“I need a way to make sure my predictions are more accurate… Can you help with that?”
It was a fascinating question, but at that moment, Scott Dickson Dagondon, Motorleaf’s Director of AI, had no idea that it was about to transform the company.
That question came in Las Vegas, at Indoor AgCon in 2017. The Motorleaf team was manning the company’s booth when a grower from a large cooperative California approached them. Motorleaf had been founded a year earlier by Alastair Monk and Ramen Dutta as a to develop hardware and software to help indoor farms monitor, control, predict and record every aspect of controlled environment agriculture. “We realized the industry was drowning in data, and we were challenged with the task to make use of what all these sensors were already recording”, says Scott. But that grower at AgCon was going to send the company in new directions.
He had a problem and he was hoping someone at the trade show would have the answer. Although he had years of experience growing tomatoes and had mountains of data gathered from various sensors that has been recorded over the years, his yield predictions were never quite accurate enough. No matter how good he was at doing the math, the actual yield was either significantly under or over what was expected by thousands of kilograms, and he was all too aware of the impacts.
“When we’re short, we lose money having to buy at high prices to cover what’s missing”, he explained. “When we’re long, we waste money on unnecessary labour costs, have to scramble to find a buyer fast and usually sell for a lower price than we could have earned if we had been able to more accurately predict our yield”.
The losses were likely in the tens of thousands of dollars a year.
Then came the request. “I need a way to make sure my predictions are more accurate, Can you help with that?”
As head of AI at Motorleaf, this was a challenge Scott couldn’t imagine turning down. With a Master’s Degree in Electrical Engineering from McGill University and expertise in AI and machine learning, Scott began working on how to combine data science with plant science and develop a brand new way to forecast harvest for greenhouses. Scott pored over journals, studies and research articles on horticulture, tomato growth and harvest forecasting, learning everything he could about this unfamiliar science.
It was both inspiring and eye-opening”, Scott says. “I gained a real appreciation for how much attention these plants need, the huge amount of time and effort being spent on predicting yield, and how important it was to help growers when millions of dollars had been invested to launch their operation”.
Soon it was clear to Scott why predicting yield accurately was so tough. “With so many conditions in a greenhouse changing over time – temperature, humidity, C02 levels, radiation – there are potentially millions of data points that need to be analyzed to determine the precise factors that impact yield”, Scott explains. “Solving the yield prediction problem takes some powerful computing that could parse through years of data, detect patterns and unlock the critical factors that influence yield for a particular greenhouse, location and strain. To me, that meant AI”.
Over the next six-months, the Motorleaf team worked closely with the grower, examining years of greenhouse data and comparing it to insights from horticultural research to develop an algorithm that could act as a growers assistant, predicting yield more accurately than any human could, and freeing growers to spend time on more important work. When live trials began, after a few adjustments, the results were in. On average, the resulting predictions were 94% accurate.
The impacts were just as impressive. “With these predictions, we were able to set ads three weeks in advance and earn a higher price for our produce”, the grower stated. “We also almost entirely eliminated needing to buy produce at inflated prices to cover shortages and were able reduce our labour and packaging costs. The ROI has been superb”.
Since those trials, greenhouse growers from around the world have begun benefiting from Motorleaf’s automated AI harvest forecasts. With clients in Canada, the US, North Africa, Asia and Europe, Motorleaf is now working to explain its services into other types of produce.
“Helping growers get the most out of their investment has been incredibly rewarding”, Scott adds. “We are already working on new applications of AI to help commercial greenhouse growers further optimize their operations and can’t wait to see the difference our AI will make for growers around the world”.