Under the Hood at FarmLogs: Creating Crop Health Alerts

January 20, 2016

It was the week before Thanksgiving in 2014. Just outside of the historic building where our team works—where lumber and seed were sold more than a century ago, and where the farmers market still runs year-round—the weather had finally dropped below freezing. The trees were empty, the ground hard.

Inside, surrounded by the exposed brick and beams of our loft-style work space, Jesse Vollmar, 5th generation farmer and FarmLogs CEO, was sitting down with a small team to discuss the future of FarmLogs.

He read from a list of goals he wanted to accomplish in the next year. Included on that list: develop a tool that would help growers find yield threats sooner.

At that time, I hadn’t yet joined the FarmLogs team as the company storyteller.

“We’re a company that’s focused on solving real problems for growers,” Jesse reiterated to me in an informal interview almost a year later.

We were sitting on the couches in an area I refer to as the company living room: a place where there’s a constant rotation of engineers and programmers and designers sitting and working throughout the day to touch base on various projects. Jesse and I were meeting because we agreed there’d be value in sharing the Crop Health Alert origin story.

What goes into creating such a product? Why should you trust it?

Of course, every good story starts with a why.

So why did we create Crop Health Alerts?

“It was time to make scouting easier and more efficient,” Jesse said.

“For the first time ever, there was enough historical satellite imagery to build an intelligent, streamlined approach to scouting.”

“As it is, scouting takes time and energy, and it’s hard to walk every field. Problems get missed, and ultimately that costs our farmers money. We wanted to build something that would help them save time and find and fix more problems.”

make-something-not-blurry.jpeg

This sign from Y Combinator hangs in the FarmLogs office to remind the team of the company’s farmer-first approach to business.

Just three short months after Jesse announced his initial idea for a new scouting tool, the FarmLogs team created and launched Crop Health Alerts.

A few months after that, word began to spread about the success of the new technology.

Screen_Shot_2016-01-20_at_3.25.25_PM.png

By optimizing multiple years of field-specific, historical satellite imagery, the new feature automatically sends FarmLogs users crop health alerts when it detects unusual behavior in their fields. The FarmLogs app then guides the farmer to the exact location of the problem spot for a more efficient scouting experience. In this image, one user let us know on Twitter (find us at @farmlogs!) that a crop health alert led to the discovery of a grasshopper infestation.

Today, Crop Health Alerts continue to help growers across the country scout more efficiently and detect yield threats sooner. 

Read on to learn more about what happened behind the scenes as Crop Health Alerts were developed.

 

Working Through the Unknowns of a Bold Idea

On reflecting about the development process, Jesse admitted that there were a lot of things we didn’t know going into it.

“To start—for us to build this tool—we knew we would need satellite imagery. At that point, we weren’t using satellite imagery at all at FarmLogs. Going into product development we didn't know how we were going to get access to a large-scale imagery provider. We also didn't know how much it would cost to acquire that imagery, or if it would be feasible to acquire it.”

We would need a very special partnership, Jesse told me, because out-of-the-box pricing for such large-scale data would be exorbitant and most-likely well outside our budget as a small company.

“Next—even if we did have the satellite imagery, how were we going to use it to extract valuable insights about our farmers’ fields and then deliver those insights to our farmers?”

“We knew that just showing images or vegetative indices wasn't going to cut it.”

“We had to build something powerful yet simple to use. It had to tell you when you had a problem and where the problem was. No one had ever built a product that could actually pinpoint a problem in a field. We'd have to define that and figure out how it could be done.”

From a behavioral standpoint, beginning stages of the development process also raised some interesting questions.

John Shriver, the former Lead Data Scientist at FarmLogs, was our go-to guy for anything imagery-related at that time. He, too, was in that initial November meeting about the future of FarmLogs. 

“For many, scouting is a pretty set behavior. It’s a lot easier to just scout your fields and make decisions through the window of your truck,” John said. “And why not? It doesn’t make sense to walk every field looking for problems, especially if you don’t know what you’re looking for. It’s just not efficient.”

