Change Leader Full Interview: Automated Information Collection and Analysis On Job Sites
This page profiles innovative and impactful applied research and development in civil and structural engineering to spur continuing thought and dialog to create a better industry. These profiles are based on interviews, and the opinions and statements are those of the subject and are not necessarily shared or endorsed by this publication.
Rory San Miguel is the co-founder and CEO of Propeller, which creates tools and software for construction companies, mines, quarries and landfills to collect, process and visualize accurate survey data.
V1 Media: Please provide a brief background of your education and career before Propeller.
San Miguel: I studied mechatronic engineering, which is a combined mechanical, electrical and computer engineering degree in Sydney. That’s what typically leads you down a robotics path. I ended up starting this robotics club, where we were running a team of about 20 people selling small sensors and LEDs and other parts to the students on campus. That turned into my drone passion.
We made all this money at this university club, but couldn’t pay ourselves a salary, so I decided, “let’s just buy some things on the internet. Let’s pick a hobby and start investigating it.” I picked drones, and started building a bunch of homemade drones myself. This is in 2011 and 2012.
One day, I was working in the office at an engineering startup. Across the hallway was a company that looked like they were building a drone, and I knocked on the window. Two days later, I was a co-founder at this company called Flirtey, which is doing drone delivery. That was really the job that led to Propeller, so I had some exposure to a startup and investors and how things worked in growing a business. We had a lot of exposure to the drones and the technology. I worked there for six months, and then myself and Francis, my co-founder, decided to peel out of that company and start our own. We wanted Propeller to do drones from a mapping perspective, to try to solve these problems for construction sites, mines, and landfills.
V1 Media: Explain your role with Propeller, and why you chose to get into construction.
San Miguel: I’m the co-founder and CEO of the company. When we started Propeller, we spoke to a bunch of construction companies in Sydney, Australia, where it’s notable that the drone regulations were a lot more supportive back in those years. So we had a commercial drone industry spun up already in Australia that we could leverage.
When we talked to these construction companies, we said, “Hey, we’re thinking of doing a drone startup. Have you used drones? What do you like about them? What don’t you like?” They said, “We really like them when we’re using them, but it takes our GPS and CAD teams a couple of days to do anything with the information. Make us this API for drone data. Make it really easy.” We thought, “That’s a brilliant idea. Let’s do that.” At the time, we were planning on building a few drones, but quickly sidelined the actual hardware. Instead, we got to work on removing the hassle of the drone process and deliver data straight to those engineering and construction companies. That’s what the business became really early on; and, for the most part, that’s what it is today.
V1 Media: Can you go through a little bit more about the Propeller Platform and some of the key technologies?
San Miguel: What happens is the drone flies around and takes a lot of images geotagged with a location that’s about as good as you can get off a mobile phone. It’s not sending great data—let’s say 5- to 10-foot accuracy. Each image and its geotag helps you stitch all those images together. This photogrammetry process looks at each photo for high-contrast points, then compares those high-contrast points to other photos, and looks for overlapping high-contrast points. When you have the same points noted in multiple images, you get perspective. You’re able to determine the depth and distance of that point from those images.
Then you run those algorithms across thousands of photos, and comparing each photo with each other photo. It’s really an intensive process. If you’re doing them on a local computer, like your laptop or a desktop computer, it might take 24 or 48 hours for that algorithm to run, depending on how many images. At the end of that, you end up with a big point cloud, similar to what you’d get out of a laser scanner, where you’ve got points in 3D space that represent the structure or surface you’ve been flying over and taking images.
We take that point map and turn it into a surface. We connect all the dots and make one big map and place the map over the 3D surface to end up with a 3D map of the site. That gets put into our Propeller vault and web software, and it sends an email notification saying, “Your map is ready.” Then you can log into Propeller and see that map in 3D; you can start measuring lines, points and volume off those 3D maps. You can compare those maps to other maps. You can pull back and forth along a timeline. We make software that helps companies digest the vast gigabytes of mapping data that get generated off a single drone flight.
