RW Holdings NNN REIT, Inc. to Merge with Rich Uncles Real Estate Investment Trust I and Acquire its External Advisor. COSTA MESA, Calif., Sept. 20, 2019

What are the competitive dynamics in the consumer drone market? What factors (e.g., software) will determine the winners?

It’s All About Product Positioning!

Instead of some “best, faster, cheaper” analysis, this post will take a deep dive into the history of the consumer drone market and how products have been positioned within it. To understand these products and how they have shaped the market, first let’s look at the major players and their chronology.

DIY Drones and 3DR

We owe “drones” as we now know them to Chris Anderson and DIY Drones (Full Story). In 2008, Chris created the online community DIY Drones to start thinking about flying robots and offer support/guidance. They borrowed vehicle designs from academia and low cost electronics from a frenzied cell phone market that had just seen the introduction of the iPhone, but there was a missing piece. The hardest part of making an early drone was attempting a DIY flight computer, so Chris’ newly formed drone company 3DRobotics partnered with a Swiss University (ETH Zurich) to create the PX4 open hardware platform and ArduPilot, a similar, but comparatively less robust system based on the Arduino prototyping platform.This dramatically spread interest in quadcopters among hobbyists and created a new branch of the already well-established model aircraft community represented by the AMA (Academy of Model Aeronautics).

So now we have quadcopters, and they’re really fun to build and fly, but we’re still figuring out what to actually do with them. Enter Parrot.

Parrot

At CES 2010, Parrot introduced the AR.Drone, our first example of excellent product positioning. It was a relatively cheap ($300) and mostly foam toy, designed to be controlled by wifi and a mobile interface. What was the most important word in that last sentence? Toy. It was an extremely successful toy for 4 reasons:

  1. These products were specifically marketed (and very well at that) as a toy in both the US and Europe, effectively penetrating these markets through Apple Stores and other prestigious electronics distributors, yet they kept a very attractive price point.
  2. Easy to operate / fly
    1. The mobile interface allowed for control via “motion events,” a.k.a. tilting or rotating the device in order to command the drone to go forward, backward, etc.
  3. High crash tolerance and low propellor risk
    1. The foam shell, overall low weight, and flexible / ducted propellers made the drone very safe and crash friendly
  4. The system was “ready to fly.”
    1. Being “ready to fly” is hugely important, as you no longer need a hobbyist level of understanding / skill to get your flying robot off the ground. No more sifting through forums to find solutions to common problems, no more researching obscure replacement parts, significantly fewer fly-aways — it works out of the box. “Ready to fly” was a big step in solving the largest problem facing drones, which is best described by the question, “When I take off, is the drone going to do what I want it to do?”

The importance of making drones reliable and less buggy, with fewer fly-aways and fewer crashes, has been a key narrative in the evolution of the consumer drone industry. However, with quickly improving hardware, safety, reliability, and capability, we are confronted with an even deeper and more troubling question: “What the heck do we actually do with these flying robots?”

Enter DJI.

DJI

DJI has defined and cornered the consumer drone market. Let’s look at how and why.

(Forbes’ Ryan Mac on DJI’s history) In 2006, Frank Wang moved from prototyping flight computers in his dorm room to manufacturing and selling them in Shenzhen. Like Apple, DJI began by designing and manufacturing the technological building blocks (a.k.a. enabling hardware) for a market they would both soon define.

Back to the ever-persistent question, “So what if the robot flies, what does it actually do?” Many people don’t understand why this is important because they look at the consumer drone market as a collection of flying robots. To better understand this, you need to reframe how you view the market. Try this: tell yourself that there are no drone companies. Stop imagining DJI, Parrot, and Yuneec as just selling drones. Parrot sells many things from headphones to navigation equipment. They also sell some great flying toys.

What does DJI sell? They are a camera company. DJI sells cutting-edge camera equipment, which in some instances, happens to fly. “But,” you retort, “DJI has sold around $1B worth of quadcopters in 2015 and the Phantom often defines what most people think of when they reference a drone! How then, can they be a camera company?”

This brings us back to our initial question. The most obvious use for drones was slapping on a camera and collecting video from a previously inaccessible area. The problem was that for a drone to move around, the entire airframe (what we call an aircraft’s skin and chassis) needs to rotate in some way, causing the thrust from the propellers to not be pointed straight down, but in a direction that angles the drone. The problem arrise that every time you need to move your drone through space, you inadvertently change where your camera is pointing.

