Ongoing evaluation by stratum (bean x technique)
library(here)
library(dplyr)
library(gt)
source(here::here("R", "load_data.R"))
source(here::here("R", "summarize.R"))
source(here::here("R", "plotting.R"))
# trials.csv only ever holds completed trials (no planned/abandoned rows) -
# see CLAUDE.md. load_trials() defaults to completed_only = TRUE.
completed <- load_trials()See DESIGN.md for the full factor list, stratum definitions, and notes on which trials changed more than one within-stratum factor at once.
latest_by_stratum(completed) |>
dplyr::select(
trial_id,
date,
stratum,
grind_clicks,
filter_type,
dose_g,
water_g,
notes
) |>
gt::gt() |>
gt::tab_header(title = "Most recent completed trial per stratum")| Most recent completed trial per stratum | |||||||
| trial_id | date | stratum | grind_clicks | filter_type | dose_g | water_g | notes |
|---|---|---|---|---|---|---|---|
| COL-IMM-01 | 2026-08-15 | COL-JAS x immersion | 23 | NA | 20.0 | NA | One-off immersion attempt on the O1 body (not O2 - this predates O2 acquisition), using #8 regulator instead of the standard O1/#4 pairing. This is an EXCEPTION to the fixed brewer/regulator-by-technique convention used elsewhere in this log - see DESIGN.md. Ran into unspecified technical issues but tasted noticeably better than the repeated 20-click pour-over brews: more sour and sweet notes. Water volume not stated. Date is an approximation assigned within the Aug 1-20 window per user instruction, not sourced from an exact conversation timestamp. |
| COL-PO-01 | 2026-08-05 | COL-JAS x pour_over | 20 | NA | 16.0 | NA | Represents several repeated brews at 20 clicks/#4 ("the last few brews"), not a single trial - consistently bitter and muted. Diagnosed as likely channeling from uneven bloom saturation on this dense, high-elevation bean. Date is an approximation assigned within the Aug 1-20 window per user instruction, not sourced from an exact conversation timestamp. Dose assumed at the 16g pour-over default since not stated explicitly. |
| KEN-HYB-03 | 2026-09-06 | KEN-KAB x hybrid | 19 | flat | 19.0 | 310 | Tasted across three smaller pours: low bitterness and sourness, slightly sweet aftertaste. Good cup, near top line but no real overflow risk. Vs. KEN-HYB-02 (22 clicks): sourness/sweetness/bitterness/body/overall all identical despite finer grind (19 vs 22) - suggests grind alone may not be the dominant driver in this range for the hybrid technique. |
| KEN-PO-15 | 2026-09-18 | KEN-KAB x pour_over | 17 | flat | 16.3 | 260 | Intended as a third poke-vs-stir isolation vs KEN-PO-05 (17cl/flat/stir), but water ran to ~97C instead of the ~92.8C baseline - confounded: grind fixed at 17, but agitation (stir->poke) AND temperature (+~4C overshoot) both changed at once. Strong sour notes dominant, not sweet at first taste, secondary bitter notes, good body, fruity nose. Bitterness of 3 is by far the highest of any poke trial (KEN-PO-08/10/11 all had bitterness 1) - strong evidence the temp overshoot, not poke, drove the bitterness (over-extraction). Sourness held exactly at KEN-PO-05's stir-baseline value (4) rather than dropping the way poke usually does at other grinds, suggesting heat may have counteracted poke's usual sourness-reducing effect. Does NOT cleanly answer poke-vs-stir at 17 clicks - that isolation still outstanding. New finding: bitterness on this bean appears quite sensitive to even a small (~4C) temperature overshoot. |
Numeric ratings are only populated for trials logged after ratings were introduced; earlier trials show NA here and are captured qualitatively in notes in the full log below.