We also automate most of the photogrammetry process, so we’ve got V-gammas on servers that do that crunching on behalf of our customers. But what’s very important and probably relevant to your audience is that we also have a 24-hour-a-day, 7-day-a-week data team made up of geomatics experts who actually do QA on every single job and help improve the accuracy of those maps down to survey-grade level. So each job is as automated as it can be. We’re also proud of the fact that each dataset gets looked at by humans who can make sure it’s good-enough quality to take measurements and help make decisions.
V1 Media: About the software, did you create this from scratch? How quickly were you able to ramp up the software itself?
San Miguel: So there’s a really great stack of open-source and already available software pieces that do this back-end photogrammetry processing. We’ve come up with our own stack, which is built on top of existing stacks, to get that “heavy lifting” done in the back end. That gives us a nice head start, and we’re very grateful to the open-source community for that.
V1 Media: What was that original open-source stack called?
V1 Media: Can you tell me a little about AeroPoints?
San Miguel: It was the start of 2015, and we signed up a quarry just north of Sydney as one of our first long-term customers. We planned on servicing this quarry by calling a local drone service provider to go out and fly the site, and then we would stitch all that data together. Now, one of the prerequisites to getting good-quality data is that you need ground control points.
So when we were getting this deal done with this first quarry, they told us they had ground control points placed all over the site–about a thousand–because they were getting manned aerial surveys done, and they needed those ground control points. So we thought we’re just going to send those drone operators out there, we don’t need to survey, and they’ll come back with the data. We’ll get the ground control points from the worksite, and we’ll do that work picture and present the data back to the worksite.
And that’s exactly what happened, except when we saw all the images back from the drone service provider a couple days later, we realized that the ground control points weren’t usable in any of the photos. They had been blown over and knocked over, so the ground control points weren’t on the site, and that was really frustrating, having paid the service provider and promising the site that they would get their data back. We really couldn’t do anything. We thought there had to be a better way to help the average drone service provider without a survey background and without the equipment to collect survey-grade data. And drones at the time didn’t have survey-grade GPS in them either.
Coincidentally, at the same time, I had been working on some low-cost GPS technology from my university days and decided to use this new wave of cheaper GPS equipment. Why can’t we leverage the recent price drop and build a GPS sensor directly into a firm target that will show up in drone maps? And that’s where AeroPoint was born. It started off as a pizza box—imagine that. We put a crosshair, with tape on it, and kitted out the inside of the pizza box with a small battery, and a little circuit board we designed and developed with this cheap GPS on it.
The product really went from there and became this option for people flying drones to purchase, where the requirement to get survey-grade data would be handled by this very “idiot-proof device.” There was one button on the AeroPoint. You press it, and it starts recording GPS data. Press it again, and it sends [the data] to us. We take it, add other sources of GPS information to correct that data, all on behalf of the customer, and you end up with very accurate x,y,z geographic points of the site.
At this first quarry site where we tested it, they now didn’t have to call their surveyor to come out and get accurate points so these maps would become accurate. And for us, that really signified that drone surveying was set to improve on cost and availability of survey data, which previously required traditional survey techniques to get out of the gate. So we were really excited to help the whole industry move forward with this step change, and that’s really what it has become. We’ve sold thousands of these things all through the world, and they get used thousands of times a month, and they have a really happy following, which we’re really proud of here at Propeller.
V1 Media: Another product I’d like to hear about is the Trimble Stratus; what that is, how it works with Propeller, and how you partnered with Trimble.
San Miguel: So there were a lot of drone data companies spinning up around 2014 and 2015 to help companies with this problem of processing data, and Propeller was unique because it had this AeroPoint technology to make the data really accurate. Trimble is the gold standard for survey-grade information and accuracy, so there was an instant connection, where they really appreciated this piece of technology and introduced us to their customers.