Roomba Creator Cyphy solves this with skew propellers and clever control system design. A fantastic solution when flying in still air, but largely ineffective when dealing with winds and intense acceleration

To finally get smooth and usable video, we use a gyro-stabilized gimbal, a motorized camera-mounting device that measures how a body (the body of our drone in this case) is rotating, and then uses those measurements to calculate the exact opposite rotation. The motors immediately perform this opposite rotation, which causes the mounted camera to remain still. This system is what we refer to as a steadicam and under the hood, so to speak, it is very similar to an autopilot. http://gph.is/1okvsd4

The autopilot also measures how the drone body is rotating, but instead of calculating the opposite rotation, it calculates how the drone would need to rotate to be in a stable hover. So DJI, who is very good at making autopilots, finds they are also very good at making gyro-stabilized gimbals.

We now have all the ingredients for the Phantom 1 and a suitable answer to our question. The Phantom 1 was huge, the catalyst for DJI and the aerial photography market. It was a ready to fly camera drone that actually took good pictures and video. Finally, people weren’t just flying robots around, they were creating value by capturing data, a big leap in overall drone usefulness. I keep harping on value and utility because that’s what defines DJI as a camera company. At the end of the day it does not matter that this robot left the ground, moved around, and then landed. What matters is the data it was able to collect and the value of that data. This robot could video data that a real estate broker can use to advertise a house, multispectral data that shows crop health in a farm, or just a great drone selfie that you just want to watch over and over.

Now the explosive growth happens, but we haven’t said why. As it turns out, photo and video is THE use case for drones, and the Phantom is the first drone to do this well. Yes, delivery by drone is exciting, but it’s still a ways off. Search and rescue, infrastructure inspection, agriculture, insurance, marketing are really just subcategories of aerial photo/video and the Phantom can tackle them all. This marks the beginning of a trend we still see today: drones as a platform for software. As we move through the Phantom 2, Inspire 1, and now the Phantom 3 (I’m really hoping DJI’s March 1st announcement is the Phantom 4), DJI consistently improves their ready-to-fly-ness, ensuring fewer fly-aways, better usability, and an easier and easier pilot learning curve. They accomplish this primarily with better software while also improving the quality of their cameras and flight hardware.

Enough with the history, what about the future?!

Well, to talk about the future, let’s start with the present and project forward.

DJI and the continued dominance of the Phantom

DJI has great flying cameras that out-perform and undercut every other photo/video drone. This is why even a heavily venture-backed and high performing company like 3DR will struggle immensely when competing with DJI’s hardware.

What’s in store for 2016 then? First off, it’s not the Phantom X. I view this purely as a market research study to test by means of public comment if DJI should invest in Lily Drone-like features and technologies. I’ll go into detail on these when talking about Lily below.

DJI will continue to improve its breadwinning photo/video products, the Phantom and Inspire series, while further testing the waters in other areas. DJI is exploring agriculture hardware with the Agras MG-1 as well as R&D hardware with the Martice M-100 and guidance kits for advanced computer vision. However, the prime directive is still to deliver increasingly better products for creative aerial imaging. So far we have only seen one investment from the DJI/Accel Partner’s Sky Fund in DroneBase, an aerial imaging services marketplace. In contrast, DJI’s largest investment was purchasing a minority stake in Hasselbald (a centuries-old Swedish camera maker); their priorities for high-quality imaging hardware seem clear. This makes sense as nearly ALL of the money being made in drones today comes from photo/video drone hardware sales. Often overlooked in DJI’s growing video empire is the Osmo, which is defining and driving the handheld steadicam market, a big hint that DJI may not be a one trick pony that struck gold with the Phantom. The Osmo was extremely well received by Vloggers and other online content creators who make up one of the fastest growing areas of photo/video. The Osmo could also be a major player the next progression of action camera technology as we react the limits of digital image stabilization.

DJI continues to champion risk and innovation in their R&D and product design. Over the summer they opened a R&D center in Palo Alto headed by Apple’s previous director of antenna design and Tesla’s previous director of autopilot design. Perhaps the most important DJI initiative that will influence 2016 is their quickly expanding mobile SDK. A robust SDK will allow DJI’s flagship products to truly become platforms for mobile software. In this way, the past competitive dynamics of the cell phone market will shape those of the consumer drone market: high quality and standardized hardware will enable cost-effective applications for every use case. This, mixed with favorable regulation and better safety/ATC systems (DJI also picked up the guy that build Uber’s public policy team from scratch), is what the collective startup that is consumer drones needs to continue growing. Without them, there is no definitive reason that drones will be guaranteed to become mainstream in the next 5-10 years.

A make or break year for 3DR

The Solo drones’ 4 smart shots and GoPro compatibility were not a strong enough product differentiator for large adoption in 2015, despite excellent marketing.

