Original measured bench data for home espresso gear — super-automatics, Breville all-in-ones, single-boiler machines, hand grinders, espresso scales, and technique experiments — from the Brewkit gear bench. Every number below was logged on our own counter, on our own water and beans — one honest bench data point, not lab data.
- Website: https://brewkit.net/
- Last updated: 2026-10-13
- License: CC BY 4.0 (data); analysis and prose remain © Brewkit
Eleven CSV datasets from published Brewkit bench studies, plus the method notes needed to read them. The CSVs are machine-readable; each study section below gives the answer-first summary so you can quote it without opening a file.
Data: data/super-automatic-maintenance-rotation-2026.csv · Last updated: 2026-10-05 · Canonical article: Best Super-Automatic Espresso Machine in 2026
Four machines — De'Longhi Magnifica Evo, De'Longhi Magnifica S, Philips 3200 LatteGo, Philips 2300 LatteGo — rotated through the Brewkit bench over two months on the same medium-roast blend. We timed first-cup-to-second-cup cycles, counted daily cleanup steps, and logged grinder response to a one-setting change.
The headline numbers:
| Observation (2-month rotation) | Magnifica Evo | Philips 3200 | Philips 2300 |
|---|---|---|---|
| Milk-path rinse + reassemble, daily avg | 25 s (carafe auto-clean first) | 14 s | 14 s |
| Re-dial-ins needed after week one | 0 | 0 | 1 (one drift, fixed once) |
| Grinder noise, same counter | Loudest of the three | Quietest | Quietest |
| Finished cappuccino, button to cup | Fastest | Middle | Middle |
| Brew group serviceability | Removable, tap rinse | Not user-removable | Not user-removable |
The quotable finding: in 2026 the super-automatic category's real dividing line is not coffee quality — every machine here makes a credible cappuccino — it is which milk system you are still willing to clean in month six.
Method: 2-month rotation, one medium-roast blend, same counter/water/beans across all machines. Readings from our counter, our water, our beans; treat as one honest data point, not lab data. Bench observations are our own; manufacturer specs (grind settings, descale intervals, capacities) are cited to official listings and manuals in the canonical article.
Data: data/philips-3200-vs-2300-bench-2026.csv · Last updated: 2026-10-05 · Canonical article: Philips 3200 vs 2300 LatteGo (2026)
Data: the Philips 3200 and 2300 LatteGo share the same 100% ceramic grinder, the same two-part no-tube LatteGo milk system, and the same AquaClean maintenance — the 3200's higher price buys a fuller one-touch drink menu (5 varieties incl. latte macchiato and americano) and a touch display, nothing in the cup. Both bench-rinsed identically at 14 s daily average.
The shot character on both: a café-style, crema-layered shot that leans smooth and mild rather than punchy — the bean-to-cup character of the whole Philips Series 2x00/3x00 platform. Neither machine exposes extraction control (no PID-style temperature setting, no pressure profiling; fixed pre-brew aroma step only).
Method: two months of side-by-side use; every spec row is labeled in the CSV with its evidence type — manufacturer spec, manufacturer claim, or bench observation — so you can separate quoted specs from measured behavior.
Data: data/df64-vs-niche-zero-bench-2026.csv · Last updated: 2026-10-06 · Canonical article: DF64 vs Niche Zero (2026)
Two single-dose grinders, same medium-roast Ethiopia/Brazil blend, 18 g doses, one sample per grinder, dialed in advance. The quotable finding: neither grinder is "better" outright — DF64 Gen 2 gives brighter flat-burr clarity and Amazon availability; Niche Zero gives a quieter 72 dB grind and a body-forward conical cup, sold direct only.
The headline numbers:
| Observation | DF64 Gen 2 | Niche Zero |
|---|---|---|
| Espresso dial-in, fresh bag | 4 shots to window | 3 shots to window |
| Reference shot | 18 g in, 36 g out, ~26 s | 18 g in, 36 g out, ~27 s |
| Cup character | Brighter, more texture separation | Rounded, body-forward |
| 6 a.m. noise verdict | Noticeably louder at the cup | Conversation quiet (72 dB spec) |
| Weekly maintenance | ~3 min, brush + chute wipe | ~2 min, brush + burr lift |
Method: three-week side-by-side, same blend, one sample per grinder. Bench observations are our own; the 72 dB figure is a manufacturer spec, labeled as such in the CSV's evidence_type column.
