- A vertically integrated snack food company sources potatoes from multiple growers spread across different growing regions contracted before the growing season for their potato-based products.
- Due to a severe and lengthy shortage of potatoes in the European market along with a spike in consumer demand, the potato inventory and availability of developed supply could not support the production requirements.
- The company was in need of sourcing potatoes from their growers before the growing season was completed, but did not know how to choose which growers and plots should be prioritized for early harvest to minimize the negative effect on both yield and quality. Harvesting plots that could have continued to increase in yield and quality instead of those that had reached, or were close to reaching, their prime would result in a direct decrease in production.
- Trellis supplied the company with their AI-powered, plot-level yield and quality Prediction Simulator to explore and compare different “what-if” harvest timing scenarios.
- In real-time, the Trellis Prediction Simulator enabled the customer to compare the effects of varying harvest timings on the production line outcome. This simulation accounted for both optimization of required supply and inventory, as well as cost of production with fluctuating quality metric levels.
- Because the cost of production is strongly influenced by dry matter content in the potatoes, the customer was interested in harvesting the plots that had not only reached their optimum yield, but also their ideal dry matter levels to minimize operational and production costs.
- Leveraging the Trellis Prediction Simulator, the customer was able to identify plots that were predicted to be of both optimum yield and dry matter content, and prioritized their supply with those plots being supplied to the factory first.
- Plots that initially were not of optimum yield and quality were harvested weeks later, and by that point they had reached the necessary growth maturity to be harvested as usual.
- Had the snack food company not supplied their production line with the optimum Trellis recommended simulation, their production costs for those plots would have been 7% more, yield would have been 12% less (28t/ha to 25t/ha), and there would be an increase of 5% waste due to grade b potatoes.