Officially sourced
A manufacturer or authorized retailer published the specification. Espresso Receipt checked the record and date.
PRE-REGISTERED PROTOCOL 1.2.0 / NOT YET RUN
The bench protocol we published before testing anything.
This page is a commitment, not a result. It was pre-registered on 2026-07-24, and no physical unit has run it. Publishing the modules, minimums, and rejection rules in advance is what keeps a later observed label from being written around a favorable outcome.
PHYSICAL RECORDS PUBLISHED00
NO CLAIMUNTIL THE RUN PASSES
- Pre-registered
- 2026-07-24
- Module lanes
- 9
- Queued questions
- 08
- Bench status
- UNFUNDED
HONEST STATUS
Pre-registered 2026-07-24. Zero physical records published.
Physical testing has not happened. Every module, minimum repetition count, and rejection rule below is written down and frozen, and the public record array is empty. Testing begins when the queue funds — acquiring retail units at retail prices, at U.S. voltage, without manufacturer samples, is the cost that gates the first run. Until then the Observatory publishes only what the site actually collects: first-party price observations and analyses modeled from sourced specifications.
THE LINE WE WILL NOT BLUR
Four labels. Four different claims.
Editorial interpretation
A fit or workflow judgment grounded in cited product architecture. It is not a physical measurement.
Modeled
A disclosed formula produced the value from visible assumptions. The output changes when the assumptions change.
Independently observed
A resolved physical unit completed this protocol. Conditions, repetitions, instruments, raw data, and limitations travel with the value. Nothing carries this label today.
PROTOCOL LANES / 1.2.0
Every result must survive the whole bench.
A machine and grinder do not run identical modules. Both must clear identity, physical fit, real workflow, sound and energy, and reset. Product-specific lanes then test the mechanism that matters. The repetition minimums are floors, not targets.
IDMACHINE + GRINDER Resolve the exact unit
Record model, public SKU, revision, region, voltage, acquisition channel, and a non-public serial fingerprint before a result can attach to a catalog record.
- Minimum
- 1 inspection
- Published output
- Unit identity and revision boundary
FITMACHINE + GRINDER Measure the working footprint
Measure closed dimensions, operating clearance, access paths, cable or hose reach, and the space needed for normal loading, filling, and cleaning.
- Minimum
- 3 runs
- Published output
- Closed and operating dimensions
FLOWMACHINE + GRINDER Time the real workflow
Record a repeatable start-to-finish task from a cold, reset station, including the touches the product requires rather than timing one isolated mechanism.
- Minimum
- 5 runs
- Published output
- Median task time and observed interruptions
BREWMACHINE Test brew repeatability
Use a fixed coffee, grinder, basket, dose, yield, preparation sequence, and instrumented brew path to compare consecutive extractions without turning temperature into a universal taste score.
- Minimum
- 10 runs
- Published output
- Temperature and beverage-mass spread
MILKMACHINE Test steam and recovery
Heat a fixed mass of refrigerated milk to the same stop temperature, recording elapsed time, recovery, and required operator steps across repeated cycles.
- Minimum
- 5 runs
- Published output
- Steam time, recovery, and workflow notes
DOSEGRINDER Test dose delivery
Run a documented seasoning and purge procedure, then measure input, output, elapsed time, and exchange across consecutive doses.
- Minimum
- 10 runs
- Published output
- Dose variance, exchange, and grind time
CTRLGRINDER Map useful adjustment
Measure particle-distribution movement and shot-flow response across documented adjustment points while keeping coffee, dose, preparation, machine, and target yield fixed.
- Minimum
- 10 runs
- Published output
- Adjustment response and usable espresso window
LOADMACHINE + GRINDER Record sound and energy
Measure equivalent sound level at a fixed distance and energy used during the declared task. Report the room floor and test duration with every value.
- Minimum
- 5 runs
- Published output
- Task dBA and energy use
RESETMACHINE + GRINDER Reset the station
Complete routine contact-point cleaning from a used state, record access, consumables, elapsed time, retained mess, and any model-specific manual boundary.
- Minimum
- 5 runs
- Published output
- Reset time, consumables, and access notes
PUBLICATION GATE / FAIL CLOSED
A result is rejected if any required piece is missing.
Interrupted and conflicting runs can remain in the internal run log. They cannot become a public observed claim until the record is complete. The gates are enforced in code by the record validator, not by editorial goodwill.
- 01
Exact retail unitPublic SKU, revision, U.S. voltage, acquisition path, and protected serial fingerprint.
- 02
Complete required modulesNo favorable subset can stand in for the declared product protocol.
- 03
Calibrated instrumentsInstrument identity, purpose, declared unit capabilities, and calibration check precede the run.
- 04
Minimum repetitionsConditions and run order remain attached to every statistic.
- 05
Raw data + limitationsEvery published statistic recomputes from frozen raw repetitions, released at a stable path with a verified checksum.
- 06
Two-person reviewTwo named reviewers, distinct from each other and from the operator. Corrections supersede; they do not silently rewrite history.
VERSIONS AND CUSTODY
Protocol 1.2.0 · schema 1.2.0 · pre-registered 2026-07-24.
The protocol version governs how a run is conducted; the schema version governs the shape of the published record. Changing either starts a new record rather than editing an old one. The public endpoint already exposes both versions, the module list, the queue, and a validated record array that is empty today.