The evolution of smart sales forecasting for greenhouse

Added on 05 July 2022

Tags

The evolution of smart sales forecasting for greenhouse

When I speak with operators starting up new growing facilities for the first time, they often budget 5% to 10% variability in yield forecasting. For most operational greenhouses, forecasting variability is even larger than that. We've seen many growers struggle to get beyond 50% to 60% accurate with long-term forecasts.

Forecasting is not an easy task, though most assume it is. To accurately forecast, you must have good, clean data on harvested weights, understand your loss from stage to stage of growth, and factor in environmental and climate changes. And even then, we're combining two biological systems into one system (weather and plants), which means plants can often break expected behavior as well.

Once we have all this data, we can begin to forecast based on classical forecasting models - such as using moving averages on past data (shorter term for short turn crops and seasonally for long turn crops).

Often, however, we're not starting with all this data. Think about how you collect harvest data. Are you weighing a few trays or gutters to represent your entire lettuce batch? Or pulling sales data for potted plants? Or collecting fruit size information for one or two of your tomato plants and estimating across the entire crop?

You're not alone. This is how most facilities collect data today. To date, it has been costly and/or labor intensive to do more than this.