Mapping The AgTech Adoption Curve
The Mixing Bowl’s Seana Day, Rob Trice, and Krista Holobar recently sat down with Ryan Rakestraw, Venture Principal at Monsanto Growth Ventures, during the PrecisionAg Vision event in Phoenix to talk through the “AgTech Hype Curve” he put together. MGV invests in a broad range of companies that can have a major impact on the future of agriculture. The slide represents his own views and is a work-in-progress.The Mixing Bowl: So, what compelled you to put this curve together? What’s the background on this document?Ryan Rakestraw: It was mostly for the team at Monsanto Growth Ventures (MGV) to think about the investment landscape, think about where technologies are in terms of their maturity levels, and think about some new technologies that are just emerging. We wanted to capture some of the more mature technology that is seeing farmer adoption, and some of the less mature technology that has yet to experience any significant adoption. This is a useful representation to remind us that some of these technologies have a little bit further to go before they get to a level where they see wider adoption.The Mixing Bowl: Can we talk through some of these categories, particularly some of these “Innovation Trigger” stage topics? What is Synthetic Aperture Radar, for instance?Rakestraw: Satellite imagery has been used in agriculture for some period of time. Many of the precision agriculture solutions available today, whether Field View, Encirca, or Farmers Edge, all rely on an underpinning of satellite images to take pictures of crops to assess their health. There are a few limitations with satellite platforms. One has been the spatial resolution. Some of these photos are from seven meters to 22 meters in resolution. It can be difficult to find different features associated with that low level of specificity. The other thing is just the number of images that are being delivered in a season. I think farmers would like to see an image every couple of days, and they aren’t getting that today. These two issues are probably going to be mitigated by another area that’s on the hype curve—nano satellites. Planet Labs and others are deploying larger constellations of them. Another limitation is that, at any given time, 50% of the earth is covered by clouds, and when these satellites are taking images they can’t see through the clouds. Those images become unusable for agriculture purposes.Synthetic Aperture Radar (SAR) actually uses a different piece of the electromagnetic spectrum. It’s using more radio frequencies—like we use in communication—that enable cloud penetration to take an image of a field. It’s not exactly the same type of image, but it does allow you to look through clouds, and to look through some of the canopy, or some of the foliage, to see things that might be structural aspects of the crop. In some cases, SAR may be able to look at features underground. SAR is an interesting technology that can be deployed either on a satellite, on an airplane, maybe even on an unmanned aerial vehicle (UAV). I view it as primarily helping with the cloud problem for satellite imagery.The Mixing Bowl: What does “deep learning” mean in agriculture?Rakestraw: Deep learning is an analytical technique that falls within a broader category of algorithms called machine learning. Companies and researchers are using machine learning to train computers to sort through data, determining various features or relationships that might be present. I think an interesting, non-agriculture example of machine learning is Google’s AlphaGo competition where they used a computer to play the game versus a Go master. Companies are using a variety of machine-learning techniques in different areas.Deep learning uses “neural networks” structured like the human brain, to match patterns or extract features from data sets. Deep learning is particularly good with images. Companies like Facebook, for example, will deploy deep learning algorithms for feature detection associated with uploaded photos to do auto tagging or capturing.In an agriculture example, this is a class of algorithm you might use to make sense of some of the images that are being taken from UAVs or satellites: What is the information in these images, what do they actually represent, and is there a certain pattern present in those images that would lend themselves to there being a particular pest or disease present? You also see these types of deep learning algorithms being developed in robotics and automation settings. Much of the work in the autonomous vehicle space is being driven by deep learning. Robotics in agriculture, whether a harvesting or picking application, or a fully autonomous tractor, would likely rely on deep learning to make real-time, in-field decisions.The Mixing Bowl: Haven’t “Uber for Tractors” and “Amazon for (farm) Inputs” been tried before?Rakestraw: I think the market wasn’t ready previously. Particularly in agriculture, there are very established channel relationships from which people tend to buy their inputs and equipment. A shift may be occurring due to demographic change—younger farmers are more familiar with transacting commerce online and heavily using smartphones. We are just getting to that place where the Farmers Business Network and others can provide platforms where farmers are comfortable buying online versus a traditional channel.We’ve also seen these sharing economy and marketplace models work a little bit better in emerging markets. There are a handful of companies in Africa and India and elsewhere where the growers don’t have their own tractor, so sharing equipment has been more useful and meaningful to them. In emerging markets an online platform is a great way to get people access to some of the equipment that they may not have today. In the U.S., “Uber for tractors” hasn’t seem to work as well. The problem in the U.S. is often timing because so many farmers are trying to get the seed for the same crop planted roughly at the same time. And everybody is harvesting the same crop in the same region at roughly the same point in time. For row crop agriculture, like corn or soybeans, equipment capacity gets filled up at the critical points. These sharing platforms are probably more applicable in the high-value crops and perennial crops.The Mixing Bowl: Are there any of these other “Innovation Trigger” technologies you think we should point out?Rakestraw: We’ve been starting to spend more time with the application of block chain in agriculture. This is a technology that could be pervasive throughout a number of industry verticals, and we feel