After only four months Autogrow’s FarmRoad platform has developed several new techniques resulting in an additional five percent improvement, with some growers hitting 95% yield accuracy.
“A five percent improvement may not sound big however when you are talking about scale that equals hundreds of thousands of dollars for enterprise sized growers, five percent is a significant improvement,” explains Chief Technology Officer Jonathan Morgan.
“We currently have enterprise scale growers in Mexico, Australia and New Zealand all in the 95-percentile prediction range. This improvement happened faster than we anticipated but it’s been down to the rich data that growers have been providing us and the models we have created.”
Autogrow has also expanded their models to include two proprietary tomato cultivars alongside Clodano by Syngenta and Endeavour by Rijk Zwaan. This adds to the work already done with Marnax by Axia Seeds, Maxeza by Enza Zaden and Merlice by De Ruiter Seeds.
FarmRoad is an enterprise class platform but lower tech growers can also take advantage of yield prediction. Head of Product Marketing Sophie Stanley notes that businesses are looking for a complete solution and yield prediction is the missing link.
“If you’re in a greenhouse relying on manual controls you can still utilize Yield Prediction by FarmRoad. The effects and the models will be different to a grower in a hi-tech environment but with enough data you can achieve the same accuracy. Our experience across multiple growing types and levels of technology enables us to point out data that may be missing and help improve data collection.”
“Accurate yield prediction gives growers stronger relationships with their own customers as they are able to provide surety of supply and deliver contracts on time. It’s really the complete seed to sale cycle and we are leading the way.”
“Our success though is down to the openness of our growers to be part of the process. Growers can be reluctant to share data like crops, yield and price to external parties. However, there is acknowledgement that, to get the benefit of new A.I technologies, you need to provide the data. The focus then is on the commercial agreement and being comfortable that their information will be secure,” says Ms Stanley.
As the concept of yield prediction is relatively new within Controlled Environment Agriculture, Autogrow warns that some vendor promises may be more smoke and mirrors than data science.
“What we have learnt from this process is there is definitely no one-size-fits-all. I would say that anyone who says they have one solution that works for everyone would be hugely exaggerating. There are too many variables across growers, cultivars, growing methods, management, and geographical locations to ever come up with one model for all,” says Mr Morgan.
“The best thing for anyone interested in yield prediction is to have the conversation with us, be part of a trial, and judge the success rate for yourself.”