Three-component consumer blending pilot

Relationship to the operating procedure

Run this pilot under Consumer-scale blending of commercial whiskies. Where this pilot is silent, that procedure controls.

Pilot objective

Demonstrate that one consumer formulator can:

define a target and assign component roles before blending;
prepare a transparent three-component mixture design without arithmetic or transcription errors;
separate descriptive, target-fit, and preference judgments;
manage order, fatigue, and carryover across bounded sessions;
obtain an independently repeated shortlist rather than a one-session favorite;
reconstruct the selected formula from the original bottles; and
preserve an audit trail sufficient for another person to repeat the work.

The pilot does not establish category-wide flavor laws, consumer-market preference, a fitted response surface, or ASTM/ISO conformity.

Authorization gate — bottle selection

Do not start liquid trials until the user approves:

Bottle A — structural lead: approachable, reconstructible, and plausible as the main body of the blend.
Bottle B — support or contrast: contributes a defined function that A lacks without failing the invariant quality floor.
Bottle C — high-impact accent: contributes a concentrated, distinctive function and is expected to require smaller adjustments.

For each bottle record expression, batch or lot when available, proof, fill level, acquisition source, replacement availability, intended role, observed role, and any stock constraint.

Target gate

Write and version:

one sentence defining the intended drinker and use occasion;
the invariant quality floor—attributes every acceptable formula must retain;
the variable target—aroma, palate, mouthfeel, finish, proof range, and contrast allowed to change;
explicit hold or reject conditions; and
a predeclared material-improvement rule for advancing a candidate.

If the target changes, open a new pilot version.

Participation and session gate

Before pouring, document participant role, suitability and informed participation, serving and total-exposure plan, stop conditions, water and food, non-drinking alternatives, and transportation. Follow current law, venue policy, and professional guidance. The public-source pass does not provide a universal safe pour or recovery interval.

Phase 1 — component qualification

Profile A, B, and C separately under consistent conditions.
Record attributes and intensity with a bounded working lexicon.
Make use, hold, or reject decisions against the quality floor.
Confirm that each candidate has a plausible role and enough stock for the complete design, independent rebuilds, and retained references.
Record proof and decide whether comparisons will use native proof, a controlled common proof, or both in separate sessions. Never change proof silently.

Phase 2 — seven-point diagnostic design

Use proportions, not arbitrary “drops.” Prepare enough material for the approved session plan while minimizing total alcohol exposure.

Formula	A	B	C	Purpose
P-A	100%	0%	0%	Pure control
P-B	0%	100%	0%	Pure control
P-C	0%	0%	100%	Pure control
P-AB	50%	50%	0%	Binary interaction screen
P-AC	50%	0%	50%	Binary interaction screen
P-BC	0%	50%	50%	Binary interaction screen
P-ABC	33.33%	33.33%	33.34%	Three-way center screen

The final 0.01 percentage point is a bookkeeping convention so the displayed three-way formula sums to 100%; measured preparation must use the recorded calculation and tool precision.

Phase 3 — preparation control

Create a versioned formula sheet with target batch volume, each component volume or mass, proof, preparer, timestamp, vessel, and code.
Confirm every coordinate sums to the declared total before liquid is combined.
Prepare retained references and preserve enough untouched stock for reconstruction.
Randomize sample codes and record the code key separately.
Record immediate appearance or preparation anomalies. Do not “fix” a sample without issuing a new version.
Do not treat an evolving infinity bottle as a reproducible component unless its current composition, proof, and batch boundary are measured and frozen for the pilot. Keep creative infinity practice outside the controlled recipe series.

Phase 4 — diagnostic sensory sessions

Question A — description and quality floor

Record observations and anchored attribute intensities before preference. Mark every formula pass, hold, or reject against the invariant quality floor.

Question B — target fit

Use a predeclared five-point local scale:

strongly misses the target;
misses the target;
partially fits;
meets the target; and
strongly meets the target.

This is a project scale, not a standardized sensory scale.

