Showcase: From Vine to Verdict

A complete walkthrough of every jinflow instrument — using real data from Domaine Zufferey, a Valais winemaking estate. This is not a toy example. Every finding, thesis, and verdict shown here was computed from actual production data.

The domaine

Marie-Thérèse Zufferey runs a 45-parcel estate in Valais, producing 1,410 wines across multiple vintages. Her data comes from FileMaker — parcels, harvests, wines, barrels, cellar operations, lab analyses, bottle movements, sales orders, and customers. Millesime ingests it all through the medallion pipeline and builds a knowledge store.

The numbers: 867 harvests. 10,640 cellar operations. 44,472 lab analyses. 75,956 sales orders. 84,714 bottle movements.

Step 1: Entities — the Gold contract

The P-world (pipeline) delivers 10 clean entities:

Entity	Rows	What it represents
Parcels	45	Vineyard plots with grape variety, area, planting year
Harvests	867	Grape intake per parcel per vintage (kg, sugar, acid)
Wines	1,410	Produced wines with variety, vintage, appellation
Barrels	800	Oak barrels with age, capacity, current contents
Lab Analyses	44,472	pH, alcohol, SO2, acidity measurements per wine
Cellar Operations	10,640	Racking, fining, filtering, bottling events
Bottle Movements	84,714	Bottles in/out of stock per wine
Sales Orders	75,956	Customer orders with quantities and prices
Customers	1,200	Restaurants, retailers, private buyers
Harvest Trends	867	Year-over-year yield metrics per parcel

These are the interface. Everything from here on is T-world — the analyst and consultant never look below Gold.

Step 2: Signals — detect anomalies

12 signals examine the entities and produce 8,259 findings:

Signal	Findings	Worst	What it detects
probe_traceability_gap	80	high	Wines with no lab analysis chain — can’t prove vine-to-bottle
probe_missing_lab_analysis	102	high	Wines bottled without lab certification
probe_yield_decline	286	high	Parcels with sustained drops in kg/ha across vintages
probe_production_cost_vs_price	1,410	high	Wines where production cost approaches or exceeds selling price
probe_customer_concentration	2,143	medium	Wines where >40% of sales go to a single buyer
probe_harvest_yield_anomaly	40	medium	Harvests with yield far outside the varietal norm
probe_barrel_age_risk	117	medium	Barrels beyond their recommended service life
probe_inventory_balance	1,250	medium	Wines with stock discrepancies (bottles in ≠ bottles out)
probe_inactive_customer	0	—	Customers with no orders in 24 months (none found)
probe_sugar_alcohol_consistency	0	—	Lab results where sugar/alcohol ratio is implausible (none found)

Two perspectives aggregate findings into entity health scores:

Perspective	Entities scored	What it measures
assessment_wine_health	1,410 wines	Overall health combining traceability, lab, cost, inventory signals
assessment_parcel_health	1,421 parcels × vintages	Yield trends, harvest anomalies across parcels

The worst-scoring wine: Z-W-2021-1205 — flagged by 4 signals, 8 findings, CHF 6,122 at risk.

Step 3: Theses — is this a pattern?

Three theses evaluate whether the signal findings represent systematic business concerns:

thesis_traceability_compliance_risk — CONFIRMED

“Strong evidence of AOC compliance risk. Multiple wines lack lab certification before bottling, and traceability chains are broken. Immediate remediation recommended before the next audit.”

• Status: confirmed (evidence score 1.0)
• Findings: 1,432 across 2 signals
• Money at risk: CHF 757,238
• Evidence: probe_traceability_gap (primary) + probe_missing_lab_analysis (supporting)

thesis_revenue_pressure — CONFIRMED

“The domaine has measurable revenue risk. Key wines show high customer concentration, thin production margins, or inventory imbalances. Diversifying the customer base and reviewing pricing strategy is recommended.”

• Status: confirmed (evidence score 1.0)
• Findings: 4,803 across 3 signals
• Money at risk: CHF 1,737,312
• Evidence: probe_customer_concentration (primary) + probe_production_cost_vs_price (supporting) + probe_inventory_balance (context)

thesis_vineyard_aging_impact — CONFIRMED

“Clear evidence of yield decline on aging parcels. Multiple parcels show sustained drops in kg/ha across vintages. The domaine should prioritize a replanting plan for the most affected parcels.”

• Status: confirmed (evidence score 0.83)
• Findings: 1,696 across 2 signals
• Evidence: probe_yield_decline (primary) + probe_harvest_yield_anomaly (supporting)

Step 4: Verdicts — why is it happening?

