Pre-Discovery: Batches Pass Initial QC
Between 11 July and 8 August 2022, four batches were produced using juniper lot PDI-JNP-220620: batch 19 (Gold Gin), batch 20 (Dry Gin), batch 21 (Jason Kong), and batch 22 (Dry Gin). All four passed post-production quality control (QC) and were approved for bottling.
Each batch was compared against a reference bottle pulled from finished goods rather than a fixed baseline. Because the reference changed each time, gradual drift went unnoticed.
Accidental Detection
On 15 August 2022, during a routine quality assessment of a later batch, I selected an existing bottled sample as a reference. The sample came from one of the contaminated batches. On tasting, I noticed a clear deviation from the house standard: a musty undertone and an atypical bitterness.
At that point, I did not know whether the deviation was real. It could have been palate fatigue, a bad sample, or a storage issue. I treated it as a weak signal and moved to confirm it before acting. By then, affected stock had already been released and was in circulation across accounts. QC had not caught it. I only found it because I happened to use that bottle as a reference.
Initial Hypothesis and Investigation
To rule out a sampling issue, I pulled a second bottle from the same batch and found the same flavour degradation. The issue was batch-wide.
Once the second bottle confirmed the deviation was real, the question shifted from whether the signal was real to how far it had spread. Before diagnosing the cause, I needed to scope the blast radius. I tasted bottles from batches produced before and after to find where the deviation started and stopped. As the same degradation appeared across multiple products, batch-level causes became unlikely. The pattern pointed to a shared ingredient, and the investigation shifted from batch-level causes to input-level causes.
Ingredient Isolation
Working from production records, I identified five ingredients shared across all four affected batches. After removing base ingredients common to all batches, two botanicals remained: juniper berries and lime peel.
A physical inspection using sight and smell did not reveal any clear issues, so I moved to single botanical distillations to isolate each ingredient and test it on its own. I started with juniper, as it made up the largest proportion of the recipes.
The juniper distillation reproduced the off-notes: a musty undertone and atypical bitterness, consistent with earlier observations. Juniper was confirmed as the source of the deviation.
Root Cause Confirmation
Supplier and inventory records confirmed that all four affected batches were linked to a single lot: PDI-JNP-220620. The contamination had been in circulation for 56 days before it was detected.
Investigation Timeline
| Date (2022) | Event | Detail | Confidence | Basis |
|---|---|---|---|---|
| 20 Jun | Lot received | PDI-JNP-220620 received from Supplier A. Visual inspection only. No flavour baseline recorded. | — | Pre-incident. No concern raised. |
| 11 Jul – 8 Aug | Production | Batches 19–22 produced using contaminated lot: B19 (Gold Gin), B20 (Dry Gin), B21 (Jason Kong), B22 (Dry Gin). | — | Pre-incident. No concern raised. |
| Jul – Aug | QC passes | All four batches pass post-production QC. No deviation detected. Approved for bottling. | — | Pre-incident. No concern raised. |
| 15 Aug | Accidental discovery | Contaminated bottle selected as reference during unrelated tasting. Musty undertone and atypical bitterness detected. | ~50% | Single bottle, single taster. Sample error, palate fatigue, and storage effect not yet ruled out. |
| 15 Aug (Step 1) | Batch confirmation | Second bottle pulled from same batch. Deviation confirmed. | ~75% | Sample error and storage ruled out. Issue is batch-wide. Cause unknown. |
| 15 Aug (Step 2) | QC record review | QC records reviewed across all four batches. No prior deviation flagged in any record. | ~80% | No historical deviation across the full batch window. Recipe and process error ruled out. |
| 15 Aug (Step 3) | Blast radius scoping | Bottles pulled from batches outside the affected window and tasted against house standards. Batches 19–22 confirmed affected. No deviation outside this window. | ~85% | Deviation bounded to a single lot window. Shared ingredient is the only remaining variable. |
| 15 Aug (Step 4) | Ingredient isolation | Single botanical distillation conducted on each ingredient. Off-notes reproduced by juniper lot PDI-JNP-220620 only. | ~100% | All other ingredients tested clean. Root cause confirmed. |
| 15 Aug | Quarantine | All four batches quarantined. Stock recovery initiated. Supplier A notified. | — | Irreversible action taken on confirmed finding. |
| Post-15 Aug | Remediation | Single-batch test distillations introduced for all new ingredient lots before full production. Quality gate updated at ingredient receipt. | — | Structural fix implemented at the point of failure. |
Ten events. 56 days from receipt to confirmed root cause.