Other than by his flowing mane, Michael Brand can be best defined by his unrelenting drive to keep creating better hardware to solve unique problems. As Dor's co-founder and CEO, Michael told us about his accidental foray into the retail world and everything he's learned there since. To read more about Dor co-founder and CTO Gregg Golembeski, click here.
Dor: What kind of upbringing did you have -- were you always this technical?
Michael: I pretty much grew up in the middle of nowhere on a 20-acre farm. There wasn’t a whole lot to do, so I found solace in going on the Internet. I was rejected by every college I applied to except for Cal Poly even though I had a 4.9 GPA and all kinds of crazy stuff. I pretty much grew up to get to college.
Yeah, but I think if I could go back, Cal Poly would still be my favorite of all the schools I talked to.
They knew better.
Exactly. [laughs] In my engineering experience in college, I wanted to go get my PhD, so I was working towards that. I was majoring in electrical engineering, working on robots, hardware, all that fun stuff, and then I got the gig at Apple with the iPod hardware team. It was probably the best team I ever worked with in my whole life. When I was working with them, I was learning so much, so I just dropped out of school to work with them. I went back for one quarter to finish my bachelor’s degree, but then came back to work with them.
I started working on the iPhone, iPad Pro. About that time, the culture I had joined it for originally had disintegrated and I thought it was time to do something a little bit more wild, so I left, rented a law office that I lived in for a while, started mining bitcoins there, built an EU lab to work on projects, and joined a startup that really needed an engineer. I offered to work for free as long as they could get funding. So I built their first prototypes, their proof of concepts for hardware, then raised a whole bunch of money.
Let me go back and pick apart a few things. How’d you get to Apple?
It was at a job fair in Cal Poly. I wasn’t expecting to get a job, I went there to get free stuff. I was already working for Lockheed Martin at the time. That was a really awesome team as well. When I talked to the people at Apple and they heard that I had software and hardware experience, they really liked it, so I went in for the interview, and it was a good fit. That team really hired a lot more for culture than technical abilities, although they obviously wanted both. It was a whole team of interns -- they usually hire interns for about eight months, then about 50 percent of those people actually get hired into the team. They’re very picky about who joins.
So you enjoy this glorious team, then switch seats, and eventually find that it’s time to move on?
Yeah. At my last gig, after Apple, it was my first experience in the startup world. I left it to everyone else to run the show. I never wanted to be CEO, I only wanted to be technical. Then at the startup I worked for, the opportunity just presented itself to start Dor. It started as a side hustle. Seeing how poorly things were run there, I saw I was in a really good position to make an awesome team and try to reestablish that team culture I had at Apple.
You stepped into the CEO role to create a culture versus becoming CEO because you wanted to run a company just for the sake of it?
"I’ve seen a product take several years and 15 million dollars. For half a million dollars and a year, we got ours to market."
With this new team at Dor, starting scrappy and lean, as Gregg said, and printing new iterations and testing product daily -- had you ever worked that quickly before?
I’ve worked that hard before, but never at that efficiency and speed. You know, how quickly we got hardware to market, the amount of money it took us to get to market? I still haven’t met a startup who’s been able to compare to us. I’ve seen a product take several years and 15 million dollars. For half a million dollars and a year, we got ours to market. I got to bring my skills as a scrappy hustler and combine it with the best talent in the world. It was really cool.
How did you and Gregg connect with Joseph Abad and Michael Lyons at your former gig?
Gregg knew the quality of their work from managing them. Gregg was probably the best software manager I’ve ever met. His team hit all their deadlines. It was this mix where everyone’s working 10-12 hour days, but everyone’s happy because they get to see something at the end of it, and he’s looking out for his team.
When I started having the idea for Dor, I wanted to bring on Gregg as my cofounder. I’d do the hardware side, Gregg would do the software side. Gregg recruited Joseph and Mike. Mike’s an awesome full stack engineer, Joseph’s still the best firmware engineer I’ve ever met -- he made a lot of this happen. I’ve talked to CEOs who have worked on an idea for years and still don’t have a prototype working. They really trust Gregg. They’ll follow him to the edge of the earth, and Gregg won’t lead them off the edge.
That’s something to be incredibly proud of as you founded this company. It wasn’t getting lucky, or that a trust fund was handed to you -- you succeeded because you’re just good at your jobs.
The story is so common of startups breaking laws and lying to employees. I hate that it burns the startup scene for some people.
So you’ve got your unique crack team of the best of the best, and how did you end up meeting Bolt?
Bolt was our first money in. We didn’t pull any members of the team until we had that money. The pitch to Bolt was some ideas on a napkin. Bolt’s interesting in that they like getting in as early as possible. They gave us the first investment check and I hired on the team with two to three months of runway. I spent the next few months networking and pitching investors.
Gregg spoke really highly of the way you and he work together, where you’re the ideas guy and he keeps you grounded.
All of the VPs at our former company were like, “Michael and Gregg are not going to get along, they’re going to be fighting all the time.” Me and Gregg pretty much disagreed about everything but never got into a disagreement and could come to a conclusion. It’s scary -- at this point Gregg and I pretty much know what each other’s thinking and we often flop to the other side and see each other’s point of view really well. I’ll be playing Gregg and he’ll be playing me.
"I was walking around and asking stores if they’d want a door counter. I was thinking there’d be pushback, but when I told them the idea, they were like, 'Holy shit, please do this! We’re hand-counting customers!' It shocked me."
So you’re an engineer. Had you ever worked in retail?
