“Software is Feeding the World” is a weekly newsletter about technology trends for Food/AgTech leaders.

Greetings from the San Francisco Bay Area.

Normally, I don’t talk to high school students very often, other than within the family or friend’s kids. Last week, a high school senior I didn’t know (12th grade student) doing projects in machine learning and synthetic biology reached out to me by email, and wanted to have a conversation with me. I was not sure what he wanted from me.

When I was a kid growing up in India, I was hesitant to ask for help, and was often intimidated by seniority or authority.

Tyler Cowen wrote in a blog post in 2018, “The high-return activity of raising others’ aspirations”

So I decided to have a 30 minute video conversation with the student, and encouraged them to try different things, be in a learning mode, have the courage to experiment, and to reach out to experts and other people to talk and learn from them. I was impressed with the student’s passion, attitude, and ethos. I am not sure if I was able to raise their aspirations or not, but I definitely tried.

If you know any young folks who want to have a conversation about future careers in tech, being a creator, or similar topics, I will try my best to have a short conversation with them.

Last week, I had said I would write about large language models aka LLMs (ChatGPT is one incarnation) and their applications in agriculture in this week’s edition. As I started researching and writing about LLMs and agriculture, I realized I could not do justice to the topic, with my current rudimentary (and probably not good) understanding of how LLMs work, and how they could be applied to agriculture. Due to this, my writeup was coming across like a solution looking for a problem, or tech being pushed for the sake of tech.

I was being something, which I had written about in edition 127 with the incident up in the air on my flight from New Delhi to San Francisco.

I was being a tech-bro!

Long story short, I realized I need to spend more time to think through different problems which face food and agriculture systems today, and how LLMs could apply to them. Stay tuned.

In this week’s edition,

1. FMS, thin markets, and loss leaders: Janette Barnard’s analysis about the business and investments in farm management software for commodity row crops

2. Smile-curve for ML/AI as a service based on the analysis by the inimitable J. Matthew Pryor of Tenacious Ventures.

3. Four investors on what’s next in Food Tech and AgTech

Also, in the news, SFTW is featured (again) in the best of the web by Precision Dealer, how IoT is impacting multiple adjacent industries, use of comics to tell stories about Mexican corn, Bayer and Kimitec form partnership for biologicals, how tech is making farm workers jobs safer, Agtonomy unveils tele-farmer solutions, California sees reduced pesticide use, camelina as energy source, Scoular connects with producers on sustainability, and Yara India reaches millions of farmers digitally.

FMS, thin markets, and loss leaders

I am and have been a vegetarian most of my life. It is not due to religious or environmental reasons, but it is the taste I have developed growing up in a vegetarian household in India.

Janette Barnard is an exciting writer, with an interesting take on innovations in the animal protein value chain. Even though I regularly read Janette Barnard’s newsletter Prime Future, many times I don’t understand some of the issues related to animal protein. Last week’s newsletter was no exception, as Janette called out the challenges with farm management software companies within the commodity row crop sector.

Janette is spot on in her analysis of a flawed market being a winner-take-all market.

The faulty winner-take-all assumption was compounded by a few other factors. When Monsanto acquired The Climate Corporation, the assumptions on productivity improvement through analytics were in the range of 15-20 bushels per acre, when the prices of corn and soy were some of the highest prices in the history of commodity markets.

Even as late as 2017, it was common to talk about improvements in profitability of $ 100 per acre, based on farm management software analytics. It created an overhang on farm management software companies like The Climate Corporation (full disclosure - I worked there from 2017 to 2020), and Granular. There was a belief about creating a data flywheel to continue to drive more value and in turn bring in more data from commodity row crop farmers and drive even more value. The reality was a bit different.

Thin Markets

As I wrote in edition 117 and 118, (Thin markets in commodity row crops), and as pointed out by Janette, the diversity among the 200K commodity row crop farmers in the US, make the market even thinner.

But the hypothesis on value identification, value creation, and most importantly value capture were very high. The cost of customer acquisition was quite high compared to the lifetime value of a farm management software customer, if you evaluated it as a stand alone business. This was true even for a company like Bayer, with a large distribution network, and strong grower relationships.

Even if the farm management software was able to identify and create high value, due to the complex nature of the open system of agriculture, it was very difficult to attribute the value created to the software. This made value capture from the potential value created by the farm management software quite difficult. Charging a high price for the FMS applications was not feasible, nor was it feasible to do value based pricing due to the value attribution problem mentioned earlier. Ultimately, it turned into a race to the bottom.