summarize_by_stratum(completed) |>
gt::gt() |>
gt::fmt_number(columns = starts_with("mean_"), decimals = 1) |>
gt::tab_header(title = "Mean ratings by stratum")| Mean ratings by stratum | |||||||||
| stratum | bean | technique | n_trials | n_rated | mean_sourness | mean_sweetness | mean_body | mean_bitterness | mean_overall |
|---|---|---|---|---|---|---|---|---|---|
| COL-JAS x immersion | COL-JAS | immersion | 1 | 0 | NaN | NaN | NaN | NaN | NaN |
| COL-JAS x pour_over | COL-JAS | pour_over | 1 | 0 | NaN | NaN | NaN | NaN | NaN |
| KEN-KAB x hybrid | KEN-KAB | hybrid | 3 | 2 | 2.0 | 3.0 | 2.0 | 1.0 | 3.0 |
| KEN-KAB x pour_over | KEN-KAB | pour_over | 15 | 13 | 2.8 | 2.5 | 2.4 | 1.5 | 3.3 |
plot_ratings_by_stratum(completed)
(This plot will be empty until numeric ratings are logged for at least one trial — see data/data_dictionary.md.)
completed |>
dplyr::arrange(stratum, date) |>
gt::gt() |>
gt::tab_header(title = "All completed trials")| All completed trials | ||||||||||||||||||||||||||||
| trial_id | date | status | bean | technique | stratum | ratio | brewer | regulator | filter_type | grind_clicks | dose_g | water_g | temp_c | bloom_g | pulse_structure | stir | pour_rate_gs | valve_open_to_g | valve_timing_notes | total_time_s | steep_dwell_s | sourness | sweetness | body | bitterness | overall | overflow | notes |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| COL-IMM-01 | 2026-08-15 | completed | COL-JAS | immersion | COL-JAS x immersion | NA | O1 | 8 | NA | 23 | 20.0 | NA | 92.5 | NA | NA | NA | NA | NA | NA | NA | 105 | NA | NA | NA | NA | NA | NA | One-off immersion attempt on the O1 body (not O2 - this predates O2 acquisition), using #8 regulator instead of the standard O1/#4 pairing. This is an EXCEPTION to the fixed brewer/regulator-by-technique convention used elsewhere in this log - see DESIGN.md. Ran into unspecified technical issues but tasted noticeably better than the repeated 20-click pour-over brews: more sour and sweet notes. Water volume not stated. Date is an approximation assigned within the Aug 1-20 window per user instruction, not sourced from an exact conversation timestamp. |
| COL-PO-01 | 2026-08-05 | completed | COL-JAS | pour_over | COL-JAS x pour_over | NA | O1 | 4 | NA | 20 | 16.0 | NA | 92.5 | NA | NA | FALSE | NA | NA | NA | NA | NA | NA | NA | NA | NA | NA | NA | Represents several repeated brews at 20 clicks/#4 ("the last few brews"), not a single trial - consistently bitter and muted. Diagnosed as likely channeling from uneven bloom saturation on this dense, high-elevation bean. Date is an approximation assigned within the Aug 1-20 window per user instruction, not sourced from an exact conversation timestamp. Dose assumed at the 16g pour-over default since not stated explicitly. |
| KEN-HYB-01 | 2026-08-23 | completed | KEN-KAB | hybrid | KEN-KAB x hybrid | 16.25000 | O2 | 8 | NA | 21 | 20.0 | 325.0 | 92.5 | NA | NA | NA | NA | 145 | Valve open through bloom + percolation to 145g, then closed for final 180g pour. | NA | 60 | NA | NA | NA | NA | NA | FALSE | No sourness, good acidic profile, easy to sip, subdued sweetness, earthy/fruity. Came close to overflowing but did not. Actual steep ~60s vs. Orea's own hybrid spec of ~13-15s closed-valve dwell. Bean was 4 days post-roast (roasted Aug 19) at brew time. |