Each confirmed thesis gets a root cause verdict:

Missing lab certification process (confidence: 90%)

Root cause: process_failure / lab_certification_gap

“The domaine bottles wine without systematic lab certification. Wines have cellar operations including bottling but no matching lab analyses. This is a process gap — either lab results exist but aren’t digitised, or analyses are performed informally without records.”

Recommendation: Establish a mandatory lab analysis step in the bottling workflow. Every wine should have at minimum pH, alcohol, and SO2 results recorded in the system before the BOTTLE operation is logged.

Narrow customer base (confidence: 85%)

Root cause: structural / customer_base_narrow

“Revenue pressure stems from a narrow customer base. Wines show high customer concentration (>40% from a single buyer), combined with inventory imbalances that suggest production outpaces diversified demand.”

Recommendation: Develop a direct-to-consumer sales channel (cave ouverte, online shop, wine club) to reduce dependency on restaurant accounts. Consider export markets for premium cuvées.

Replanting deficit (confidence: 60%)

Root cause: structural / replanting_deficit

“Yield decline is driven by a replanting deficit — aging parcels with vines >30 years have not been replaced on schedule. Parcels show sustained yield drops, and rising production costs erode margins on affected cuvées.”

Recommendation: Establish a 10-year replanting plan prioritizing parcels with vine age >35 years and yield below 60% of the 3-year rolling average. Consider interplanting younger vines alongside aging ones to maintain production continuity.

Step 5: SMEbits — expert knowledge

Two pieces of expert knowledge captured from Marie-Thérèse herself:

smebit_zufferey_old_vine_yield (business_rule)

“Parcels Z-P06 and Z-P08 have vines planted before 1980 — yield decline is expected and accepted for quality reasons.”

Marie-Thérèse’s heritage parcels Z-P06 (Cornalin, planted 1975) and Z-P08 (Petite Arvine, planted 1978) consistently produce 30-40% less than younger parcels. This is not a problem to fix — it is a deliberate quality strategy. The concentrated fruit from these old vines is what makes her top cuvées special.

This SMEbit anchors to probe_yield_decline — it tells the analyst: “yes, the signal flagged these parcels, but the SME says this is intentional.”

smebit_barrel_numbering_convention (structural)

“Each domaine uses a different barrel numbering prefix — Z-BRL for Zufferey, C-BRL for Clavien, B-BRL for Bétrisey.”

This enables safe cross-tenant barrel identification. A structural SMEbit — no check needed, just institutional knowledge that explains the data.

Step 6: BitBundle — the curated narrative

bb_confrérie_data_landscape

“The Confrérie’s Data Landscape — Three Systems, One Story”

The BitBundle curates 5 SMEbits into a narrative about how the three domaines of the Confrérie des Trois Coteaux each brought different systems (FileMaker, Navision, LabSystem) and how those differences shape what’s visible in the data.

SMEbit	Note
smebit_barrel_numbering_convention	The barrel prefix convention — proof the three families planned for cooperation from day one
smebit_navision_date_midnight_bug	The integration tax — Navision’s timestamp limitation shapes what’s possible across tenants
smebit_zufferey_old_vine_yield	Not every signal is an alarm — Marie-Thérèse’s heritage parcels are declining by design
smebit_betrisey_lab_only_tenant	Each system has a boundary — Bétrisey’s LabSystem is deep on science, silent on business
smebit_clavien_customer_restaurant_dependency	Business structure shapes data — Clavien’s concentration is a known reality, not a discovery

Step 7: Reports — the deliverable

6 reports (18 PDFs in DE/FR/EN):

Report	What it delivers
report_executive_summary	High-level overview for leadership — confirmed theses, total exposure, top recommendations
report_vintage_quality	Vintage-by-vintage quality metrics with lab analysis coverage
report_appellation_performance	Performance by AOC appellation — yield, quality, price positioning
report_cellar_inventory	Current barrel and bottle inventory with age risk flags
report_customer_portfolio	Customer concentration analysis with diversification recommendations
report_data_landscape	Data quality overview — completeness, validity rates, traceability coverage

The full chain

One thread, every instrument:

From a single wine (Entity) to a business recommendation (Report) — through every layer of the analytical pyramid. Every step is declarative, reproducible, and browsable in the Explorer.

Try it yourself

This KLS is available as a demo. Open it in the Explorer:

jinflow explore --tenant millesime.domaine_zufferey

Browse the findings. Click a thesis. Read the verdict. Check what the SME said. See the full report.

Then see the same engine at work in a completely different industry: From Shipment to Savings — the InterLogic logistics walkthrough.

Talk to your data — so it speaks to you.