diff --git a/recognization/code/bench_ocr.py b/recognization/code/bench_ocr.py
new file mode 100644
index 0000000..e644bc2
--- /dev/null
+++ b/recognization/code/bench_ocr.py
@@ -0,0 +1,42 @@
+"""OCR cost per crop, on production-shaped crops.
+
+The decision record claims ~18-26 ms per crop on Mac and re-measured 18.0 ms on
+40 production crops. This is that measurement, so the number is reproducible
+rather than remembered.
+
+    .venv/bin/python code/bench_ocr.py
+"""
+import json, statistics, sys, time
+sys.path.insert(0, 'code')
+import cv2
+from eval_identity import LocalLaneAdapter, crop_of
+
+lane = LocalLaneAdapter('stock-lopia-clean')
+labels = json.load(open('eval/labels/image6.json'))
+image = cv2.imread('add/image6.jpg')
+
+crops = []
+for facing in labels['facings']:
+    if facing.get('label') in (None, 'exclude'):
+        continue
+    crop = crop_of(image, facing['box'])
+    if crop is not None and crop.size:
+        crops.append(crop)
+    if len(crops) >= 40:
+        break
+
+# Warm the recognizer: the first call loads the Vision model.
+lane.text_lane.read(crops[0])
+
+times = []
+for crop in crops:
+    started = time.perf_counter()
+    lane.text_lane.read(crop)
+    times.append((time.perf_counter() - started) * 1000)
+
+times.sort()
+print(f'{len(times)} crops')
+print(f'  p50 {statistics.median(times):.1f} ms')
+print(f'  mean {statistics.mean(times):.1f} ms')
+print(f'  p95 {times[int(len(times) * 0.95)]:.1f} ms')
+print(f'  min {times[0]:.1f}  max {times[-1]:.1f}')
diff --git a/recognization/docs/research/2026-09-28-recognition-architecture/DECISION.md b/recognization/docs/research/2026-09-28-recognition-architecture/DECISION.md
index 3aeac22..edc0916 100644
--- a/recognization/docs/research/2026-09-28-recognition-architecture/DECISION.md
+++ b/recognization/docs/research/2026-09-28-recognition-architecture/DECISION.md
@@ -2,7 +2,9 @@
 
 Audited 2026-09-28 · submodule `25dbf13` on `fix/live-recognition-quality` plus
 its uncommitted diff (snapshot: `git stash create` → `2785c07`) · auditor:
-Claude Opus 5.5
+Claude Opus 5.5. Implemented and re-measured 2026-09-29; every section that
+carries a later number says so, and the audit-time figures are kept beside them
+rather than replaced.
 Scope: identity pipeline (matcher, gallery, decision, evaluator, catalogs,
 labels), live track/label logic, the uncommitted diff. Not audited: detector
 quality, tag/planogram pipeline, iOS/Android rendering, on-device runtime.
@@ -214,7 +216,7 @@ Ordered by how much of the gap to a good product each class explains.
 | D | Retrieval → Van's matcher | yes (DINO) | DINO recall collapses with scale (7/42 @10 at 493 SKUs) | DINO adds ~41 ms per crop | **Reject DINO retrieval.** |
 | D′ | RootSIFT descriptor index (FLANN) → vote → verify top-10 | yes | 35–37/42, equal to or slightly better than exhaustive | 1.7–3.9 ms per crop flat from 7 to 493 SKUs; build 171 ms for 64k descriptors | **Adopt past ~150 SKUs.** |
 | E | Retrieval → learned verifier | no | — | — | Only if C wins. |
-| F | OCR text lane against catalog text signatures | yes (macOS Vision) | Veto: precision 62 → 82 % at τ=0.5 for −1 correct. As a lane: +10 correct at live, including unenrolled SKUs. **Audit-time figures on v1 labels**; the current measurement is in § Re-measured 2026-09-29 (E-3): the lane adds 16 pp of coverage at full resolution while raising precision. | ~18–26 ms per crop on Mac (re-measured 18.0 ms on 40 production crops); on-device on iOS (Vision) and Android (ML Kit) | **Adopt.** |
+| F | OCR text lane against catalog text signatures | yes (macOS Vision) | Veto: precision 62 → 82 % at τ=0.5 for −1 correct. As a lane: +10 correct at live, including unenrolled SKUs. **Audit-time figures on v1 labels**; the current measurement is in § Re-measured 2026-09-29 (E-3): the lane adds 16 pp of coverage at full resolution while raising precision. | ~18–26 ms per crop on Mac; re-measured 23.3 ms p50 / 37.2 ms p95 on 40 production crops (`code/bench_ocr.py`); on-device on iOS (Vision) and Android (ML Kit) | **Adopt.** |
 | G | Fusion (geometry + colour + text + temporal + tag prior) | partly (geometry + colour + text) | Largest measured gain | — | **Adopt**; the text and geometry errors are complementary (text reads siblings apart; geometry covers text-poor faces). |
 | H | Store-trained identity head on DINO | no (it was trained on the eval shelf) | Leaked; no valid measurement exists | — | Retire with DINO. |
 | J | Shelf-tag / planogram prior | no (`tags.py` exists) | Every facing run on image6 sits above a readable tag with name and price | — | Strong hypothesis, H-02. |