| KEN-HYB-02 | 2026-08-29 | completed | KEN-KAB | hybrid | KEN-KAB x hybrid | 16.33158 | O2 | 8 | flat | 22 | 19.0 | 310.3 | 92.5 | NA | continuous spiral, no pulses | FALSE | 5 | 138 | Valve open through bloom + percolation to 138g, then closed, continuing to 310.3g actual (171g standing). Recipe targeted a 15s closed-valve dwell; actual dwell was ~60s - same overshoot pattern as KEN-HYB-01, so this still isn't a real short-dwell data point. | NA | 60 | 2 | 3 | 2 | 1 | 3.0 | FALSE | Easy to sip, very enjoyable; earthy/fruity with a citrus undertone; body leans a bit watery. Liquid was visually near the top line but with no actual risk of spilling - notably different from KEN-HYB-01 (180g standing), which was a genuine near-overflow. The ~9g difference in standing volume (180g vs 171g) appears to have meaningfully changed the risk level, suggesting the safety margin is narrower/more volume-sensitive than previously assumed. Dose (20->19g) and grind (21->22 clicks) both changed vs. KEN-HYB-01 alongside the dwell, so no single factor can be cleanly credited for the cup outcome. |
| KEN-HYB-03 | 2026-09-06 | completed | KEN-KAB | hybrid | KEN-KAB x hybrid | 16.31579 | O2 | 8 | flat | 19 | 19.0 | 310.0 | 92.5 | 40 | continuous spiral, no pulses | FALSE | 5 | 140 | Valve closed for initial ~5s of bloom then opened (deviation from fully-open bloom design); open through 140g; closed for final pour to 310g (170g standing); dwell ~30-45s vs 15s design target. | 235 | 38 | 2 | 3 | 2 | 1 | 3.0 | FALSE | Tasted across three smaller pours: low bitterness and sourness, slightly sweet aftertaste. Good cup, near top line but no real overflow risk. Vs. KEN-HYB-02 (22 clicks): sourness/sweetness/bitterness/body/overall all identical despite finer grind (19 vs 22) - suggests grind alone may not be the dominant driver in this range for the hybrid technique. |
| KEN-PO-01 | 2026-08-23 | completed | KEN-KAB | pour_over | KEN-KAB x pour_over | 15.56886 | O1 | 4 | wave | 23 | 16.7 | 260.0 | 93.0 | 60 | 4x50g | FALSE | NA | NA | NA | NA | NA | NA | NA | NA | NA | NA | NA | Good cup, lightly sour but pleasant cranberry-like undercurrent. |
| KEN-PO-02 | 2026-08-27 | completed | KEN-KAB | pour_over | KEN-KAB x pour_over | 15.56886 | O1 | 4 | flat | 21 | 16.7 | 260.0 | 93.0 | 60 | 4x50g | FALSE | NA | NA | NA | NA | NA | NA | NA | NA | NA | NA | NA | Good in both nose and taste, no sourness complaint. Note: grind (23->21 clicks) AND filter (wave->flat) both changed vs. KEN-PO-01, so the improvement can't be cleanly attributed to either alone. |
| KEN-PO-03 | 2026-08-29 | completed | KEN-KAB | pour_over | KEN-KAB x pour_over | 15.95092 | O1 | 4 | wave | 22 | 16.3 | 260.0 | 93.0 | 60 | 60/110/160/210/260 | TRUE | NA | NA | NA | 210 | NA | 3 | 2 | 2 | 1 | 3.5 | NA | 22 clicks/16.3g/wave/stirred after bloom, vs. Trial 2 (21 clicks/16.7g/flat, no stir noted). Grind and dose deltas treated as equivalent for analytic purposes per user - the effective comparison is filter type (flat->wave) and bloom stir (added). Watery body, low sweetness, near-zero bitterness, subdued earthy/fruity like a jam, not bright but not dark/oily either - a solid, satisfying cup. |
