Epic 8: Grinder Adjustment Learning System

[jodli]

Description

Track and learn from adjustment outcomes to refine recommendations and understand each user's specific grinder behavior.

User Stories

• As a user, I want the app to learn my grinder's actual adjustment impact
• As a user, I want to see if previous suggestions were successful
• As a user, I want increasingly accurate suggestions over time

Acceptance Criteria

• Track: previous grind → new grind → extraction time change → taste outcome
• Calculate grinder's step impact: "0.1 step ≈ 1.5 seconds"
• Show confidence level for suggestions based on historical success
• Refine suggestion amounts based on learned patterns
• Display success rate: "Following suggestions improved shots 75% of time"
• Minimum data threshold before making confident predictions
• Handle outliers and anomalies gracefully

Learning Algorithm

• Record all adjustments and outcomes
• Calculate average impact per step size
• Build confidence intervals
• Detect non-linear grinder behavior
• Account for bean-specific variations
• Weight recent data more heavily

User Feedback Loop

• "Did you follow the suggestion?" prompt
• "How did it turn out?" follow-up
• Track ignored suggestions separately
• Learn from both successes and failures

Part of #26 Phase 2
Critical dependency for accurate recommendations
Requires #29 and #30 from Phase 1