Data: data/baratza-sette-270wi-vs-encore-esp-bench-2026.csv · Last updated: 2026-10-06 · Canonical article: Baratza Sette 270Wi vs Encore ESP (2026)
A stepped-collar grinder (Encore ESP) against a grind-by-weight grinder (Sette 270Wi), same 18 g reference doses, one sample per grinder, dialed in advance. The quotable finding: the 270Wi's integrated scale removes dose error, not grind error — when a shot ran fast on the bench, the fix was still a micro-click, not a re-press.
The headline numbers:
| Observation | Encore ESP | Sette 270Wi |
|---|---|---|
| Espresso dial-in, fresh bag | 5 shots to window (stepped collar) | 3 shots to window (micro steps) |
| Dose accuracy, 10 shots | ±0.3 g with external scale + pulse | ±0.1-0.2 g, auto-stop by weight |
| 18 g dose time | ~9 s (setting 10-ish) | ~5-6 s |
| Static at espresso fineness | Noticeable; clumps need a shake | Minor; straight-through path helps |
| Weekly maintenance | ~2 min, quick-release burr brush-out | ~3 min, arms off, chamber brush |
Method: three-week side-by-side, same 18 g reference doses. Bench observations are our own, logged during normal daily use.
Data: data/gaggia-classic-pro-vs-rancilio-silvia-bench-2026.csv · Last updated: 2026-10-06 · Canonical article: Gaggia Classic Pro vs Rancilio Silvia (2026)
The two 58 mm single-boiler classics on the same bench. The quotable finding: the Evo Pro's small aluminum boiler is ready first and steams sooner, but shot-to-shot temperature drift is noticeable until you surf or PID it; the Silvia's 12 oz brass boiler is slower to everything yet barely breaks stride on the fourth shot in a row.
The headline numbers (bench rows):
| Observation | Gaggia Classic Evo Pro | Rancilio Silvia |
|---|---|---|
| Ready to pull from cold | ~1 min class (fast small boiler) | Patience — mass first, shots second |
| Reference shot | 14 g in, 28 g out, 25–27 s | 16 g in (stock double), 32 g out, ~27 s |
| Shot-to-shot temp drift | Noticeable; surf or PID it | Small once settled — brass mass |
| 150 ml milk to microfoam | ~35–45 s, two-hole wand | Comparable result; longer wait to steam temp |
| Fourth shot in a row | Recovery gap appears | Barely breaks stride |
| Weekend descale + backflush | ~15 min | ~20 min (0.3 L brass to flush) |
Method: same-counter side-by-side with spec rows (boiler, wand, reservoir, build) labeled manufacturer spec and behavior rows labeled bench observation in the CSV's evidence_type column.
Data: data/jx-pro-s-vs-timemore-c2-bench-2026.csv · Last updated: 2026-10-06 · Canonical article: 1Zpresso JX-Pro S vs Timemore C2 (2026)
Two hand grinders, same 18 g dose of the same medium-roast blend. The quotable finding: step size decides espresso dial-in — one JX-Pro S click is about a quarter of one C2 click (12.5 µm vs ~85 µm measured), and the same 18 g : 36 g ~28 s recipe needed 4 clicks of correction on the JX-Pro S versus 1 click on the C2, where that single click jumped the shot from 22 s to 34 s.
The headline numbers:
| Observation | JX-Pro S | Timemore C2 |
|---|---|---|
| Measured click size at fine end | 12.5 µm per click (spec) | roughly 85 µm per click (caliper bench check) |
| Fines (<250 µm) at espresso setting | ~21% | ~27% |
| Shot time spread (5 shots, same dose/ratio) | ±2.1 s | ±4.7 s |
| Grind time for 18 g at espresso fineness | ~40 s | ~55 s |
| Filter grind (V60, 20 g) | effectively a tie | effectively a tie |
Method: same beans, same target ratio, both grinders at their best achievable espresso setting; fines from a 5-run sifting average, shot-time spread over 5 shots (bench, September 2026). The ~85 µm C2 click size is our caliper bench check against Timemore's ~80-micron-per-click class documentation; labeled in the CSV by evidence type.