| KEN-PO-04 | 2026-08-30 | completed | KEN-KAB | pour_over | KEN-KAB x pour_over | 15.95092 | O1 | 4 | flat | 20 | 16.3 | 260.0 | 93.0 | 60 | 60/110/160/210/260 | TRUE | 5 | NA | NA | 211 | NA | 3 | 2 | 2 | 1 | 3.5 | NA | 20 clicks/16.3g/flat/chopstick stir (~10s) - vs Trial 3 (22 clicks/wave/stir, 210s), two variables changed (grind, filter) but total time (211s vs 210s) and all five ratings came out identical to Trial 3. Notable null result: muted, sour-ish berries, earthy but fruitier side, faint/intermittent sweetness, body present but not substantial. Suggests the watery/muted/moderately-sour character in this bean's O1 profile isn't primarily filter- or grind-driven in the 20-22 click range - a bigger lever (dose/ratio, temp, or a different technique like O2 hybrid) may be needed to shift it meaningfully. |
| KEN-PO-05 | 2026-09-01 | completed | KEN-KAB | pour_over | KEN-KAB x pour_over | 15.95092 | O1 | 4 | flat | 17 | 16.3 | 260.0 | 93.0 | 60 | 60/110/160/210/260 | TRUE | 5 | NA | NA | 270 | NA | 4 | 3 | 2 | 1 | 4.0 | NA | 17 clicks/16.3g/flat/chopstick stir - single-variable grind swing vs Trial 4 (20 clicks/flat, 211s); total time nearly doubled to 270s (4:30). Best result yet on this bean and the first trial to move the profile after the 20-22 click plateau (Trials 1-4 all landed near 3-3.5 overall). Bright, quick-receding sourness leads; sweet aftertaste initially masked by the sour note; low bitterness throughout; body unchanged at 2. Confirms grind is a real lever on this bean, just not within the narrow 20-22 click range tested earlier. |
| KEN-PO-06 | 2026-09-03 | completed | KEN-KAB | pour_over | KEN-KAB x pour_over | 15.95092 | O1 | 4 | flat | 15 | 16.3 | 260.0 | 93.0 | 60 | 60/110/160/210/260 | TRUE | 5 | NA | NA | 285 | NA | 2 | 2 | 2 | 1 | 2.5 | NA | 15 clicks/16.3g/flat/chopstick stir - vs Trial 5 (17 clicks, 270s, overall 4), continued finer but overall dropped to 2.5. CAVEAT: tasted fresh, but attention was divided with a meeting rather than dedicated exclusively to tasting - ratings flagged as noisier/lower-confidence for that reason (not due to cooling). Watery body, low sourness/sweetness, easily drinkable but thin, no oiliness. User recommends re-running this exact recipe under undivided attention before concluding 15 clicks overshot the sweet spot found at 17 clicks. |
| KEN-PO-07 | 2026-09-05 | completed | KEN-KAB | pour_over | KEN-KAB x pour_over | 15.95092 | O1 | 4 | flat | 15 | 16.3 | 260.0 | 93.0 | 60 | 60/110/160/210/260 | TRUE | 5 | NA | NA | 300 | NA | 3 | 2 | 1 | 2 | 2.0 | NA | Clean repeat of Trial 6 (15cl/flat/16.3g), tasted with undivided attention. Total time ~5:00, longest yet. Sourness relatively high (3), body thin (1), tasted under-extracted-feeling warm/hot despite longest brew time. No visually noticed channeling, but the long-time+weak-extraction combo is suspicious for it (same failure mode as the earlier Colombian bean). Combined with Trial 6, two independent runs at 15 clicks both underperform Trial 5's 17 clicks - solid evidence 15 clicks overshot the sweet spot. |
| KEN-PO-08 | 2026-09-07 | completed | KEN-KAB | pour_over | KEN-KAB x pour_over | 15.95092 | O1 | 4 | flat | 19 | 16.3 | 260.0 | 93.0 | 60 | 60/110/160/210/260 | TRUE | 5 | NA | NA | 250 | NA | 3 | 3 | 3 | 1 | 4.0 | NA | Bloom agitation instinctively changed from the standard chopstick stir to poking holes with a chopstick instead, since stirring seemed to only catch the top layer of grounds rather than saturate the full bed. Two variables vs KEN-PO-05 (17cl/stir): grind (17->19) and agitation method (stir->poke), cause not isolated. Ties KEN-PO-05's overall 4 with better balance - lower sourness, higher body, standout chocolatey nose. Very satisfying cup. Next trial isolates whether grind or poke technique drove the tie. |