Considering the poor GPS performance, gimbal shipment delays of almost 4 months, and the $1800 total price tag, the Solo needed to display earth-shattering performance or assisted piloting / smart shot features that 10x’ed the market.

However, the poor GPS performance nullified the small value added by the smart shots, which quickly became a standard adopted by competitors (DJI, Parrot, Yuneec, EHang, and Hexo+).

So going back to my final point on DJI, I would like to see 3DR abandon the Solo to focus their efforts on standardized, market-leading flight computers and open source software like the immensely valuable DroneCode and DroneKit projects. With their simple, market-leading, and beautiful products, as well as their expansive but closed source SDK, DJI is truly stepping up as the Apple of consumer drones. 3DR is then the best positioned and most capable player to rise as the Android analoge. I strongly believe it will be nearly impossible for them to compete with DJI on hardware, but they can dominate software, which history has consistently shown to be the winning bet in the Silicon Valley.

Parrot

The Parrot Bebop aimed to fill a known gap in the market, a smaller, easy to use, sub $500 drone that can also take high-quality video. So they created a compact, wifi-based (i.e. cheap communications), $500 quadcopter with a gimbal-less software-stabilized 4K camera. These were all great design choices that, in theory, would have well positioned the Bebop. However, with poor connectivity from the wifi and a camera that was primely positioned to take the brunt of any crash and even then only took so-so video, the Bebop was also a no go. “But the sky controller,” you say. Well, now the Bebop cost $800 and is still greatly outperformed by the $500 Phantom 3 Standard. Unfortunately, the Bebop 2 failed to fix any of these fatal flaws while raising the base price to $550, smh.

To win in 2016, Parrot should continue to improve their minidrones, which are excellently positioned toys, and try one more iteration of the Bebop, but with a substantial redesign. The purpose of this is to position the Bebop a less complex and less expensive Phantom that will rise above the almost innumerable low cost Phantom copies produced by what drone analyst calls China Inc.

For the Bebop to be successful it needs to be <$400 (a $300 price point is preferable), easy to fly, easy to transport/carry, better protected camera lens, crisper video picture, and most importantly it must maintain a reliable connection (I’m talking >90% uptime) at 500ft range. Parrot currently claims the Bebop 2 has 300m range (~1000ft), however, this cannot be allowed to happen, ever.

Yuneec: A Well Supported Challenge To DJI

Yuneec began making radio controlled model aircraft in 1999 and over their 17 years in business they have been a pioneering force in brushless electric motors, radio equipment, and electric powered manned vehicles including light-sport, ultralight, and general aviation aircraft. In a logical progression from model aircraft, yuneec used their radio and motor technology to lead the very early hobbyist quadcopter market as its largest manufacturer. It’s important to note that until very recently Yuneec and always white labeled their products to be marketed and distributed by US partner companies, most notably Horizon Hobby. This is why Horizon’s Blade Chroma series looks almost identical to Yuneec’s Typhoon and why it features the same integrated display on the controller. In all of this Yuneec has proven to be a manufacturing powerhouse, delivering 1 million units per year for the past 10 years.

Yuneec made a huge splash in August when Intel Capital announced they were investing $60M in the drone maker. Just 2 weeks later, Qualcomm introduced the SnapDragon Flight, a version of their flagship mobile processor designed specifically for UAS and announced that Yuneec would be the first OEM to use the chip. So, Intel and Qualcomm made their bets, on the same pony…

This lead Yuneec to win nearly all the CES drone awards this year (most notably from Engaget and PCMag) with its Intel RealSense enabled Typhoon H. The new Typhoon, however, did not use the Snapdragon Flight, which was instead was demoed on Tencent and Zerotech’s Ying drone. Regardless, Yuneec has the track record and support to truly challenge DJI’s dominance, but this will be a longer play that extends beyond 2016. Despite the hype, Yuneec still has a ways to go before the Typhoon and Tornado can adequately take market share from the Phantom and Inspire. Both sets of competing products are relatively close in terms of performance, cost, and positioning. The one key differentiator is Yuneec’s integrated display, which, unlike DJI products, does not require an additional mobile device to access key features such as live video feed, telemetry data, and autonomous mission planning. The major differentiator is the modular sensor bay that can hold a 360 degree sonar, Intel RealSense, or infrared camera array. These sensors have already been integrated to enable sense and avoid technology developed by recent Intel quisition, Ascending Technologies. AscTech has been the go-to high end research drone for academia making notable appearances in almost every groundbreaking UAS technology demonstration by UPenn’s GRASP Lab, ETH Zurich, and Intel themselves back in early 2015. AscTech has been sucessfully by selling state of the art drone hardware and processors with relatively open firmware. This is immensely important to researchers because if they were to use a DJI product to test new autonomous or sense and avoid technology, they would have to remove all of DJI’s proprietary firmware and rebuild it from scratch to incorporate whatever hardware and software they are testing. Yikes. So in exchange for easy to develop and open software/hardware, their products cost a small fortune only affordable to top universities and Intel.