Our new app would require folks to get out of the truck. We’d guide them to where they needed to go, but would they trust the app enough to take action—to modify a set behavior and create a new habit—on the new intelligence the app would provide?

“We had to trust they would,” John said. Because that—I would discover after talking to members of the crop health dev team—is the the nature of innovation.

To make big leaps forward in any industry, sometimes you have to supplement the user’s need with the innovator’s drive and instinct.

“So we went for it,” Jesse said. “We had enough scientific research to support the decision to move forward, even though it would need to be done in 3 months—a timeframe that most companies wouldn’t be able to deliver on.”

 

Getting Started with Satellite Data 

As the former Lead Data Scientist, John Shriver was charged with collecting and analyzing new data, determining what it means, spotting trends, and recommending ways to apply it to FarmLogs’ product offerings.

“In our search for imagery, we talked to aerial providers and drone providers, too. But nobody had the data we needed,” John said.

john-at-work.jpg

John Shriver, the former Lead Data Scientist for FarmLogs. Shirt reads: “Stand back. I’m going to try science.”

Not initially, anyway. But then the team discovered Planet—the satellite company that FarmLogs still uses today.

“It was a big moment,” Jesse said. “We were able to negotiate a special contract with Planet to get imagery for a large portion of the country.”

Normally growers don’t use satellite imagery to make field decisions, because obtaining imagery is cost prohibitive.

“Because of our contract with Planet, we would now be able to provide satellite imagery and insights to farmers that they can’t get on their own.”  

“We found Planet, which was great,” John said. “But remember that we were asking for something literally no one else had ever asked for before. That data we needed—six years of historical, high-res imagery—would have to come from their archive.”

And getting that imagery would take time.

Processing the archive imagery took 30 servers crunching data 24/7 for an entire month.

Keep in mind: Jesse had challenged the team to build Crop Health Alerts in three months. Only two months remained to meet the goal.

“Once we got access to Planet’s archive, we needed to build a data pipeline,” John said. “We designed the algorithm, loaded it into high-performance cloud servers, and began uploading the processed WDRVI images. We also built a local database to store the processed dataset so that we could efficiently query for specific, field-level images."

After that, John spent much of his time working on the algorithms that would address some of these questions:

  • “How do we look at all these images of one field and actually detect change in a field in a way that’s reliable? How do we find the best signal amongst all the noise?”
  • "What threshold of change do we need to see between images to notify a user? Which times of the year is this going to be most relevant to our farmers?"

The team also knew they’d have to apply a new system of colors to the images they’d provide to users.

John said: “But color has meaning. So if everything is red and green, does green mean good and red mean bad?”

Not necessarily.

Quality control was also top of their minds. “These images have clouds in them,” John said. “How do we know the images are good enough to deliver to a user?”

Ultimately, to answer that last question, they built a cloud filter.

“People spend their lives building cloud filters,” John said. “So for us to say we’re going to build a cloud filter in three months—that was huge for us.”  

Regarding his initial questions—about triggers, relevancy, and colors—he’d need to work closely with the User Experience Design team to find the right answers.

 

Creating an Actionable and Intuitive User Experience (UX)

“We knew we were going to get these images, and we knew we were going to have to show crop health in some way,” said Sam Pierce Lolla, Director of Product Design at FarmLogs.

Sam’s job is to enhance user satisfaction with FarmLogs products by focusing on usability, accessibility, and creating delightful experiences.  

Sam, too, was in that initial November meeting with Jesse.

“John and I had to work through a lot of high-level concepting and abstract design processes to get to the final product,” he said.

sam-at-work.jpg

Sam Pierce Lolla, Lead User Experience Designer, working through Crop Health Alert user design sketches.

To start developing the user experience and the user interface, Sam and John worked closely to develop two or three strong use cases.

A use case identifies a need and then a series of possible interactions between a user and a system that allows the user to achieve a goal.  Workflows are then developed based on possible use cases.

It’s a critical first step to any product development process.

something-is-wrong.jpg

A high-level workflow developed early in the process.

 

event-vs-stage.jpg

An early sketch composed of Post-its that explores the ways users might be notified about crop health alerts.

