It’s that busy time of year again: pre-season supply planning. Also known as “a second harvest,” to VP Winemaking and Supply Planners. You’ve done this before – you take into account last year’s yields, your inventory position, forecasted sales, and you make some educated assumptions with hopes for a more balanced portfolio. Making these crucial business decisions for a finished product that will only hit the shelves three years from now is nearly impossible. With supply and demand fluctuating dramatically each year, you are like an archer trying to hit a moving target during an earthquake. This vintage it is especially challenging considering you and most of your competitors are drowning in more months of inventory than you could give away as holidays gifts if you tried. (Check out the State of the Wine Industry 2020 report to see how challenging things are across the market).
Leveraging your data in a smart way to cross this minefield of strategic planning is an extremely valuable resource. Accurate yield predictions early in the season help you find the best sourcing channels and purchase the right quantities at the price your company requires across your portfolio. When these stars align, you save millions of dollars in storage, COGS and sales, while reducing the months of undesired inventory moving forward.
An “as-is” historical average means 30% supply uncertainty
The idea of estimating your 2020 supply based on a historical three-year average should work well enough… in theory. We all know that everything regresses to the mean, and an average should provide stable enough ground to hit that moving target. And come on, how can you possibly be expected to provide an accurate yield forecast months before there are even leaves on the vines, let alone the impact of a potential September heat spike? Nevertheless, there are a few tricks you can keep up your sleeve that will serve you well in early season supply planning.
Every year counts
Firstly, because of the multitude of nuanced changes that occur over the years, businesses often utilize only a few years of historical data to estimate their supply. However, if you have recorded more than a few years of historical data, you should definitely leverage it for your planning. More years of data can improve your estimates by 15%, get more granular information, and help you draw insights that will elevate your ability to confidently make those tough decisions. Those tough years of sweat equity have earned their place in helping you improve your future!
Don’t neutralize …normalize!
What about the increased volatility that comes with including more years of data? Follow this advice, and you’ll easily clear that hurdle, while benefiting from all the advantages of the extra data. Let’s take, for example, a year or two that experienced extreme weather conditions, such as 2015. Don’t leave those years out, but don’t leave them “as-is” either, because that will completely throw off your estimates. Here is where normalizing the data comes in handy. For example, think of how a teacher might account for some exams being more difficult than others during a semester to provide a final score for a student.
Similarly, map the yields of those “outlier” years to a more typical scale to get them closer to the average, and then include them with the rest of your historical data in order to gain greater accuracy. Besides, a three year average where one of those years was an outlier (think 2018 vs. 2017) requires some normalization as well, and with much less data to ensure you’re getting a reliable estimate.
Measure your measurements
Another way to work with a multi-year dataset is to consider measurements and samples that were taken in previous seasons. Your team utilized that information during those vintages to estimate that particular year’s supply, but it can contribute to the coming year’s estimate as well. Considering historical cluster counts for a specific vine to quantify the volatility across years can be neutralized, and adding this data to your estimate can do wonders for your supply plan.
Happy Birthday, Vines!
New vines? Old vines? Depending on what percent of your portfolio are newly planted or old vines, your season estimates can be thrown off dramatically if you don’t remember to account for their development. For example, a multi-year average that the updated age of the vine will boost your accuracy, as vines produce 20-40% less in their first years and also begin to produce less as they age (after 25 years or so).
Right wine, wrong label
Considering more data means you can get more granular in your sourcing of optimal quantities to best suit your portfolio, whether by program, blend or grade. Hitting that nail on the head will reduce the amount of down blending you will need to do come harvest, which can cut your wine COGS dramatically. When using a limited data set, you don’t separate your estimates, but rather group all grape data into one average, which allows a lot of irrelevant data to sneak its way into this calculation and the most valuable insights to get lost.
If you’re looking for the best platform to help you get the most accurate, timely predictions throughout the season, taking into account proprietary databases, Trellis.ai is for you. Trellis delivers predictions, insights and recommendations that maximize crop yield, improve production efficiency.