Dietary filters, keto / paleo / vegan / GF / DF / nut-free, how are they modeled?

One EAV attribute per dietary axis on the recipe entity, all multi-select-friendly Yes/No flags: diet_keto, diet_paleo, diet_vegan, diet_vegetarian, diet_gluten_free, diet_dairy_free, diet_nut_free, diet_shellfish_free, diet_soy_free, diet_low_cal (under 600 cal/serving), diet_kid_friendly, diet_pescatarian. Chef admin checks every applicable flag when uploading a recipe.

Customer side: the customer’s dietary profile lives on their account as a JSON column with two lists, hard exclusions (allergens, nuts, shellfish, gluten, dairy) and soft preferences (keto, low-cal, vegan). Hard exclusions filter the menu page to compatible recipes only (legal-and-safety reason; never show a nut-allergic customer a nut-containing recipe). Soft preferences re-rank but don’t hide.

Daily Harvest and CookUnity do this best. Customer onboarding flow asks the 12 questions in 2 minutes, then every weekly menu page is personalized. Net effect on retention: cuts "wrong meal for me" complaints by ~40% and reduces month-3 churn by 5-8 percentage points in the cohorts I’ve measured.