing the same clothing.

Speaker 1 (32:59.79)
and they're running at each other and colliding. And we built the AI that can do a pretty good job of figuring out who's who, tracking through those occlusions, keeping track of all these things. And we realized, we can take the same core and apply it to AI in retail and business cases. very cool. And so we actually have an active product now in, I think, a few...

I don't want get over my screds, but a few hundred stores or something, actively recording and doing analytics and video highlights, but in that domain, basically tracking customers through a store.

Did they look at that advertisement is this end cap enticing people who who spends the most time on what and and are you are you actually delivering ROI for your People that are paying for advertisements I used to work at a Best Buy and I worked in merchandise where I set up all of the end of line You know end of aisle ads and all that and it's always like man I know how much these people paid for this I hope people are buying this product more than they would if it was just on the shelf to two steps down. Yes

And what, and I'll get to the hardware for that, but one thing that's amazing, we've been doing a lot of is our goal isn't to build a dashboard of just like five people came to the store. The goal is to do like what we do in Trace, to deliver a highlight reel, which is like, so you can put, you draw a sensor, it's called in the field, like a digital sensor where filming everyone who crosses this line. And then there's a playlist and app you can look at and it just click, click, click just

You get to see every single person that crossed that line over, a day or whatever you want. And it's such a wonderful highlight reel. We showed it to, we demoed it to someone and he was like, wow, this is awesome. Because he can see there's a difference between they walked by it and like glanced at it, or they were just walking to get somewhere else. It's really hard to put that into a dashboard, but humans love watching video. It's really engaging. Short form videos are great.

Speaker 1 (35:05.07)
And so like, we want to deliver the highlights of that, the highlights that you care about. And so he said like, oh, I love it. Like, this is clear, okay, how we need to change this end cap. Because it was so easy for him to just watch what people do and how they interact with it.

I'm even thinking like exactly from a user experience perspective before he would have had to watch all the video Maybe he would have like saw a few people walk, but it would not compute Yeah, like back to back to back creating highlight for like retail is fantastic. I love it. Yeah

Yeah.

And so, for us, we're looking in the hardware space, what makes sense here to help us. was again, I'd say we were reluctant, which is we were like, everybody already has cameras, we don't need to do it. But now we're looking, well, all those cameras aren't great, or they're really expensive, or it would be nice to have, for customers who don't already have cameras, who want to improve it, to be able to ride them to that. Because a of times the customer comes to you and wants that.

And so I think that's down the line, kind of the next place we want to start looking. we're going to build, ideally, something that it's almost the same brains that's in the soccer camera is now utilized in the retail space. Probably a different housing, yeah, lot of the same, reuse a lot of the same components.

Speaker 2 (36:19.596)
Yep.

Speaker 3 (36:24.334)
You don't think it'll be nice and maybe blue hanging from the ceiling?

Yeah, yeah, yeah. I've thought about it. If the product team will let me, I would make it the same beautiful bright teal trace colors, and I would put nice addressable LED lighting on it. I would go all the way, but they probably won't let me.

It's like literally the opposite of every other white and blend in with the rest of the ceiling Well the last question I've got for you Alex is around the tariffs going on with the company your size and Even like the brief conversation we had before we kicked off the podcast This is kind of really where it's impacting. So I'm curious like what's been your experience so far and what's happened?

I want it to be white.

Yeah.

Speaker 1 (36:54.85)
Yeah, yeah, yeah.

Speaker 1 (37:14.414)
Yeah, so obviously, everyone in hardware space, know, day by day, I'm, you know, checking things. The biggest way it impacted us, of course, was, you know, we import from Taiwan, currently, is the key region, and we rushed and got our vendor to produce almost all of our inventory for the rest of the year before the... I think July 9th was the nominal, the first...

90 day pause that was going to end on July 9th. And we did it. You know, it was awesome in that our partner was able to step up. We did some things to improve our capacity, you know, bring in supplies. I think a lot of people were doing that. You know, the the trade-off though now on the hardware side that sucked about that is now it was locked. There were some improvements I wanted to make to the hardware for this season. Other investigations I wanted to do that it was like, now we don't have time.

you know, just ship what we've been shipping. And then it also meant making a huge investment in CapEx. know, cash flow matters a lot in this business. And so it's a total change to have to have, you know, pre-ordered so much stuff ahead of time.

Yeah, even young with a traditional like I'll call like hard product vendor that sits on inventory and sells it You don't even get to cash out the inventory sell it has some amount of time or it's ROI. Yes service So it's like an even longer play and going back that finance charge. We talked about like the NRE side Yeah, this same calculation comes in and this is for anyone listening trying to start a company. It's all about cash flow It's not an engineering or software or marketing if you can fix cash flow everything else goes into place because you have money to fix the problems, but

If cash will ever go to zero, bad days happen.

Speaker 1 (39:03.758)
Yeah, yeah, exactly. You got to the balance sheet. And you're in a fundamental way, like how much cash you have to put up front, how early you have to put it up front, limits your growth rate, because you'll run out of cash, essentially. So think about that. going forward with the tariffs, it does mean we're looking closely at our supply chain and making backup plans. What would it take?

if we did need to rip and replace and move these components. What components can actually be sourced here and maybe final assemblies here, but other key components are still external. And I'd say at this point, it's really, in the, a ton of work to look at all of these various options right now, because we're just out of the phase that was panic, just get everything built before the terrorists might hit.

And you bought yourself an- Yeah, just an inventory runway to make a plan.

Yes, exactly, to make the plan. And that's where we are now.

Everybody, Alex Krause from Trace. This is the Hard Tech Podcast. I'm your host, Deandre Hericus, my co-host Grant Chapman. Thank you so much for tuning in. I look forward to seeing you guys next week.