What Cellar is optimizing for
- Open bottles when they are likely to deliver the most value.
- Protect bottles that are too young or too rare to waste.
- Surface decline risk before the bottle becomes a regret.
This documents the Phase 1 probability-based drinking timeframe model used in the Cellar inventory tab.
Cellar estimates a likely drinking window, not a guaranteed expiration date. The model is designed to help you decide what to hold, what to open, and what may be at risk of fading. It is intentionally explainable, conservative, and built from fields you can realistically enter.
| Variable | Weight | Why It Matters |
|---|---|---|
| Varietal Aging Potential | 30% | Grape structure strongly influences ageability through tannin, acidity, sugar, and phenolic density. |
| Region / Appellation | 15% | Climate, tradition, and appellation standards affect structure and longevity. |
| Vintage Quality | 15% | Weather conditions influence ripeness, acidity, concentration, and balance. |
| Producer Tier | 15% | Better producers tend to deliver better fruit selection, structure, consistency, and historical aging performance. |
| Wine Style | 10% | Structured and traditional wines usually age longer than soft, fruit-forward, early-drinking wines. |
| Alcohol Balance | 5% | Balanced alcohol supports longevity; excessive or poorly integrated alcohol can shorten the useful window. |
| Price Proxy | 5% | Price is imperfect, but it often proxies fruit quality, oak program, vineyard source, and production intent. |
| Bottle Age Context | 5% | The current age helps classify the wine as too young, approaching prime, mature, or declining. |
APS =
varietalScore * 0.30 +
regionScore * 0.15 +
vintageScore * 0.15 +
producerScore * 0.15 +
styleScore * 0.10 +
alcoholBalanceScore * 0.05 +
priceScore * 0.05 +
bottleAgeScore * 0.05Every input is normalized to a 0–100 score before the weighted formula is applied.
| APS | Peak Start | Peak End |
|---|---|---|
| 0–20 | Vintage + 0 years | Vintage + 3 years |
| 21–40 | Vintage + 2 years | Vintage + 6 years |
| 41–60 | Vintage + 4 years | Vintage + 10 years |
| 61–80 | Vintage + 8 years | Vintage + 20 years |
| 81–100 | Vintage + 12 years | Vintage + 35 years |
Too Young More than two years before estimated peak start.
Approaching Prime Within two years of estimated peak start.
Prime Window Inside the first 65% of the estimated window.
Late Plateau Inside the final 35% of the estimated window.
Declining Past estimated peak end.
Phase 1 does not yet use community tasting notes, critic re-tastings, bottle condition, storage temperature history, cork variation, or producer-specific vertical tasting data. That is the future moat. This version is the disciplined foundation, not the final intelligence layer.
Export JSON as your master backup. CSV is for spreadsheet review.
Phase 1 Drinking Window Engine
Introduced the Cellar name and added the first probability-based drinking-window model.
- Aging Potential Score.
- Estimated peak start/end years.
- Drinking-stage labels and model confidence.
- Methodology tab foundation.
Visual Intelligence Layer
Added visual maturity cues so drinking-window predictions became easier to understand at a glance.
- Maturity curve visualization.
- Drink Soon / Prime / Decline / Too Young cards.
- Inventory-card model bindings fixed.
Open Tonight Recommendations
Added a decision engine to recommend which bottle to open based on context.
- Occasion, meal, style, and strategy inputs.
- Ranked bottle recommendations.
- Recommendation reasons and last-bottle warnings.
Bottle Evolution Feedback Loop
Added opened-bottle review tracking so Cellar can compare predicted maturity against actual bottle experience.
- Maturity, fruit, structure, tertiary, and open-again signals.
- Opened-bottle history.
- Inventory quantity reduction after review.
ML-Ready Data Persistence
Added durable D1 storage for tasting entries, cellar inventory, drinking-window outputs, and evolution reviews.
- wine_entries: tasting log and scoring data.
- cellar_inventory: bottle inventory, model scores, drink windows, and status.
- bottle_evolution_reviews: opened-bottle feedback for future prediction refinement.