| KEN-PO-09 | 2026-09-08 | completed | KEN-KAB | pour_over | KEN-KAB x pour_over | 15.95092 | O1 | 4 | flat | 19 | 16.3 | 260.0 | 93.0 | 60 | 60/110/160/210/260 | TRUE | 5 | NA | NA | 277 | NA | 3 | 2 | 1 | 2 | 2.0 | NA | Isolation trial vs KEN-PO-08 (19cl/poke, overall 4): only agitation changed back to stirring, grind held at 19. Overall dropped to 2 - worse than KEN-PO-08 on every axis except sourness (flat at 3). Sweetness down to 2, body down to 1 (watery/thin), bitterness up to 2. Dominant note sourness with slight sweet aftertaste; not as satisfying as KEN-PO-08. CONCLUSION: clean, decisive evidence that the poke agitation technique - not the grind bump to 19 - drove KEN-PO-08's improvement. Poking now confirmed as the better technique for this bean. |
| KEN-PO-10 | 2026-09-10 | completed | KEN-KAB | pour_over | KEN-KAB x pour_over | 15.95092 | O1 | 4 | flat | 20 | 16.3 | 260.0 | 93.0 | 60 | 60/110/160/210/260 | FALSE | 5 | NA | NA | 250 | NA | 2 | 4 | 3 | 1 | 4.0 | NA | 20 clicks/16.3g/flat/poke (not stir) - reuses KEN-PO-04's exact baseline (20cl/flat/stir, overall 3.5) with only agitation changed. Beats it outright: overall 4, highest sweetness of any trial, lowest sourness among the good results, no bitterness. Notes: chocolate/stonefruit, subdued sourness as undertone, good but not overly substantial body, very easily drinkable, pleasant every sip. Second independent isolation (after KEN-PO-08 vs 09 at 19 clicks) showing poke beats stir, now at a different grind point - consistent direction strengthens the case for poke as genuinely better. |
| KEN-PO-11 | 2026-09-14 | completed | KEN-KAB | pour_over | KEN-KAB x pour_over | 15.95092 | O1 | 4 | flat | 22 | 16.3 | 260.0 | 93.0 | 60 | 60/110/160/210/260 | FALSE | 5 | NA | NA | 235 | NA | 1 | 3 | 4 | 1 | 4.0 | NA | 22 clicks/16.3g/flat/poke - continues coarsening-with-poke trend from KEN-PO-08 (19cl) and KEN-PO-10 (20cl). Fastest of the three poke trials (235s vs 250s/250s), consistent with coarser grind. Sourness continued falling (3->2->1), body climbed to its highest yet (3->3->4), sweetness dropped back from KEN-PO-10's peak (3->4->3, not monotonic). Overall ties at 4 for the third poke trial in a row. Described as very smooth, possibly the best cup yet despite the numeric tie - three trials in the 19-22 click range now share overall 4 with poke, a wider good zone than stir ever achieved in this range (3-3.5). |
| KEN-PO-12 | 2026-09-15 | completed | KEN-KAB | pour_over | KEN-KAB x pour_over | 15.95092 | O1 | 4 | wave | 22 | 16.3 | 260.0 | 93.0 | 60 | 60/110/160/210/260 | FALSE | 5 | NA | NA | 228 | NA | 3 | 3 | 3 | 1 | 4.0 | NA | 22 clicks/16.3g/wave/poke - NOT the planned replicate of KEN-PO-11 (user switched to Wave filter instead of flat), so this became a filter isolation instead. Pleasant nose/body, subdued-citrus sourness dominant with sweet aftertaste, no bitterness; only critique was a slightly subdued overall flavor profile. Vs KEN-PO-11 (flat, same grind+poke): Wave raised sourness (1->3) and dropped body (4->3), sweetness/bitterness/overall held steady - evidence flat filter reduces sourness and boosts body relative to Wave when poke is used, consistent with Orea's bypass-reduction claim for flat. Also a third poke-vs-stir data point: beats KEN-PO-03 (wave/stir, overall 3.5) at same grind+filter, though not a same-session isolation. True flat-filter replicate of KEN-PO-11 for confidence-building still outstanding. |