What does all of this mean for 2016? In the short term Yuneec will take a small amount of market share from DJI in photo/video drones, but their real play is for 2017 when truly autonomous UAS enter the market. As was said before, it is very difficult to compete with DJI current products, however, autonomy is poised to change who’s hardware dominates and the first mover will have a tremendous advantage, similar to what DJI enjoys today. This is a nice segway into autonomous drones and new technology/advancements that will actually matter in 2016, so let’s start with the crowd favorite, the Lily Camera.

The Madness Of Lily And The Challenges Of Autonomy

Lily exploded onto the web last May with their ridiculously viral product announcement video and a hugely successful preorder champaign. This got them $34M in presales and $15M series A round lead by Spark Capital, but to make their $999 price point and deliver on the wide array of features shown in the promo video, they’ll need every penny. To understand my reservations about Lily and “follow me” drones we need to look at the state of the art for intelligent autonomous systems.

A flying robot that follows you around requires a list of increasingly complex things. First, it needs to be able to maintain a stable hover in one place. Companies have figured this out by using GPS and downward facing optical flow cameras that watch how the robot moves with respect to the ground to catch drifting. Now that the robot can reliably stay in one place, it needs to move, just not into the nearby tree. So the robot needs information on its environment which it can get from cameras, sonar, or LiDAR. In each case, the incoming data needs to be heavily processed to be useful so most systems employ what know as SLAM (Simultaneous Localization And Mapping). SLAM first uses the data to build a 3D model/map of the robot’s environment and then figures out where it is in that map. This process involves taking thousands of sensor measurements to create a cloud of data points that make up the 3D model. Now our robot needs to figure out where its target is and plan how to get there. After calculating hundreds of options, it selects the the best path and finally the autopilot calculates what how fast each motor should spin to move the robot in the right direction.

RECAP: Be stable, build map, find yourself on map, find where to go, plan steps to get there, execute first set, and repeat, one thousand times per second.

Now we see why some AscTecs come with a i7 processor and 4GB of RAM yet can only autonomously move at a walking pace. The time required to figure out the perfect move only allows for limited amount of moves per second. Researcher have worked around this by collecting less data, planning less paths, and accepting greater overall error. This allows the robot to move quicker, but it’s no longer making perfect choices and risks a crash. This is the greatest barrier for intelligent autonomous systems.SidebarYou make have seen videos of autonomous drones moving very quickly and completing incredible precise missions. These systems cheat by using an array of infrared cameras that track markers on the robot. This is called a motion capture or MoCap system and they take care of the SLAM process as long as the robot is in the middle of at least 4 extensively calibrated cameras (if the camera is moved at all, the recalibration process is slow long I could only find a timelapsed example).Who has the best autonomous drones and what are they capable of? Besides the mysterious work of Skydio, the most advanced demonstrations go to DARPA, MIT, Intel, and Qualcomm. This means that Lily is attempting to compete with Yuneec’s manufacturing prowess, Intel’s processors and cameras, and AscTec’s years of autonomous UAS R&D. Why do this? So a robot can follow me around and take pictures… I think we need to have a serious conversation about ROI.

Why Certain Drone Technologies Are Less Important (Spoiler Alert, it’s ROI)

Remember when we discussed how photo/video was the best use case for drones? That’s because aerial photo/video holds the best ROI. This is the same reason drones have yet to revolutionize agriculture. In their UAS economic impact report, AUVSI predicts 100,000 drones will be sold for precision agriculture by the end of 2016, which is roughly 1 drone for every 10 US farms with annual sales >$10,000. For all of these farms to benefit from UAS, every dollar they spend on buying and operating the drone must create more than one extra dollar in crop sales. For now, manned aircraft and satellite imaging continue to be more cost effective than agricultural UAS. This is the same reason slight advances in battery technology don’t matter just yet. The 20-30 minute flight times that most consumer models enjoy is enough for almost every application and swapping batteries, while annoying, remains cost effective. Another factor is the FAA’s ban on beyond line of sight drone operations, which is also why we are unlikely see drone logistics and package delivery until 2020.