Analytics Dashboard
Added portfolio-level cellar intelligence and fixed Analytics button feedback.
- Prime readiness, decline risk, drink-soon count, and cellar value.
- Maturity, aging-potential, varietal, and region distributions.
- Optimal opening timeline, urgent bottles, and learning signals.
Smart Alerts & Opening Queue
Added alert queues that turn analytics into immediate action.
- Immediate attention queue.
- Approaching-prime queue.
- Protected holdings queue.
Opening Plan Layer
Added planning tools that rank bottles by maturity stage, decline risk, bottle count, aging potential, and selected planning style.
- Planning-window selector.
- Openings-per-month control.
- Saved opening plan.
Data Quality & Model Confidence Audit
Added audit tools that identify missing high-value fields that reduce drinking-window confidence.
- Average data quality score.
- High-confidence and cleanup-needed counts.
- Per-bottle missing-field issue tags.
Collection Intelligence Engine
Cellar evolves from a tracking application into an intelligent cellar operating system capable of evaluating collection quality, maturity balance, concentration risk, and strategic opportunity.
- Collection health scoring.
- Strategic maturity signals.
- Priority bottle intelligence.
- Portfolio-level cellar analysis.
Consumption Forecasting Engine
Added inventory runway forecasting so Cellar can estimate how long the cellar will last and identify which bottles should be prioritized first.
- Monthly consumption modeling.
- Projected cellar runway.
- Priority-consumption signals.
- Inventory longevity forecasting.
Guaranteed Goal Delete UI Fix
Rebuilt delete-button behavior with explicit button types, inline event prevention, and hardened localStorage removal logic.
Hard Goal Deletion Fix
Rebuilt goal deletion logic with direct localStorage removal and hardened button bindings.
Database Goal Deletion
Deleting a collection goal now removes the entry from the UI, local storage, and the D1 database.
Functional Goal Edit/Delete Actions
Fixed non-responsive Edit and Delete buttons on the Goals tab by replacing fragile delegated handlers with direct button bindings.
Goal Management Enhancements
Added goal editing, deletion, and target-vintage tracking for more advanced collection planning.
- Edit existing goals.
- Delete collection goals.
- Target vintage support.
- Vintage-aware duplicate detection.
Goal Duplicate Entry Fix
Fixed duplicate goal creation caused by overlapping button bindings and added duplicate-goal prevention logic.
Collection Goals & Acquisition Targets
Added long-term cellar objective tracking so Cellar can guide future purchasing and collection shaping.
- Varietal and region targets.
- Aging-potential goals.
- Inventory growth targets.
- Progress tracking against cellar objectives.
Predictive Cellar Timeline
Added year-by-year cellar forecasting so Cellar can project future maturity opportunities and decline risk across the collection.
- Future peak-window forecasting.
- Upcoming maturity milestones.
- Projected decline-risk visualization.
- Long-term cellar trajectory modeling.
Strategy Profiles & Adaptive Recommendations
Added personalized drinking strategies so Cellar can adapt recommendations based on collector behavior and maturity preference.
- Balanced Collector profile.
- Aggressive Early Drinker profile.
- Peak Window Maximizer profile.
- Long-Term Cellarer profile.
- Special Occasion Saver profile.
Cellar Intelligence Feed
Added a centralized insight engine that synthesizes alerts, maturity data, inventory concentration, and evolution feedback into actionable intelligence.
- Critical decline-risk insights.
- Prime-window opportunities.
- Collection concentration warnings.
- Age-worthiness and evolution signals.
ML Dataset Refresh Fix
Fixed the Refresh ML Dataset button so it visibly rebuilds metrics, preview rows, status text, and last-refreshed timestamp.
ML Dataset Export & Training Readiness
Added tools to convert Cellar data into exportable model-ready rows for future drinking-window machine learning.
- Training CSV export.
- Training JSON export.
- Raw cellar and evolution JSON exports.
- Training-row preview and labeled-coverage metrics.
Deployments Tab Fix
Fixed the Deployments tab so the changelog renders as a dedicated top-level view instead of appearing blank or leaking into product tabs.