ETA accuracy benchmark · 🇬🇧 Southampton

How Port Flow's predicted ETA compares to each vessel's own broadcast ETA at Southampton, measured on closed voyages over the last 7 days. We publish the number — good or bad. Lower error is better.

Port Flow predicted ETA (RMSE)

2.19 h

MAE 1.45 h · 101 voyages

Broadcast ETA (RMSE)

51.64 h

crew-declared, 45 voyages (same set)

Our advantage

95.8 %

lower error than the broadcast ETA

Broadcast ETAs more than 7 days from the actual arrival are excluded as AIS sentinel / placeholder values (49 excluded). Both figures are computed on the same remaining voyage set — apples to apples.

Closed voyages

error = predicted/broadcast minus actual arrival (hours)


229822000	product	10/12, 04:24 AM UTC	-1.78 h	-2854.41 h
232019192	container	10/12, 04:13 AM UTC	0.42 h	1939.78 h
232019192	container	10/12, 03:17 AM UTC	0.29 h	—
246724000	container	10/11, 11:31 PM UTC	5.46 h	1.62 h
255915758	container	10/11, 11:29 PM UTC	0.71 h	-1.49 h
259307000	container	10/11, 10:39 PM UTC	0.19 h	-2.63 h
255804590	container	10/11, 06:12 PM UTC	-11.10 h	-272.21 h
209247000	container	10/11, 05:38 PM UTC	-6.23 h	-756.64 h
636021249	container	10/11, 03:14 PM UTC	0.32 h	-3309.24 h
255801540	container	10/11, 02:38 PM UTC	-0.33 h	-10.14 h
235009804	container	10/11, 11:35 AM UTC	-0.13 h	—
257268000	container	10/11, 11:33 AM UTC	-0.01 h	-2.31 h
257368000	container	10/11, 10:25 AM UTC	-0.05 h	0.57 h
257368000	container	10/11, 10:11 AM UTC	-0.09 h	0.80 h
235040503	chemical	10/11, 10:04 AM UTC	0.73 h	-34.82 h
255893000	chemical	10/11, 09:26 AM UTC	1.94 h	-5.45 h
255801540	container	10/11, 06:32 AM UTC	-0.09 h	-131.54 h
235505000	container	10/11, 06:32 AM UTC	0.63 h	-79.53 h
563229400	container	10/11, 02:49 AM UTC	-14.35 h	-2527.83 h
235007413	chemical	10/10, 09:32 PM UTC	-0.03 h	-5.53 h
538008142	product	10/10, 12:29 PM UTC	3.26 h	-49.00 h
235040503	chemical	10/10, 10:41 AM UTC	-0.25 h	-11.44 h
255915622	container	10/10, 07:47 AM UTC	-14.81 h	-653.30 h
235007880	chemical	10/10, 06:57 AM UTC	0.21 h	0.04 h
244092000	lng	10/10, 05:46 AM UTC	0.16 h	-16.27 h
235009956	chemical	10/10, 05:42 AM UTC	0.64 h	-93.71 h
636017870	container	10/10, 04:07 AM UTC	0.38 h	-1098.12 h
563051300	container	10/09, 11:31 PM UTC	0.55 h	-2396.52 h
235040503	chemical	10/09, 10:34 PM UTC	-0.22 h	0.67 h
244092000	lng	10/09, 09:52 PM UTC	0.19 h	-113.87 h
215746000	chemical	10/09, 04:52 PM UTC	0.72 h	-454.87 h
232019192	container	10/09, 03:39 PM UTC	-0.00 h	—
232019192	container	10/09, 03:22 PM UTC	-0.06 h	2000.63 h
636091569	product	10/09, 01:50 PM UTC	3.00 h	-680.83 h
636015952	general-cargo	10/09, 01:22 PM UTC	0.37 h	-2362.38 h
235007880	chemical	10/09, 12:11 PM UTC	-0.03 h	-3.19 h
232019192	container	10/09, 12:07 PM UTC	0.01 h	2003.88 h
235007880	chemical	10/09, 09:16 AM UTC	0.45 h	-0.27 h
232019192	container	10/09, 09:13 AM UTC	-0.02 h	2006.77 h
636019984	container	10/09, 08:25 AM UTC	0.94 h	-88.68 h
232019192	container	10/09, 06:18 AM UTC	-0.28 h	2009.70 h
311000321	container	10/09, 06:12 AM UTC	0.14 h	-28.13 h
235500000	container	10/09, 04:54 AM UTC	-0.07 h	2011.09 h
311001547	container	10/09, 04:16 AM UTC	0.41 h	-2773.27 h
232019192	container	10/09, 03:17 AM UTC	-0.03 h	2012.71 h
636093154	general-cargo	10/08, 08:19 PM UTC	0.37 h	-2783.32 h
257731000	chemical	10/08, 03:09 PM UTC	2.56 h	-23.66 h
235009956	chemical	10/08, 02:41 PM UTC	0.26 h	-54.70 h
211230400	container	10/08, 01:33 PM UTC	0.40 h	-217.55 h
232019192	container	10/08, 12:42 PM UTC	-0.25 h	2027.30 h

See it on a live desk

Predicted ETAs, congestion and sanctions screening across 51 ports — try it in one click, no signup.

Open dashboard See pricing →

How it's measured

Predicted ETA: distance-to-berth over speed-made-good, refreshed on every AIS position, with a seasonal correction from rolling medians.
Broadcast ETA: the crew-entered ETA in the vessel's AIS message — the de-facto industry default.
Both are compared to the actual arrival timestamp on the same closed voyages. We report RMSE (penalises large misses) and MAE (typical error).
Broadcast ETAs more than 7 days off are dropped as AIS sentinel / placeholder values, and our model is scored on that same set — so the comparison is honest, not cherry-picked.
Coverage is public AIS only; ports with weak coverage show fewer closed voyages. We'd rather show a small honest sample than a big misleading one.