To make matters worse, the value identification and value creation were hobbled due to lack of good quality data coming from the farm. The precision agriculture data coming from planters, and combines often has many data quality issues, which makes robust and repeatable analysis quite challenging.

In an open system, many different factors like seed genetics, weather, soil, practices like fertility, pest and disease have an impact on the final outcome. The challenging problem of decomposing your final outcome (for example yield) into its components is made even more difficult due to bad and incomplete data, and a difficult data entry user experience.

FMS are loss-leaders for input companies

In such an environment, it makes sense that many independent farm management software providers have faded away, or have been folded into the digital offerings of large input providers of seed and chemicals (for example, Bayer and Corteva). The FMS applications from input companies act as a loss leader, and help make digital connections between input sales people, agronomists, and service providers and growers stronger.

The FMS application provides better service, and increases the lifetime value of the customer (LTV) through larger share of their input basket, or higher brand loyalty.

For example, when I worked at Bayer, the introduction of Climate FieldView for Channel seed customers led to increased share of the seed basket, higher lifetime value through brand loyalty, and more than offset the cost of providing the software to those customers.

Due to this standalone software FMS companies have struggled, as they didn’t have anything to sell other than software to recoup their investments in software, and customer acquisition/retention. Famously, Farmers Edge has crashed and burned (They had many other issues. Please read Shane Thomas’s excellent analysis on Farmers Edge)

Having said that, many improvements have been made in the data capture experience, organizations have learnt some lessons on how to get good quality data. Janette rightly concludes,

I believe markets and solutions often go through many iterations, and pivots before they figure out the right business and deployment model. Even though the first iteration of FMS startups and applications did not fare well, it has led to an inflow of technical talent to the space, changed the thought process for many agriculture companies about the role of technology in their digital transformation, and the lessons learnt have been included in the new models.

The FMS applications have provided a baseline of data for other applications along the supply chain which would not have been possible without farm level information.

For example, Bushel acquired Farmlogs to help connect post-harvest workflows with farm level information, CPG companies, and grocery retailers can help understand farm level practices in their sourcing, crop insurance companies can improve the entire crop insurance workflow of reporting, policy management, and claims adjustment much more efficiently, equipment dealers and OEMs can service their customers better, offtakers can source grain based on certain farming practices, and many other applications. So even though standalone FMS investments might not have recouped their money, it has set the baseline for many applications in the future.

We have an interesting future ahead of us, with many more lessons to learn, and many more mistakes (hopefully new ones) to make.

Smile curves

J. Matthew Pryor and Sarah Nolet of Tenacious Ventures are two of the foremost thinkers in the Agtech space. Tenacious Ventures has published a podcast episode on Machine Learning as a Service within agriculture and AgTech. Do give it a listen.

As part of the analysis, Matthew Pryor published a smile curve by drawing on some of the analysis done by technology analyst Ben Thompson (Stratechery fame) in 2014.

​
​
​

The smile curve does a very nice job of addressing the question of if agriculture AI has a need and a market or not. You should read the Twitter thread posted by Matthew.

Matthew has been a proponent of the “factory (farm) has no roof” and the hypothesis that remote sensing will be able to solve/address 80% of the questions at the farm and field level.

This requires significant investments to capture high quality ground truth data at scale, train machine learning models, and deploy them globally to provide low cost actionable insights. An example of this would be to provide a crop type model at the field level, by using remote sensed data. Due to this companies, which can procure data at scale, and turn them into low cost insights will be able to provide a differentiated product.

This is easier said than done and requires significant continued investment, as I had pointed out in edition 89. (Jan 16, 2022)​

Companies still need to invest in significant resources to clean up the roof, and also to collect high quality ground truth data (or clean it) coming from precision agriculture equipment, and other sensor types with different data modalities.

If this clean data is provided to other innovators, and customers, value can be captured in terms of efficiency improvements in building models and deploying solutions. Similar to the earlier discussion about value capture, attribution to value created to these models will be challenging and ML service providers will have to come up with creative solutions to do so.

The other side of the smile curve, requires high fidelity ground truth data to provide localized high impact solutions. This could see the prevalence of many small players which provide point solutions, or a few providers with an efficient engagement and delivery model, which includes human support and expertise to help you handle your slightly unique situation, coupled with flexible yet simple pricing models.

As Matt points out the differentiation, and defensibility on the left will come from business and delivery models, and not entirely from technology differentiation.

​Four investors on What’s next in Food Tech and AgTech

1. Agriculture without reducing soil quality

This feels obvious. Technologies like automation, smaller equipment, biological products, and market interventions like incentives for sustainable agriculture will help with this trend.