It just started off with Trimble; we had a call, and we hit it off. That was midway through 2016 and, by mid 2017, we signed off on a distribution agreement with Trimble for the Trimble Stratus product, which is a thinned version of the normal Propeller Platform with a few internal integrations built on top of it. We had a distribution agreement to sell Trimble Stratus and our AeroPoints through Trimble dealers and locally.
That was an excellent outcome for us. As a small Australian startup, we wanted to be working in the United States, and we needed some help to become more credible with our customers. We wanted someone like Trimble to stand behind the workload we’d developed and the accuracy of the products, so it all worked out really well, and Trimble has been a delight to work with. When we were looking around for an office in the United States, it was a “no brainer” to put it here in Colorado, because of their large presence here.
Now we’re working with probably 50 or so of their dealers now in the United States (and more internationally). The majority of our new revenue every month is along our Trimble channel. Overall, it’s an excellent partnership where we continue to work on more integration. We really want to make this Stratus product differentiated and viable for Trimble customers. That’s really the path we’re going on, and it’s going great.
V1 Media: Could you say a little more about the difference between Propeller and Timble Stratus, and also what makes Propeller different from others who are doing this kind of thing in the market?
San Miguel: In Australia, when working with the quarries, we were working in a public coordinate system, so were running Australia-wide coordinate systems that made sense for that market. When we came to America, we started working more with construction companies and more with Trimble, we needed to support local site calibration.
This is essentially defining a “zero point,” and everything is referenced back to that point.
So we were running predominantly on published systems, and we needed to support global systems to really do anything significant in the US market and with Trimble. So we worked alongside Trimble to build in native support for their coordinate systems definition. The have a JDAL file, which is a translation of x,y,z points in space from a global system to a particular local system. We can provide that same translation in our systems and generate data in the local systems relevant to the construction site. We wanted to include data out of Trimble Stratus coordinated accurately for construction sites, and that’s one of the cornerstone integrations. Anyone who is a Trimble Stratus customer can open up Trimble Stratus, drag their calibration file in and instantly that Propeller setup environment will be generating data compatible with their other systems.
When you’re at a construction site, you’re trying to affect the topography so you end up in some particular ideal setting usually described in a 3D design file. We can load those Trimble design files directly into Propeller as well, which means we end up with those designs loaded into Trimble Stratus. As a construction company, it’s not just about saying how much material you’ve moved, it’s about saying how much material you need to move as well. We can answer that question.
If you have a Trimble account already, you can log into Propeller straightaway, and you can also sync files back to your Trimble account from Trimble Stratus or Propeller. And that’s really the main difference. In terms of functionality of the product, it’s identical, but getting set up and making sure that everything works together is really where the integration will shine with Trimble Stratus.
V1 Media: What about overall differences from competitors?
San Miguel: Both startups and big companies are in this space; even Trimble has a photogrammetry processing engine and an ability to measure quantities in cotton fields, that sort of thing, over time, so there are a lot of companies doing similar things. I think Propeller has two real advantages. Number one, compared to some of the bigger incumbent software companies, Propeller is on the web, and that is a massive difference for our customers. Gone are the days of needing to install multi-gigabyte software on your computer with lots of buttons to get these outputs. We’ve made it a really simple thing. You just log in, you can see all your data, you can use all the tools, and get those coordinates really quickly on any device.
The second advantage is that we can actually and genuinely manage our customers’ processing demands for them. Because we have this 24-hour support team and infinitely scalable server architecture, we can genuinely process and take on that responsibility for our customers, and we do that in a way that Trimble, for example, just cannot, and that’s what they pay us for. That’s how I think the web-service approach is really favorable here.
Compared to our startup competitors, who also are web-based, the big difference is that Propeller is incredibly focused on earth-moving workflows. We’re not trying to do agriculture, we’re not trying to do inspections, we’re not doing anything more. There are all these other things you can do with drone mapping. And what happens when you try to focus on too many things is that you essentially focus on nothing. You end up with a product that’s fragmented and difficult for an end customer to get the value they’re after.