Nope. I knew nothing about retail. It’s actually really weird how that idea came about. It started out having nothing to do with retail. My concept was, I want to own data that no one else has. We don’t have as much data available about the physical world versus the digital world. Dor was just me wanting to know when people go in and out of doorways. I don’t care who they are, it doesn’t matter, I just want to know that one number. I was planning on making this super cheap door counter to just stick in doorways and gather data. Where money hits people walking through doors is retail.
I was in SoMa at the time, and I was walking around and asking stores if they’d want a door counter. I was thinking there’d be pushback, but when I told them the idea, they were like, “Holy shit, please do this! We’re hand-counting customers!” It shocked me, coming from the tech world where everything is analyzed. I was like, “Isn’t there anything on the market for you?” Retailers said they’re too expensive.
That’s when I realized there’s this huge demand from businesses to get what I assumed everyone had. I started looking at the competitive landscape and they cost $2,000-5,000 per door, with heavy monthly fees, insane installation ... no one had attempted to create this product. That’s when I realized this needed to be a retail analytics company, not a data mining company. Then the networking kicks in. I start talking to as many people in retail as I can. So now I would say I know a lot about retail!
Why did you choose thermal-sensing tech instead of other solutions?
Back in the day, we were originally using ultrasonics. It kinda worked. It was a proof of concept and we started developing around that but ran into limitations. Fluffy clothes would muffle sound and not detect a person, echos would reverberate throughout a room. So we pivoted and created an infrared sensor. We basically hacked a TV remote to do infrared detection. Then we deployed those to a couple of our alpha customers, ran into other limitations such as black clothing would absorb all the light or it’d reflect off the floor and think a person was there. It was working well enough, but the accuracy numbers weren’t there.
Gregg’s buddy told us to check out this thermal sensor and we talked to the manufacturer and got a few samples. The data coming out of it looked pretty good but it would take a lot of algorithmic dev to make it useful for us. We started along that dev and it worked, and that’s what Dor is today.
So we worked on ultrasonic for three months, threw it all away, worked on reflective IR, threw it away, worked on thermal and ended up keeping it. I don’t think the team ever thought of it as throwing away work -- it’s all going toward our goal.
And that lean mindset helped with having to keep moving on to better ideas.
How would you explain Dor to an engineer?
The data that comes out of the sensor looks like an audio stream. So we have to tell it that it’s a human vs. a gust of wind. We implemented a machine learning algorithm that was taught with known data. We started out in front of a door and count people going in, and train our algorithm by saying, this was a person in and a person out. We have about 14,000 data points like that that we collected.
How do you make that data live somewhere it makes sense?
That’s really the secret sauce. We developed the sensor in a way that we could upload new firmware to it remotely. The algorithm is stored on our own servers internally run it through our machine learning system and train the algorithm. What it spits out is actually a simplified algorithm that the sensors can run through that follows all the parameters of the learnings. So we can put in new data into our server, then push that out to all of our sensors out in the field. Where it gets cool is that we can pull in data from sensors out in the field, retrain it on our server, then push it back out.
That’s the flexibility of using a cellular network, right?
Yes, the cellular network, and that brings a lot of challenges that other products don’t have. Like, if you have a “smart” vacuum, it probably doesn’t contact the server and reprogram itself. All of our sensors have the ability to reprogram themselves.
For instance, there were several stores having a problem where gusts of wind coming under a crack in a door looked like people. We flagged all those sensors to send us their data at night, and we reprogrammed it to say, “This is not a person,” and then we created a new decision tree and sent it out to all of our sensors in the field. Now that problem doesn’t exist any more. It’s awesome. We built all these fleet management tools -- there are whole companies that just build these IoT management tools -- I would say ours are better, and they’re built from scratch.
”There were several stores having a problem where gusts of wind coming under a crack in a door looked like people. We flagged all those sensors to send us their data at night, and we reprogrammed it to say, ‘This is not a person,’ and then we created a new decision tree and sent it out to all of our sensors in the field. Now that problem doesn’t exist any more.”
Have you had to strike a hard balance between wanting to protect consumer privacy and also having these unique points of data no one else has?
A lot of store owners actually want demographics data, who they are and how to sell to them. The holy grail is, you walk into this store, I know your name and what kind of products you like to purchase, everything. We didn’t see as much value in that as much as staffing and marketing optimization. Having demographics data can lead you to some conclusions, but we see the value for your store coming from the macro data, not the micro scale. Dor’s concerned with the ecosystem around the store. We let store managers deal with the internals, because they’re there all day. We look at weather, staffing, benchmarking against other stores.
So, really, our play is about combining data from everyone’s stores anonymously and using it to help each one individually. As an independent third party, stores are actually happy giving us their data, because they can’t see financial info about their competitors, but we can anonymize it to help both of them grow their businesses with our insights without sacrificing that privacy.
I’m sure that’s an interesting sell to companies, that this anonymous, macro data is going to be more powerful for them that creepy, granular, snag-your-IP-address data.
As long as you have the right reasons to use that data. Right now, store owners maybe don’t have the best motives, when the right reason to get granular would be to provide a better experience for the customer.
It’s the difference between tangible predation versus walking into your favorite store and getting a notification saying, “Welcome back, pick out anything on this shelf for free just for being a valued customer!” Finding that sweet spot is the future.
"Our play is about combining data from everyone’s stores anonymously and using it to help each one individually."
What’s one of your favorite things about Dor?
Well, I’m a hardware guy. At the end of the day, all data comes from the real world, so there has to be some kind of interface that happens. The fact that we got the power usage so low on the sensor is pretty mind-blowing.
Any wonkish side hobbies? What do you do on your down time to stay sane?
Bitcoin is a big hobby. I’m super fascinated by the crypto-currency world -- a bit of the anarchist in me. I love building niche community websites too.