Four years after spinning out of Google, Heritable Agriculture is moving from AI promise to a live commercial product.
The company has launched a web-based prediction platform designed to help seed companies, nurseries and breeders answer one of agriculture’s most expensive questions: which plant genetics are best suited to which fields, regions and management systems?
Farmers may be next.
“In April, we launched on the web a prediction platform that lets folks say, ‘Hey, I’ve got a bunch of seed varieties. Which one is the best for which area?’” says Tim Beissinger, CTO and co-founder of Heritable Agriculture.
How Heritable’s Prediction Engine Works
The platform combines:
- Historical weather data — roughly the past 50 years
- Soil models, without requiring soil tests
- Satellite imagery converted into digital site characteristics
- A customer’s proprietary performance data, such as yield or trial results
- Genetics-by-environment-by-management modeling
“With our tool, you get 10-meter accuracy down to your field, and you can pick based on matching the genetics to the field. Same thing with a corn plant, same thing with a tomato — you name it,” Beissinger says.
He describes the platform as a way to match genetics to geography.
“Given all of that information, we can figure out how genetics is interacting with the weather, soil, and site characteristics derived from satellites, and be able to say, for a new site, genetic variety A will perform this well, and genetic variety B will perform better or worse,” he says.
Primary Users Today
For now, the platform is primarily aimed at organizations that already have trial or performance data across multiple varieties and locations, such as seed companies, nurseries, breeding programs, specialty crop companies and forestry or tree genetics groups.
“We get the performance data from our customers,” Beissinger says. “We layer in their proprietary data, match it with weather, soil, and satellites, and then take it from there to predictions.”
A farmer could technically use the platform, but only if that farmer has historical data showing how multiple varieties performed across their own fields.
A Pay-Per-Prediction Model
The cost model is also new and notable. Beissinger says users can upload data and train a model at no cost; the cost comes when predictions are made–about $1 to $1.50 each. For example, if a farmer or seed company wanted predictions for 15 varieties on a new field, the cost would be about $20.
Crops Served
“Our biggest customer segment today is vegetables in open fields,” Beissinger says.
The platform currently supports a broad range of crops, including corn, rice, rye, barley, oats, soybean, sugar beet, potato, sunflower, sweet potato, canola, cassava, pulses, cotton, groundnut, winter wheat and millet.
Beissinger says Heritable is also interested in permanent crops, including apples and almonds. That category is especially compelling, he says, because growers are often locked into variety decisions for years.
Reducing the Trial-and-Error of Variety Placement
Heritable’s larger value proposition is helping companies reduce risk before committing to a variety in a given geography.
Instead of physically growing every candidate variety in a new region, companies can use Heritable’s predictions as a first-pass filter.
“Imagine you want to open up to a brand-new region. Say it’s a Central Valley company here in California growing grapes all across California. If they want to expand to France, in the past that would have meant trying all of their grape varieties in France,” Beissinger says. “Now we can say, ‘Of all the grapes they’re growing in California, which ones are most likely to perform well in France?’ Then they can go from there and take it to Chile, and so on.”
That’s where the speed comes in. Heritable can help companies narrow the list in minutes instead of waiting through a full growing season.
Partners and Projects
Heritable also has two sides to its business. In addition to the prediction and placement platform, Beissinger says the company continues to work on biotech and breeding discovery, including identifying genes worth pursuing.
“On the biotech side we work on identifying genes that are worth pursuing, designing for those genes, and doing more breeding and biotech, as opposed to prediction and placement,” he says.
Beissinger says Heritable has worked with KWS on the biotech side and with Syngenta on variety placement.
“We work with Syngenta on variety placement,” he says. “Much of our conversation is relevant to what we’ve done with Syngenta.”
The company is also part of a three-way partnership with Paul J. Mastronardi, Red Sun Farms in Ontario, and CIV, an Italian germplasm and strawberry company, focused on indoor strawberries in Canada.
“We are finding the right varieties quickly through these predictions — figuring out, without growing as many trials, which ones are performing and how well they’re performing in this environment with this management,” he says.
What’s Next: Farmer-Facing Tools
As for farmer-facing development, Heritable received a $5 million Gates Foundation grant to bring its technology to smallholder farmers in Africa. Unlike the company’s current commercial platform, that project is designed to put recommendations directly in farmers’ hands.
“We got a $5 million Gates Foundation grant to bring it to smallholder farmers in Africa,” Beissinger says.
For that project, Heritable is ingesting historical performance data for the varieties farmers are using, then making predictions based on location and management.
“They’ll be able to see, ‘Here are the 500 corn hybrids that might match your field. Based on what you’re doing and where you live, this is the hybrid that’s most likely to be a high yielder on your field itself,’” he says.
From Secret Tech Group to Commercial Company
Since its spinout from Google, Beissinger says the startup is acutely focused on customer value.
“We were a secret technology group,” Beissinger says. “Now we are a company that wants to be not just creating hypothetical value, but delivering actual value.”
