Machine Learning in Wine Marketing?

We talked a lot in class / in blogs about the wine shopping experience: consumers are often overwhelmed by the plethora of brands and have a difficult time making a choice. There are just too many options. Even if a consumer knows roughly what she is looking for (e.g., California Cabernet Sauvignon or Burgundy Chardonnay), chances are that she will still have to decide from a dozen or dozens of brands in her preferred price range.

However, at the end of the day the consumer has to make a purchase decision. So how does she actually selects a bottle of wine amidst uncertainty and lack of information? The possible approaches are:
a) Pick randomly ("I'll just try this bottle for no reason")
b) Consult someone at the store / online ("Let me ask the wine specialist / Vivino")
c) Go with her "gut" ("This bottle looks interesting")

I'd like to dive a bit deeper into approach c). What is the "gut feeling" of a customer, and based on what is the gut feeling formed? More specifically, if a consumer looks at 10 bottles of wines from the same region / grapes at the same price, what would influence her gut feeling?

The most straightforward hypothesis is that label design, when all else are equal, heavily influence consumers' purchase decision. I think it would be very interesting to use regression analysis / machine learning to understand the specific factors that drive consumer adoption. Several factors / regressors to consider in this analysis, for example:

Brand name:
- Length
- Binary variables for category (e.g., has name of location, has animal, has family name, has number...)
- Language (English, French, German, not a real word, etc.)

Label design:
- Brand name font and font size
- Color scheme (number of colors, background color, dominant color(s), etc.)
- Center image category (estate, nature, grape, animal, human, coat of arms, signature, no image, etc.)
- Center image style (watercolor, oil painting, sketches, modern art, wood-carved, etc.)

Label information:
- Grape varietal shown
- Vintage shown / vintage year
- Appellation / AVA / geography shown

There are other nuanced but important decisions to make for this experiment. For example:
- Do we want to include price as a regressor, or run experiment on wines at the same price?
- How do we separate brand name impact from pure label design impact? Should we run experiment of lesser-known brands only?
- How do we categorize the "emotion" invoked by label design?

It would be interesting to see if any data scientist or startup company can dig into the science of wine label design and reveal the obscure trends in consumer preference.