2. Differentiated plant-based products

Even though plant-based products like plant-based-meats have taken a beating, these companies need to innovate more to provide plant based products which consumers want and at a price point consumers will be excited about.

3. Regenerative farming techniques

VC Isabella Fantini thinks “investments will go toward technologies that are a little "less shiny" meaning there will be more interest in upstream technologies that are closer to the farmer & through the supply chain, rather than downstream toward the consume.”

Regenerative farming techniques like no-till have seen rapid adoption, though use of cover crops has lagged. Cover crop efficacy is very context dependent, and cover crop adoption will grow only in situations where it provides a clear incentive to adopt, or economic, agronomic, or environmental benefits on a reasonable time horizon.

4. Food as Medicine

I personally detest the idea of “food as medicine.” Food is meant to nourish us. Food is a huge part of our culture, and identity. Food is a connector and a source of joy, when you have the right kinds of food, in the right quantities, at the right price.

I don’t like the reductive idea of food as medicine. Just like medicine, it takes on the meaning of trying to fix something which is broken.

We should celebrate food in all its glory, and work towards making nutritious, high quality food available to everyone in the world. Anything else will be a difficult (medicine) pill to swallow.

In the News

AgTech and Agronomy

Last week’s newsletter was featured on the “Precision Farming Dealer’s Best of Web” for the week of Fe 1, 2023. They have featured editions of this newsletter in their best of the web section a few times now.

“Newer IoT technology is designed to be easier to install, lower cost, more reliable, and with more straightforward configuration and setup, making it a cost-effective solution for the agriculture industry, which is under constant price pressure.” Good write up on how IoT is impacting multiple industries, including agriculture.

I always love it when people tell stories using comics. The organic seed alliance tells the stories of Mexican corn adaptation using comics.

​Bayer and Kimitec form a strategic partnership to accelerate the development of biologicals. “The biologicals market is expected to grow to nearly €25 billion by 20281, as consumers’ demand for low- and no-residue food products, and retailer food sourcing standards drive growers to look for new innovations in crop protection. Since 2007, Kimitec has been developing effective biologicals that provide effective alternatives and complementary options to synthetic crop protection, aligning them well with Bayer’s strategy to provide best-in-class solutions for growers through the integration of future and existing technologies.”

Robotics and Automation

​Tech makes farmworker’s jobs safer. “The most accurate science fiction tale is actually “The Jetsons.” No, automated harvesters won’t be wearing maid uniforms and carrying feather dusters like Rosey the Robot. But they will make sure that farm employees have a safer, more efficient, less physically rigorous way to do their jobs, all with the potential for greater career development.”

​Agtonomy unveils telefarmer solution for specialty crop farmers to empower farmers to remotely execute labor-intensive field tasks such as weeding, spraying, mowing, and transporting.

Sustainability

​California records double digit decline in statewide pesticide use! “In 2021, almonds, wine grapes, oranges, tangerines, strawberries, table grapes and raisin grapes were treated with the most total pounds of pesticides. Crops that decreased overall pounds of pesticide application from 2020 to 2021 include almonds, wine grapes, table grapes, raisin grapes and walnuts.”

​Turning camelina into a renewable fuel source. I have been excited about camelina for some time now, and it is great to see companies running experiments to test its efficacy as a rotation crop. “Camelina acres are tiny compared to corn, soybeans, and wheat, but Kevin Monk, Vice President of Ag Technology for Sustainable Oils, says the company knows it’s not going to replace those row crops. Instead, Sustainable Oils is looking for ways to partner with those crops by targeting two main growing periods: those following wheat-fallow rotations (where camelina could be planted in place of fallow) and those double crop areas that would benefit from a short-season, overwinter crop – both of which are ideal conditions for camelina.”

​Scoular’s farm team gives producers a voice on sustainability. This is a very prudent, and thoughtful approach by Scoular to form an advisory council of 15 producers from across the Scoular origination network.

Smallholder

​Yara India reaches 11.8 million farmers digitally (Yara’s sustainability report) “The sustainability report centers on the 5C approach to sustainability at Yara: Commit, Channelize, Care, Concern, and Contribute.”

What do you think?

💗 If you like “Software is feeding the world”, please share with a friend.

🙏 If you don’t mind answering 3 questions anonymously (2 are optional), I would love to get your feedback.

About me

My name is Rhishi Pethe. I lead the product management and technology delivery teams at Mineral, an Alphabet company. The views expressed in this newsletter are my personal opinions.