Data: data/espresso-channeling-wdt-2026.csv · Last updated: 2026-10-06 · Canonical article: Espresso Channeling: 7 Causes, Ranked (2026 Fix Guide)
A paired technique experiment on a Breville Barista Express (BES870): 20 shots, same beans, same grind setting, same 18.0 g dose and 36 g yield — 10 prepared by tapping the portafilter flat and tamping, 10 with a quick 5-second needle stir (WDT) first, judged on a bottomless portafilter.
The headline numbers:
| Metric (10 shots each) | Tap + tamp | WDT + tamp |
|---|---|---|
| Shots with visible side jets or a wandering stream | 5 of 10 | 1 of 10 |
| Time to 36 g (spread) | 22–34 s | 26–29 s |
| Shots both sour and bitter in the cup | 4 of 10 | 0 of 10 |
| Spent pucks with a visible crater | 3 of 10 | 0 of 10 |
The quotable finding: five seconds of stirring needles was the single highest-return technique change we tested on this machine — visible channeling dropped from 5 of 10 shots to 1 of 10.
Method: single-session kitchen data, same operator, not a laboratory study; direction matches published distribution experiments cited in the canonical article.
Data: data/breville-barista-express-vs-gaggia-classic-pro-bench-2026.csv · Last updated: 2026-10-06 · Canonical article: Breville Barista Express vs Gaggia Classic Pro (2026)
The all-in-one against the upgradable 58 mm classic, same bench. The quotable finding: the Breville gets you 90% there automatically; the Gaggia will beat that 90% once you learn it — and its two-hole wand rolls 150 ml of milk into microfoam in roughly 35–45 seconds while the Breville's single-hole wand is slower and less forgiving.
The headline numbers (bench rows):
| Observation | Barista Express (BES870XL) | Gaggia Classic Evo Pro |
|---|---|---|
| Reference shot | 18 g in, 36 g out, pulled in 26–28 seconds | 14 g in, 28 g out, 25–27 s |
| Warm-up | Ready fast — under a minute to brew temp | Boiler ~5 min; a full 15 for the brass group to saturate |
| 150 ml milk to pourable microfoam | slower and less forgiving | roughly 35–45 seconds |
| Loudest moment on either counter | grinder running 8–10 seconds per dose | quieter than pre-2019 Classics |
| Shot quality once learned | gets you 90% there automatically | will beat that 90% once you learn it |
Method: spec rows (portafilter, grinder, boiler, controls, tank, power) labeled manufacturer spec; behavior rows labeled bench observation in the CSV's evidence_type column, from the same-counter comparison in the canonical article.
Data: data/barista-express-impress-touch-bench-2026.csv · Last updated: 2026-10-06 · Canonical article: Barista Express vs Impress vs Touch (2026)
The three-tier Breville lineup on one bench. The quotable finding: the assisted-tamp Impress hits a balanced 27 s pull in 3 dial-in shots and contains the mess, while the touchscreen Touch automates milk down to ~10 s of hands-on time for a cappuccino.
The headline numbers (bench rows):
| Observation | Barista Express | Express Impress | Barista Touch |
|---|---|---|---|
| Time to first drinkable shot, from box | ~35 min | ~20 min | ~15 min (guided screen setup) |
| Dial-in shots to a balanced 27 s pull | 4 | 3 (dose trim helps) | 3 |
| Shot-to-shot consistency, days 2–7 | Good once dialed | Very good — tamp is a constant | Very good — timed + temp-stable |
| Cappuccino milk, hands-on time | ~45 s of wand work | ~45 s of wand work | ~10 s (load jug, press) |
| Counter mess per session | Grinder scatter + puck knock | Lowest — tamp contains grounds | Moderate |
Method: spec rows (heating, grind settings, tamping, milk system) labeled manufacturer spec; the observation table rows labeled bench observation in the CSV's evidence_type column, verbatim from the canonical article's bench table.