that this will make business transactions in agriculture much smoother, whether that’s selling or marketing grain, purchasing inputs, buying services like custom spray applications, etc. Block chain can also enable different business applications. Instead of paying for a custom spraying, this technology could unlock more creative business models, like sharing value with the grower. Instead of paying somebody $6 an acre to spray, a service provider would get a certain percentage of the profit after the season.The Mixing Bowl: Let’s move to the “Peak of Inflated Expectations” category. We’ve noticed almost nobody at PrecisionAg Vision is talking about drones. What’s up with that?Rakestraw: I think there are a couple of reasons. Nobody has found the perfect use for drones today. I think that they’re being adopted in agriculture. I think there are a lot of farmers who are buying their own DJI Phantom platform and playing around with it, but nobody has really found the perfect application. Even for a scouting application, which I think is probably the lowest hanging fruit, nobody has exceptionally delivered on the drone experience. Regulation has also been a limiting factor. Up until very recently it’s been very expensive to fly by the books, following the regulation as it has been outlined, with the required certifications to fly. I think that the Federal Aviation Administration (FAA) has been very progressive in lightening some of those regulations, and that should give more people the opportunity to adopt this technology.The Mixing Bowl: Anything more on the technologies at the peak of inflated expectations?Rakestraw: We’re starting to see several companies that fall into this IoT (Internet of Things) space. They are connecting sensors that are deployed across the farm– in the field or other places of the farming operation. We have seen a large number of technology startups that are doing this, but aren’t necessarily delivering value yet. There are certain applications—measuring dryness in a grain bin or soil moisture in high value crops—that work well. But a lot of people are just saying they do IoT —particularly technology companies that are promoting IoT across a variety of different industries—without necessarily understanding the direct applications of IoT on the farm. There is a lack of differentiation in these companies’ offerings. At this point, many people are using IoT more as a buzzword without creating much value.The Mixing Bowl: You have “aerial imagery” in the “Trough of Disillusionment.” Why? And how does that compare with satellite imagery that you have moving into the “Slope of Enlightenment?”Rakestraw: I think satellite imagery is relatively well adopted. Many growers today are getting satellite images of their fields. Many companies across the agriculture value chain are using satellite imagery to look at supply chains and assess risks. The Landsat satellite program has been around for 30-plus years, and the U.S. Department of Agriculture (USDA) has used satellite imagery for about that period of time so this pretty mature. Aerial imagery- images taken from fixed-wing aircraft — is at the cusp of going up the slope of enlightenment. It should be able to deliver incremental benefits on top of satellite imagery, like higher resolution, and the ability to capture more images per season. But at this point, nobody has been able to really make the business model work well. There are a couple of emerging companies that are doing especially well, but historically in aerial imagery, nobody has been able to scale outside of a small regional footprint.The Mixing Bowl: Looking at the Plateau of Productivity, are there takeaways from these technologies that we should look at as success cases or lighthouses that other ag technologies should be looking at to mimic for success?Rakestraw: A few common themes emerge amongst those technologies that have reached the Plateau of Productivity. One is that these technologies got to the point where they are relatively easy to use. Incorporating auto steer in a tractor or combine matured to the point where it worked without extra effort from a grower, and it saved the grower time. Another theme is that the return on investment (ROI) associated with these technologies became very clear. The in-cab display technologies in tractors and combines were pretty useful when they enabled the farmer to have a really good real time assessment of what was going on to avoid double spraying particular areas or make sure there weren’t planting issues. So in-cab displays were easy to use, had a direct interface with the grower, and were something very amenable to the existing process. If we take the example of yield monitor technology, I think this was data that the farmers were highly interested in. Yield monitors have been around for years in various forms and have seen continuous, incremental innovation to get better and better. Now the yield monitor is a fundamental tool for a lot of the decisions that a farmer might make. Looking at soil sampling technology, there has been an understanding that the soil is a critical aspect of growing, and people have been soil sampling for a long period of time, and we have seen incremental adoption. It seems that today we are getting better at understanding the soil properties that lead to optimal growing conditions, so that the analysis of soil samples is leading to greater actionable insight and thus of greater value to the grower.The Mixing Bowl: Any high level closing thoughts on AgTech innovation? Where are we and where are we going?Rakestraw: I think everybody—particularly those of us at the PrecisionAg Vision conference—feel AgTech will get to the plateau of productivity, where these technologies are going to help farmers make better decisions, to optimize profitability, to facilitate sustainability throughout the industry. We are going to get there. I think the big question is when. These products need to develop further in terms of the sophistication of the technology. Our understanding and knowledge of core agronomy also needs to advance a bit. Particularly with the software, hardware, and data tools we have now, how do we make sure that our agronomic understanding is incorporating the new data we have at our disposal?I think we’ll get to this plateau of productivity—but if you talk to a startup company developing technology in this space, they’re probably a bit more bullish on where the AgTech industry as a whole is on the hype curve. Whereas if you talk to an average farmer in the U.S., they’ll say that we are closer to the peak of inflated expectations.