 

above-the-line.jpg

Pictured above: a workflow map that helped the team focus on the interactions that would help them meet their goal.

Sam said of John: “We went into a room together and had a bunch of ideas, and we built patterns of how the design [of the new feature] might work.”

“We posted those patterns on the wall and tried to arrange them into stories. We wanted to say, okay, if this were to happen, then this is what our users would see, and then they would see this, and then this. It was important that every story we ran through started with someone’s need. And it needed to end with something useful.”

IMG_2880.jpg

Pictured above: one of four images that captures Sam and John working through various use cases and workflows.

john-hanging-images.jpg

 

john-pointing.jpg

 

sam-john-thinking.jpg

Moving through the use case process helped them better define the new feature’s functionality. It also helped them find answers to their most critical questions.

“One fundamental question we thought about a lot at the beginning was what is a crop health alert? Is it something that goes away? Can you clear it? If you clear it, can you not see it anymore?,” Sam asked. 

“To answer that question, we had to ask if an alert is an event, or if it's something that’s in an active state? An event is a single point in time.” It’s like snapping your fingers once. There’s what happened before you snapped your fingers and what happened after.

A state has a defined start point and an end point. It’s like harvesting a field. There’s what you did while you’re in the combine, and what you did before and after you’re in the combine.

It was an important question to answer, because if a field could be “under a state of alert,” how long would it persist?

Once the team determined that a Crop Health Alert was in fact an event, rather than a state, they could plan for how users would interact with a crop health alert in the app and via email.

iconography.jpg

Pictured above: various icons the team explored as they designed the app interface. Knowing that crop health alerts signaled an event, rather than an ongoing state, helped them adopt the yellow triangle to notify users of potential yield threats.

 

near-final-design.jpg

Pictured above: a sketch that closely represents the final Crop Health Alert app user interface.

Of course, it would be remiss not to mention the engineering team in all of this. Front-end, back-end, iOS, Android: our diverse team of engineers needed to be able to take John’s models and Sam’s design and translate it into something that would run into production.

And translate it they did. In three months’ time, the entire FarmLogs crew came together to launch a new feature that did exactly what it was supposed to: alert farmers when something was happening in their field that they couldn’t see from the road.

crop-health-alert.png

Pictured above: the mobile user interface for Crop Health Alerts as of November 2015. As a user,  you can see all of your fields in one place. A yellow triangle indicates which field has an alert.

 

mobile-ui-crop-health-jpg.png

Pictured here is the mobile interface users encounter after tapping a field with an alert on it. As a user, you can see and find (via GPS) exactly where the threat is located. Since launch, updates to the technology and the UI have continued to be made based on feedback we’ve received over the 2015 free trial of the feature.

 

A Success Story that Started at Home

After Crop Health Alerts launched, one of the team’s first big feel-good moments happened when Jesse’s father, a 4th generation farmer, received an alert on his own corn field.

They were at dinner when the alert came through.

Jesse asked his dad if he’d been out in the field lately.

His father explained that he’d been there just a few days ago. He didn’t know why they’d be getting an alert in that field.

After dinner, Jesse decided to drive out to the field to take a look on his own.

“From the road the field looked fine. My dad was right,” Jesse said.

“I couldn’t see that the field was showing any stress. But then I opened the app and I walked into the field and, sure enough, as soon as I got through the first few rows of tall corn, I entered into an area where the corn was only 1/3 the height of the rest of the corn. It was yellow and sick.”

As it turns out? The area had a big nitrogen deficiency problem.

“There it was, this massive problem in our fields that we couldn’t see from the road,” Jesse said. “My dad is a great farmer and I have a lot of respect for him. But even he wouldn't have spotted this problem without an alert.”

 

Want to know even more about Crop Health Monitoring?

You might be interested in these short support articles: 

What makes FarmLogs' crop health images special?

What do the colors in my crop health map represent?

When should I receive crop health images?

 

New Call-to-action

Ashley Rutter

Author

Ashley Rutter is the Content Marketing Writer at FarmLogs.