Issue link: http://winesandvines.uberflip.com/i/63937
WINEMAKING the results This is where things begin to get com- plicated. It's clear from a first glance at the data on total and free SO2 levels that there is quite a bit of variation among bottles of the same wine. But how can this be quantified? And is this an acceptable level or a problem? To make sense of the variation found in the study, some state- ment had to be made about the level of consistency sought. In this case, it was de- cided for the chemical analysis that a suit- able quality standard would be that each wine should not vary from its own mean by more than 10%. (This is, of course, a theoretical level at which to set the qual- ity standard. It was decided by looking at the level of variation shown by the three least variable wines, which were taken as a benchmark. It can be debated whether or not this is an appropriate level.) It is then possible to do a simple sta- tistical test to get the sample coefficient of variation (the ratio of the standard deviation—a measure of how variable the results are—to the mean.) For TCA analysis, it is a bit more straightforward: The wine has to have a threshold below 0.5ng/liter in order to pass, and a single bottle of wine either passes or fails this test. Thus, it is possible to define whether or not a wine is in technical compliance with a defined standard. The sensory work is more complex to analyze, because even skilled tasters are not measuring devices, and they perform with a degree of inconsistency in sen- sory trials. Since many judges tasted the same wine in this trial, and judges differ, the statistician had to come up with a mathematical model in order to infer the presence or absence of bottle variation. The results of these analyses were very interesting—and actually quite worrying. Using a 95% confidence level (a standard statistical way of dealing with issues of significance), the study reveals that only 23% of the sampled wines are compliant with the technical quality standard, which allowed for a 10% deviation. That means that 77% of wines failed. Breaking these data up a little more we can see that the overwhelming majority (90%) of wines pass the total SO2 a mathematical correction for the small sample size in this study, the estimated TCA defect rate is quite high, with only 43% of wines passing the test of P = 0 tency test, while only 60% of wines pass a similar test for free SO2 quality-control consis- . After applying (probability = zero TCA) at 95% confi- dence. It is when these figures are com- bined that the final (and rather startling) data point is reached: Only 23% of wines meet the desired criterion that all techni- cal data are in compliance. sensory attributes For the sensory data, the bottle variation is quantified as the number of bottles that have a significant variance from the wine- specific mean value. That is, the average score on the sensory attributes for the wine is taken as the standard, and the study sees whether bottles of that wine differ signifi- cantly from this average or not. Using this measure, bottle variation is detected in 12% of all measured attributes. The sensory work is important, even though it is hard to do and difficult to analyze. It is all very well showing that bottles of the same wine vary, but a key question has to be: Does that variation in free SO2 translate into sensory differences that can be picked up by the consumer? If I were a brand owner, this is where I'd begin to worry. Certainly with TCA, supermarkets in the UK have reported a reduction in customer complaints in brands that have shifted from natural Luc & Jodie Morlet Phone: (707) 967-8690 ext. 111 Email: info@morletselection.com Wines & Vines MAY 2012 59