Our strategy at Propeller has been: let’s pick these verticals, and then let’s work with these sites and understand all their problems, whether they’re drone related or not, and let’s just start building exactly what these verticals want in a really specific way, so Propeller is the most industry-specific-focused version of these workflows. And that’s really what led to our growth in the market. It’s the obvious choice if you’re a construction company trying to do it right, as far as I’m concerned.
V1 Media: So how do you go about educating potential customers such as construction sites and designers, about what you do and how you can help them do their jobs better?
San Miguel: It’s a never-ending battle on the education piece. Some people want to adopt new technology; they want to be the first to do it. Some people don’t want to adopt new technology; they only want to adopt things that have been proven. So our education approach needs to cover both aspects.
We’ve got a company strategy that helps people who want to adopt the new stuff. There’s a company strategy for people who want it to be the “tried and tested” way of doing things. And it’s getting both categories spun up in the right quantity and at the right time to make that successful. So we leverage a lot of online tools, a lot of our content. Our referrals are a deep part of our education plan now. We get some companies who have had success to communicate to other companies that are thinking about it, just how it happened, what does success look like, what the ROI looks like for the business. It’s pretty normal in that respect.
The industry is becoming more modern with each new generation that comes into construction. Some of that hesitation is starting to subside a little bit. It’s definitely not a fast-moving industry, and in some ways, it’s a good thing. It means there’s an opportunity for Propeller to figure out how to get new customers in a way that other companies cannot. But it’s a shift in mindset over time and to witness new technology going from being unusual and risky to being normal. That’s a really proud thing for us here at Propeller.
V1 Media: So how do you see Propeller evolving and improving in the future?
San Miguel: The focus of the business has gone from processing peoples’ drone data to becoming more of a comprehensive worksite tracking and improvement solution. So the business really is becoming more focused on the problems we are solving and how we solve those problems. So we’re trying to solve those problems, but do it in ways that make even more sense to customers. So you can imagine Propeller taking other sorts of survey data in the future, drone or not drone, to really consolidate all that information and make it as easy as possible for end customers in construction to come and get value from. That’s what the plan is here with Propeller.
V1 Media: And so let’s talk a little bit about the future of construction and land development. How do you see those changing in the future as all this new technology is coming into the market?
San Miguel: I read reports and hear from customers how often they’re over time and over budget on their jobs, especially in the United States. To me, it’s just not surprising when it’s very difficult to track what you’re doing. It’s very difficult to track your progress against the plan. These construction companies are looking to fix the problem of tracking their progress and productivity, and feed that back into the decisions they’re making.
We’re moving from guessing to making data-informed decisions. These worksites will have a stream of information coming in from their sites; foremen and superintendents will have a stream of information, and they’ll be able to make really good decisions to actually improve that problem and have projects running on time and on budget. And that’s the most obvious and exciting trend for me. In the next couple of years, information collection and analysis will be automated, and it’s going to be served up to these superintendents on a platter so they can make the right call, almost instantly.
V1 Media: In your area of expertise, how can our readers do their jobs better using current technology?
San Miguel: Well, it depends on the job. I think there’s often a lot of work that needs to be done inside the organization trying to adopt new technology. There’s going to be people inside a company who are resistant and excited together and separately, so there’s a bit of work that needs to happen internally and for the company to decide that to reinvent itself and to become or continue to be relevant and successful and profitable in the future, it needs to be focused on weaving best-in-class workflows and technology into the existing business.
You can look around and see drones flying over construction sites; people aren’t here to waste their time. If you’re not flying a drone on your construction site, then it’s probably worth asking “why?” And it’s probably worth talking with some other companies to ask if they’re doing it, if they’re getting value out of it. And think about adopting it yourself.
But if you are on the adoption path, make sure you’re putting the groundwork into your organization to actually support these new workflows as they become valuable to the company. I think we’re past the point of no return with drones on construction sites. It’s not actually that risky anymore. The drones have matured; the regulations have matured; now’s the time for the mass market to start adopting it.