Why these 30 machines
The desk is an editorial comparison set of 30 machines worth comparing across five buying tasks: semi-automatic espresso, super-automatic, drip & batch, pod & capsule, and manual & portable. Machines are chosen for specification diversity and common buyer constraints (footprint, grinder/milk workflow, skill level) — not sales rank, popularity scores, or hands-on lab testing.
Current allocation: drip (6), manual portable (5), pod capsule (6), semi automatic (7), super automatic (6).
What editorial order means
editorial_order (desk #1–30) is the default sequence within category chapters. It reflects buying-task grouping and editorial judgment about which models belong on the shortlist — not a quality score, award, or recommendation rank.
Non-default sorts (price, width, reservoir, brand) reorder the full filtered set globally. Null price, width, or reservoir values always sort last.
Source policy
- Manufacturer pages are preferred for dimensions, features, and list prices.
- Manuals and support PDFs corroborate capacities, intervals, and workflow limits.
- Secondary sources are used only when manufacturer pages are unreachable; they are labeled in the evidence ledger.
- URLs are spot-checked; known 404s are removed or replaced. Bot-protected (403) pages may be retained with
manual-check-requiredverification status.
Price policy
US prices use separate price_verification (verified-visible, manual-check-required, unverified) and source_access (reachable-2xx, anti-bot-403, manual-check-required, stale-404) fields. Verified-visible prices are independently observable on a linked page; manual-check prices are numeric but require human confirmation; unverified means no honest numeric price is recorded. Prices change — confirm before buying. 23 of 30 machines currently carry a numeric price; the remainder show “Price not confirmed.”
Availability
Each record uses current, legacy, discontinued, or unverified honestly. A machine is marked current only when a current manufacturer or store page supports it.
Known vs unknown vs not applicable
Every nullable specification field is classified exactly once:
- Known — value present on a linked source.
- Not confirmed (
unknown) — value is null and not ruled out for this machine class. - Not applicable — value is null because the field does not apply (e.g., portafilter size on a drip brewer).
Detail pages, CSV, JSON, and the catalog UI share these semantics.
Correction policy
Errors in specs, sources, or prices are corrected when verified against manufacturer or manual evidence. Discontinued SKUs are replaced with current equivalents or marked honestly. Coffee Decoded did not perform hands-on lab testing on this desk.
Schema v2, evidence classes, and comparability
Machine records use schema v2 with a stable source_registry, versioned estimate methods, and a field_meta entry for every comparison field. Supported field states: manufacturer_declared, regulatory_test_lab_measured, independent_measured, calculated_estimate, editorial, unknown, and not_applicable.
- Cross-category rows compare apples-to-apples workflow, footprint, milk, and cleaning fields — not pump maximums or thermoblock-vs-boiler equivalence.
- Category-specific technical rows appear only within the same machine class.
- Energy rankings stay within the same category, protocol, and evidence basis. Rated/nameplate watts are capacity labels, never used as kWh fallbacks.
- Manufacturing origin fields never infer assembly country from headquarters, “designed in,” or seller location.
- Conflicted observations remain separate and do not participate in rankings.
Energy cost estimates
Daily and annual kWh scenarios use formula coffee-energy-v1.0.0 only when measured brew-cycle and state-power inputs exist. The compare desk shows rated watts as nameplate capacity; calculated scenarios are labeled:
Calculated, not measured. Based on the usage assumptions and energy inputs shown below using Formula coffee-energy-v1.0.0. Change the assumptions to match your use.
Missing required inputs block a product point estimate — we do not invent consumption from rated watts.
Buyer-fit editorial flow
The buyer-fit question flow on the machine desk produces a transparent 2–3 machine shortlist from ten routine questions. Each candidate shows best-for, avoid-if, tradeoffs, assumptions used, unknown facts, and why it matched. Editorial inference is labeled and separated from sourced facts. There is no star score or “best overall” without a documented testing dataset.
Future merchant links (inactive)
Merchant offers live in a separate inactive data/offers.yaml boundary. Machine editorial records do not contain affiliate IDs, tracking parameters, or merchant priority. Inactive or absent offers produce no buy CTA; availability never alters ranking or editorial labels.
Top 10 field note
The Top 10 guide is a separate editorial snapshot with manufacturer-linked field notes and intentionally omits live prices. The Top 30 catalog is the full comparison desk.