| KEN-PO-13 | 2026-09-16 | completed | KEN-KAB | pour_over | KEN-KAB x pour_over | 15.95092 | O1 | 4 | flat | 22 | 16.3 | 260.0 | 93.0 | 60 | 60/110/160/210/260 | FALSE | 5 | NA | NA | 310 | NA | 3 | 2 | 2 | 2 | 3.0 | NA | Attempted flat-filter replicate of KEN-PO-11, brewed while partially distracted by a work call. Total time 310s, 32% longer than KEN-PO-11's 235s despite an otherwise identical recipe and same poke technique - likely uneven/delayed pulse timing during the actual pour, not just noisy tasting. Enjoyable but not great; dominant note subdued sourness, watery mouthfeel despite the standard 260g pour. NOT a clean replicate - the divided attention plausibly affected the brew itself (timing), not just the tasting. Second instance of work-call multitasking degrading trial data quality (first was KEN-PO-06's divided-attention tasting). True clean replicate of KEN-PO-11 still outstanding. |
| KEN-PO-14 | 2026-09-17 | completed | KEN-KAB | pour_over | KEN-KAB x pour_over | 15.95092 | O1 | 4 | flat | 20 | 16.3 | 260.0 | 92.8 | 60 | 60/110/160/210/260 | FALSE | 5 | NA | NA | 280 | NA | 3 | 2 | 3 | 2 | 3.0 | NA | Attempted clean replicate of KEN-PO-10 (20cl/poke/flat/16.3g) - first trial with actually measured temperature (199F/~92.8C via thermometer, confirming the long-assumed ~93C baseline was accurate). No distraction this time; 280s total, 12% longer than KEN-PO-10's 250s despite no obvious cause. Meaningful gap from KEN-PO-10 (sourness 2, sweetness 4, bitterness 1, overall 4) - did not reproduce the chocolate/stonefruit result. Second failed attempt to confirm KEN-PO-10 (after KEN-PO-12's filter detour) - raises real question of whether KEN-PO-10 was the true expected result at 20cl+poke or a favorable single draw from unmeasured brewing variance (poke depth/pattern, pour steadiness, timing). KEN-PO-10's result should be treated as possibly an optimistic outlier, not yet confirmed reproducible. |
| KEN-PO-15 | 2026-09-18 | completed | KEN-KAB | pour_over | KEN-KAB x pour_over | 15.95092 | O1 | 4 | flat | 17 | 16.3 | 260.0 | 97.0 | 60 | 60/110/160/210/260 | FALSE | 5 | NA | NA | 275 | NA | 4 | 2 | 3 | 3 | 4.0 | NA | Intended as a third poke-vs-stir isolation vs KEN-PO-05 (17cl/flat/stir), but water ran to ~97C instead of the ~92.8C baseline - confounded: grind fixed at 17, but agitation (stir->poke) AND temperature (+~4C overshoot) both changed at once. Strong sour notes dominant, not sweet at first taste, secondary bitter notes, good body, fruity nose. Bitterness of 3 is by far the highest of any poke trial (KEN-PO-08/10/11 all had bitterness 1) - strong evidence the temp overshoot, not poke, drove the bitterness (over-extraction). Sourness held exactly at KEN-PO-05's stir-baseline value (4) rather than dropping the way poke usually does at other grinds, suggesting heat may have counteracted poke's usual sourness-reducing effect. Does NOT cleanly answer poke-vs-stir at 17 clicks - that isolation still outstanding. New finding: bitterness on this bean appears quite sensitive to even a small (~4C) temperature overshoot. |