Data: data/espresso-scale-response-bench-2026.csv · Last updated: 2026-10-06 · Canonical article: Best Espresso Scale (2026)
Four espresso scales measured with a 1 g/s simulated flow (an espresso shot pours slower; we stress-tested). The quotable finding: the BOOKOO Themis Mini was the fastest to settle of the four (~0.2 s), while the Timemore Black Mirror Basic 2 chased a 36 g cutoff and never overshot by more than 0.4 g.
The headline numbers (bench rows):
| Observation | Black Mirror Basic 2 | Maestri House S2 | Themis Mini | Weightman |
|---|---|---|---|---|
| Display settle at 1 g/s simulated flow | within ~0.3 s | within ~0.5 s | fastest of the four (~0.2 s) | 1–2 s to settle at a fast pour |
| 36 g cutoff overshoot | never more than 0.4 g | — | — | — |
| Drip-tray clearance needed | 26 mm of height | 15 mm plus its glass top | — | 17 mm |
Method: spec rows (resolution, timer, charging) labeled manufacturer spec; response rows labeled bench observation in the CSV's evidence_type column, from the 1 g/s simulated-flow stress test described in the canonical article.
Data: data/bambino-plus-vs-casabrews-cm5418-bench-2026.csv · Last updated: 2026-10-13 · Canonical article: Best Espresso Machine Under $300 (2026)
The under-$300 bracket's two real contenders, two weeks each on the bench. The quotable finding: the Bambino Plus's shot-to-shot temperature repeatability was indistinguishable from our reference single-boiler, while the CM5418 reached brew temperature in about a minute and its single-hole wand produced paintable microfoam — with one real compromise: its thermoblock needs a pause between steaming and pulling the next shot.
The headline numbers (bench rows):
| Observation | Bambino Plus (BES500) | Casabrews CM5418 |
|---|---|---|
| Shot-to-shot temperature repeatability | indistinguishable from our reference single-boiler | — |
| Time to brew temperature on bench | (spec: ready ~3 s) | reached brew temperature in about a minute |
| Pressure gauge during dial-in | n/a — no pressure gauge | tracked plausibly with grind changes |
| Milk texturing result | auto-milk poured credible latte art on the first try | paintable microfoam for latte art |
| Between steaming and the next shot | — | pause needed between steaming and pulling the next shot |
The quotable framing: at this price you choose between thermal sophistication (Bambino Plus) and value-plus-upgradability (CM5418) — and on either bench, the grinder, not the machine, is the bottleneck.
Method: two weeks per machine on the Brewkit bench; at least four shots daily across three roast levels; milk steaming twice daily with whole and oat milk; heat-up and steam-recovery cycles timed with a thermocouple in the group head; shots judged blind against the reference machine; every dialing-in session started from the same beans (a 10-day-off-roast medium and a 16-day-off-roast dark). Spec rows labeled manufacturer spec (official Breville BES500 and Casabrews CM5418 pages, retrieved September 2026); the pump-rating row labeled manufacturer claim; behavior rows labeled bench observation; the session-design rows labeled bench protocol in the CSV's evidence_type column.
Each study ran on the Brewkit bench (same beans, same counter, one variable at a time where applicable) with numbers logged during the rotation or side-by-side period named in the study header. Readings are from our own machines and kitchen — one honest, reproducible data point per study, not laboratory measurement. Manufacturer specifications are quoted from official product pages and manuals (retrieved September 2026) and labeled as such.
Yes — the CSV data is released under CC BY 4.0. Attribute "Brewkit (brewkit.net)" and link the canonical article named in each study. The analysis, verdicts, and prose on brewkit.net remain copyright Brewkit; the CSV numbers are free to reuse with attribution.
Brewkit — an independent espresso-gear publication that measures the gear it writes about. Bench studies land on the site first; this repository republishes the underlying numbers for anyone who wants machine-readable data. Study requests: hello@brewkit.net.