QPRC 2016

Wine Blending Optimization with an Additional Constraint:  We Must Use All of the Grapes!

Daniel R. McCarville, Lisa Custer, Doug Montgomery, Chuck Bellante

Producing fine wines has similar manufacturing problems and opportunities as other product manufacturing industries.  Wine makers can certainly apply standard experimental strategies such as factorial experiments and mixture experiments to improve the quality and quantity of the products produced.  It is common to hear winemakers describe their wines with qualities such as nose, color, taste, fruitiness, finish, acidity, earthiness, and others.  Quantifying these measures for experimental purposes can provide some challenges.  Large producers of high quality wines can choose their best barrels of wine or their best vineyards to be included in their top wines, and use other barrels for lesser quality wines.  Bordeaux wine makers identify these lower quality wines as “second label” wines.  The poorest quality barrels of wine or vineyards are combined into “box wines”, or even sold to high volume low quality producers.  Small producers of high quality wines, often referred to as boutique wineries, have a very limited supply of grapes, and must use virtually all of their barrels of wine they produce.  This presents an interesting constraint for these small producers when determining their wine blends prior to bottling.  Clearly a mixture experiment can be performed with the components being barrels from different vineyards, or yeasts, or other processing treatments A second constraint requires all of the barrels of wine produced be used in one or more blends. Additional constraints such as vineyard designations, appellation designations, and varietal designations on labels that can also affect both the experimental design and the use of the resulting model.  This presentation will provide an explanation of these constraints while performing a wine blending mixture experiment for a boutique winery, and applying the resulting models to identify the optimal, sub-optimal, and residual solutions.