The Modern Bar Cart Podcast Episode 310 - Data-Driven Drinks with Dr. Kevin Peterson
Mar 5, 2026
Dr. Kevin Peterson, cocktail bar founder and author who researched customer cocktail-rating data. He discusses why matching people to their ideal drink is much harder than perfecting classics. The conversation covers high-dimensional preference variables, algorithmic and bartender-led ways to zero in on tastes, why spicy and bitter elements polarize, and how menu language and profiling can guide better matches.
AI Snips
Chapters
Books
Transcript
Episode notes
Closing Castalia And Saving Handwritten Drink Data
- Kevin Peterson closed Castalia after nearly a decade and gathered physical pen-and-paper customer surveys during service.
- He saved massive stacks of handwritten tasting sheets, noting handwriting degradation and drawings that added contextual data for later analysis.
Perfect Cocktail ≠ Perfect For Every Drinker
- Cocktail Theory optimized drinks by varying ratios, temperature, and dilution to find ideal ranges for classics.
- Kevin realized a perfectly made classic can still miss if the guest's palate doesn't match that style, prompting a shift from drink optimization to matching drink to person.
Cocktail Preferences Are High Dimensionally Nonlinear
- Guest preferences don't decompose neatly into additive ingredient likes; interactions and single small ingredients (like absinthe) can wreck a drink.
- Kevin found taste is high-dimensional and nonlinear, so you must consider whole-drink interactions, not just summed components.