Question C — local preference

After description and target fit are complete, record preference or ranking separately. Preference does not replace target fit, and neither response identifies a causal attribute.

Session construction

Divide the seven formulas into bounded blocks set by the approved participation plan; do not force all samples into one sitting.
Use an overlapping coded anchor across blocks so block-to-block drift can be observed.
Standardize glassware, proof treatment, temperature, rest, instructions, and response form.
Randomize or counterbalance order as appropriate to the question.
Record suspected carryover, heat, adaptation, and fatigue. Split or stop the session when later judgments may be contaminated.
Reveal identities only after all observations for that session are locked.

Phase 5 — independent repeat

Reprepare the seven coordinates or the retained shortlist in a new session rather than treating repeated sips from one pour as independent evidence.
Use new codes and a newly randomized order.
Compare session-level quality-floor decisions, target-fit tier, attribute direction, and preference.
Advance a region only when it survives the repeated session. A single identical rank order is not required, but material reversals require hold and diagnosis.

Phase 6 — bounded refinement

Retain at most two or three promising coordinates.
Define a narrower feasible region around those coordinates and state any minimum or maximum bounds before generating formulas.
Change the structural base in larger measured steps and the high-impact accent in smaller measured steps, with actual increments declared in the design rather than inherited as universal percentages. The apparatus must support calibrated sub-percent trials when a trace component is plausible; reported cases justify fine search resolution but do not establish a universal effective threshold.
Preserve a control and the prior best in each branch.
Do not fit or optimize a response surface unless the design has adequate coverage, independent evidence, and a declared model check.

Phase 7 — reconstruction

Rebuild the provisional winner from the original bottles using only the stored formula and preparation record.
Compare the rebuilt sample with a retained reference under coded conditions.
Record whether any perceptible or material difference is observed. Do not call this a formal triangle test unless the licensed procedure and statistical rules are followed.
If reconstruction fails, diagnose measurement, proof, vessel, rest, stock, transcription, or sample-condition causes before changing the formula.

Pilot pass criteria

Advance the pilot from In Revision only when:

all three components passed qualification and remain traceable and replaceable enough for the intended use;
all seven diagnostic coordinates and subsequent branches have complete arithmetic and preparation records;
descriptive, target-fit, and preference responses remain separate;
no unresolved carryover, fatigue, code, or session-control defect undermines the shortlist;
the promising region survives at least one independently prepared or independently evaluated repeat;
the provisional winner can be reconstructed from the original record;
claims are limited to the tested bottles, formulas, participant(s), sessions, and context; and
every hold, rejection, disagreement, correction, and stopping decision is retained.

Hold or fail conditions

Hold or fail the pilot when bottle identity or proof is uncertain, stock is inadequate for reconstruction, the target changes midstream without versioning, arithmetic or coding cannot be audited, carryover or fatigue remains uncontrolled, the winning formula reverses materially without explanation, or reconstruction cannot reproduce the retained result.

Required record

The completed pilot must contain the target brief, bottle profiles, participation plan, formula coordinates, calculations, code key, session plan, individual sensory forms, carryover notes, repeat comparison, refinement logic, reconstruction check, and pass/hold decision.

Evidence boundary

The seven-point mixture screen is drawn from NIST/SEMATECH mixture-design guidance. Practitioner sources provide formulation architecture and measured iteration. The California Gold and Barrell case reports add independently reported examples of material sub-percent adjustments; they justify testing fine increments, not a universal threshold. The Punch infinity-bottle reports define an evolving-blend boundary and document carryover as a failure mode. Public ASTM and ISO catalog scopes provide alcohol-participation, method-family, participant-role, carryover, descriptive, discrimination, and hedonic boundaries. The licensed standards were not accessed; exact sample sizes, formal statistics, trained-panel criteria, standardized serving rules, and conformity claims remain outside this pilot.

Review gate

Current status: In Revision. The procedure is structurally ready, and the 2026-08-29 web-source intake has been integrated. Liquid execution is blocked only by the user-approved Bottle A, Bottle B, Bottle C, target brief, and participation plan.