Mistake Monday: Crop Estimation Is Pure Luck

When I worked in California, I was told to estimate yields (aka crop estimation). The procedure of counting clusters, weighing those clusters, and calculating an estimate of yield. I used to do this, but I don’t anymore. Because doing that and being right is pure luck. Don’t believe me? Here’s how the story goes:

I operated a vineyard that had a 30-acre block of split-canopy Chardonnay. This one block was almost 10% of that ranch. The spacing was 8′ rows and 4′ vines, so 1,361 vines per acre. A little over 40,000 vines total in this one block. Pretty important to get the yield estimate right, right?!?! I decided to follow all the crop estimation procedures and impress EVERYONE with my ability to be right. It’s just math and I’m great at math! I spent 4 days, 40 hours worth of work counting the clusters on 1,000 vines, and taking cluster weights from 500 of those vines. I worked my ass off to make sure I counted everything properly. I did this at lag phase because that’s what the book said. I took them from all positions on the vine because that’s what the book said.

AND? My estimates showed there should be 6.23 tons/acre. I triple-checked all the numbers. I had it LOCKED DOWN!!! 6.23 tons/acre baby!!! I got it!!!

During the season, we had meetings to discuss our yield expectations in the vineyards. The winery depended on those estimates. During several of those meetings, I was specifically asked on multiple occasions if I had done my check and if I felt confident in my estimate? It was a big meeting with various people attending so I said “yes” with supreme confidence. Though, as the younger me, I was tempted to say F*&K YEAH I AM!!! I was fully confident.

But guess what? Confidence has absolutely nothing to do with grape yields. Neither do hopes, fears, dreams, or nightmares. Unfortunately, I was setting myself up for a nightmare. Little did I know, my boss had visited this vineyard a couple times and looked at this block. He new my number was wrong. How wrong? WAY wrong. I just didn’t know it.

When we picked the block there was not 6.23 tons/acre there. Nope, there was 10.17 tons/acre there. I was off by 63.2%. And yes, that is the right math, not 40% like it looks like at first glance. And therein, lies the nugget of truth about crop estimation. Fortunately, I was able to learn from my mistakes. That 30-acre block was harvested over a 3 week period so I had the opportunity to revise my estimate on 6 different occasions. All of which were wrong, and became a running joke. “Geez, Daniel, will you get the right number before you pick all the block or will you pick all the block before you get the right number?” A great time was had by all. Except me.

See, I didn’t grow up in grapes. I didn’t have 15 years experience like so many around me. I wasn’t a third generation grower or winemaker. I was just doing what the book said to do. How in the heck could I be so wrong? WTF? Well… math and statistics is WTF. I wound up underestimating that one block by 120 tons because of math and statistics. Fortunately, I’ve had a lot of math and a lot of statistics in my day and wound up getting the last laugh. Two years later, my crop estimation was the most accurate. I didn’t even count or weigh a single cluster that year. How about them apples?!?

Math and statistics is how I did that. AND, math and statistics are why I don’t count clusters and weigh them anymore unless requested by a client. It is a recipe for being constantly wrong. It is a parlor trick of the mind with some numbers layered on to make it seem like it should work.

With crop estimates, you have two things that need to happen to be right. You must be accurate AND you must be precise. Just one of those is not enough. That means to estimate “right”, you need to have a large enough sample (accuracy) and a uniform enough block (precision), which in statistics, means you need to sample at least 5% of a vineyard to be precise, assuming the vineyard is uniform. If the vineyard is not uniform, you’ll have to sample 10% or 20% to get close to the real yield. That 30-acre block? I would have had to “sample” 10% of the block to get the yield estimate to a range of 9.5 to 10.5 tons/ac. I’m not going to sample 10% of the vineyard. No way in hell. I cluster counted 2.5% of the vines in that block and I missed by 63%. I missed by 5 entire semi-truck loads. Most folks sample 0.1% or maybe 0.5% of a vineyard. How will that work? It won’t. And it doesn’t. It’s a mind trick that you think you’re doing okay on your yield estimates, even though you’re not. Here’s why:

Most vineyard yield ranges are between 3 and 8 tons/ac, right? It won’t be -15 and it won’t be 72. There’s a finite range. If you have 6 varieties in your vineyard, you could write down the numbers 3,4,5,6,7,8 and put them in a hat and randomly draw numbers. And guess what would happen? You would be very close on 2, way off on 2, and kind of close on 2. And I’d be willing to bet dollars to donuts that if you complete crop estimates of any kind where you count and weigh clusters, that’s the result you get. Right?!? You do a ton of work and the accuracy is no better than drawing numbers from a hat because both methods are really just broad guesses. Sucks right?!? I know the feeling

But it’s true. Because you have 6 varieties, you have 6 chances to be at least a little correct and only 6 possible tonnages. But if you do the math for those estimates, it is entirely possible that you would be MORE wrong doing that than just plain guessing. How about THEM apples?

So how in the heck did I get better at my estimates than guys who’d been doing this forever? Simple. History, averages, and listening. I used math and statistics in my favor for a change. I reviewed records of previous harvests. All I had to do was average the last 3 harvests for low variance varieties and the last 5 harvests for high variance varieties. That was my starting estimate. Three and five-year averages. Then I simply listened to my buddies (the ones with 15 years experience) say that the crop looked normal, light, or heavy and I adjusted up 20%, down 20%, or left it alone. That’s all I did. It took me 60 minutes.

“OH, but DANIEL, I don’t have 3 and 5 year historical data points”, you might say. “What then, Mr. Smarty Pants?” I’d tell you this: Assume your vineyard will yield between 3 and 7 tons/acre, depending on variety. Then walk 10 minutes in each variety and rank them from lightest to heaviest, and simply slot them in from 3 to 7 tons/ac. That’s your start. Then just listen to what others say about their vineyards, show a couple people with experience your vineyard and see what they think, and adjust up or down by 20%, accordingly. If you get a hail or a frost or a freeze, all bets are off. Nobody, even guys like me, will be able to precisely tell you what the yield is after a hail until we pick it. I’ll bet you some 10.17 ton/acre split canopy Chardonnay in North Coast California that you’ll be as close or closer than anybody who uses the “count and calculate” method.

Think of what you can do with all your free time now!?! You could take a nap. You could take a vacation. You could take extra tissue samples. Heck, pull some of those weeds. But whatever you do, RESIST the urge to think that counting 0.25% of your clusters will give you an accurate crop estimate for the other 99.75% of your crop.

Because here’s the cold-hard truth: The largest vineyard owners and wineries in California are, at best, +/- 10% on yield estimates. I’ve worked with them and run them. Most of the time, they are about +/- 15% accurate and sometimes +/- 20%. These are literally million and billion dollar companies. With hordes of employees doing their absolute best to calculate good yield estimates.

This brings me to the final point: Does it seem really hard to crop estimate a High Plains vineyard? A vineyard that has who knows how many missing vines, maybe a freeze, maybe a little hail, some nutrition deficits, and maybe some plugged drip tubes? Well, it is hard. It’s really freakin’ hard.

I have learned to find a range, like 4.5 to 5.5 tons/ac, and be confident in that by walking a lot of acres. It’s a much better use of my time. Because a guess is a guess. Even if that guess involves math and spreadsheets. You can count or not count. But you will miss either way. I mostly miss by a lot more when I count and weigh. So why even count and weigh?

Sure, is this a little hyperbole? Yes. You should have a rough idea of the tonnage you will be harvesting. I am not suggesting that you blow it all off and just say any random number that pops into your head but, until we have a robot that can scan large amounts of the vineyard, then it’s just a huge guess; a very rough idea. And it is silly to expect precision and accuracy from a method that is 6 parts luck and 1 part science. So how about we stop pretending that there is a tried and true way to do crop estimation and we focus on ranges?

The alternative is just seeing what it weighs when we pick it because that’s both 100% accurate AND 100% precise!