Australians have been planting grapevines since the 18th century. Today, grape growing is the largest fruit industry in Australia, with wine grape production the most predominant of the three viticulture industries, producing 1.7 million tonnes of grapes in 2014. Although traditional industries are diverging rapidly into a digital world, Australian grape producers are still plagued with the inefficiencies that come with doing things the old way. Even today, many grape growers are estimating their yield by recording bunch counts on paper while out in the vineyard, followed by doing some basic mathematics. For years, this has meant that predictions of annual grape production are out by up to 30 percent. However, with grape producers exiting the industry due to economic instability and wineries feeling the pinch, innovators are racing against the clock to revive the industry.
One startup looking to eliminate inefficiencies in the industry is GrapeBrain, a ‘viticultural intelligence system’ which leverages Microsoft’s Azure Machine Learning to significantly improve grape production forecasts. By feeding large data sets into Azure Machine Learning, GrapeBrain identifies patterns and insights from current and historical data, and uses what’s happened in the past to make predictions of what might occur in the future. Over time, yield estimates become more and more accurate. Added to that, the technology can be applied to other agricultural industries, which could mean significant improvements in global food supply.
GrapeBrain, a product of parent company Seer Insights, was founded by three students from the University of Adelaide. Harry Lucas, 20, is currently studying Mechatronic Engineering, while Liam Ellul, 23 is completing a combined degree in Commerce and Law. Petros Bakopoulos, 21, who joined the company later, is studying Mechanical Engineering and Finance.
The idea for GrapeBrain came about through Ellul and Lucas’s participation in the Tech eChallenge program, run by Adelaide University’s Entrepreneurship, Commercialisation and Innovation Centre and the School of Computer Science, in partnership with Microsoft. This was mid-2014. However, they wanted to take their concept outside the university environment and throw themselves into the world of real business.
Lucas said the original idea was to create drones for broadacre farming; the drones would have been designed to help grape producers identify problems with crops. But through market research, they learned that the most underserved problem for wineries and grape growers was the inability to accurately predict yield – that is, the number of grapes that will come off a vineyard at the end of the season.
“The unanimous feedback we received was that wineries don’t need drones, they need results. They’ve been gathering data for years and lost a lot of money due to problems surrounding a process called Yield Estimation,” said Lucas.
“We started looking into this interesting problem and we just got more and more engaged. It turns out that the logistical implications of not knowing how many grapes you’ll have at the end of the season are massive, some wineries lose over $200,000 per year here in the Barossa. We think its crazy that it’s not possible to do this in this day and age, so we set out to solve the problem.
The problem is costing Australia around 100 times this amount, not taking into account the challenges faced in other agricultural sectors. Lucas said the platform is applicable to other crops as well, and that the company has already been approached by bodies who administer these other crops.
“[We’re] excited to discuss our future potential collaboration with them,” said Lucas.
It may be difficult to comprehend how technology can predict grape production when there are various environmental factors that come into play, however Lucas insisted they adopt many different techniques to improve accuracy. As mentioned earlier, one major contributor is GrapeBrain’s use of the Azure Machine Learning system, which Lucas admitted has “opened up the whole field of data science to us”. GrapeBrain collects all available data that contributes to yield prediction, like weather and soil chemistry, and subjects that data to a predictive analytics process, which is part of the Azure Machine Learning system.
“The thing about modelling natural processes is that there’s many different factors which affect them. If we’re talking specifically about grape yields, there are a myriad of factors which impact a solid yield estimate, [such as] weather, soil chemistry and even the type of pruning used. That’s only three factors, and it certainly gets a lot more complex in the field,” Lucas explained.
“Our system certainly takes this data into account, but it goes far beyond this. The accuracy of our predictions depends on all of these factors as well as human input. With poor data comes poor results, as you can imagine.”
What’s powerful about Seer Insights’ first solution is not only that it can make sense of data, but enables grape producers to act upon that knowledge. They can manage their soil and water better and make other informed decisions to ensure the best outcome.
GrapeBrain is currently web-based and also available as a smartphone application. The smartphone application enables grape producers to do mandatory bunch counts when out in the vineyard, eliminating the need to use a pen and paper. The data is processed and sent to the cloud.
Ellul said GrapeBrain’s ideal customer is a corporate scale winery that has an international presence – this is strategic for the startup, as it allows GrapeBrain to scale through organisations. Ellul also admitted that they’re in the process of securing deals with some large organisations so they can start testing the platform this growing season.
“The organisations we’re testing with are actually our future customers and have been specifically selected due to their large overseas presences, allowing us to scale THROUGH organisations, rather than opening up entirely new sales channels overseas,” said Ellul.
GrapeBrain does not yet have a monetisation strategy set in stone. The startup will doing extensive testing before executing a business model.
“We’re still taking feedback on this, but we’re currently intending to charge in a fashion that scales with organisation size. It’s very important to us to make this software as accessible to smaller, independent customers as it is to corporate wineries,” said Bakopoulos.
The startup had been self-funded up until recently when it secured a grant from the Bank of South Australia – a much needed capital injection. Bakopoulos admitted that the last few months have been tough for the startup both development-wise and financially.
“We’ve been feeling the pressure trying to get our software into shape for roll out this year, and before we received the support from BankSA we were struggling to support the financial implications of running a startup. We’ve passed through that hurdle now and we’re excited to see our platform in use at the end of this year,” Bakopoulos said.
The co-founders are no longer actively looking for investment, however are open to the prospect of raising funds at a later stage depending on circumstances.
Lucas, Ellul and Bakopoulos have a lot to be proud of. Prior to being named as South Australia’s Young Innovators of the Year, they were personally invited by South Australian Premier Jay Weatherill to present to Cabinet. They were also invited to Seattle by Microsoft. However, what the co-founders are most proud of is the progress they’ve made with the product and the global implications of their solution.
“I’m very proud of the fact that we’re solving a problem which has some incredible future applications. Our software has the ability to be adopted to other crops, and I feel like this has some incredible applications in third world countries and feeding the world. I’m most proud of the fact that we have the ability to help people,” said Bakopoulos.
Lucas communicated a similar sentiment: “I’m most proud of the software that we’ve built. There’s nothing more satisfying than taking our platform out to a grower or winery and seeing how excited these people get to try our system out. I don’t think I’ve ever felt anything quite like it.”
Over the upcoming months, the co-founders will be finalising their initial product offering, testing the software and executing a sales strategy. They’re also eager to test their software in California’s Napa Valley before the end of next year.