ETA accuracy benchmark · 🇸🇬 Singapore

How Port Flow's predicted ETA compares to each vessel's own broadcast ETA at Singapore, measured on closed voyages over the last 90 days. We publish the number — good or bad. Lower error is better.

Port Flow predicted ETA (RMSE)

49.76 h

MAE 14.54 h · 979 voyages

Broadcast ETA (RMSE)

65.12 h

crew-declared, 337 voyages (same set)

Our advantage

23.6 %

lower error than the broadcast ETA

Broadcast ETAs more than 7 days from the actual arrival are excluded as AIS sentinel / placeholder values (431 excluded). Both figures are computed on the same remaining voyage set — apples to apples.

Closed voyages

error = predicted/broadcast minus actual arrival (hours)


255799000	container	10/12, 04:48 AM UTC	-25.05 h	-3292.81 h
563034300	product	10/12, 04:08 AM UTC	-8.86 h	-74.13 h
563021100	product	10/12, 03:39 AM UTC	6.53 h	7.34 h
563882000	product	10/12, 03:04 AM UTC	-18.67 h	-240.07 h
636015007	product	10/12, 02:55 AM UTC	5.50 h	-2.91 h
538012478	product	10/12, 02:47 AM UTC	8.52 h	-2461.78 h
563021100	product	10/12, 02:45 AM UTC	4.84 h	-45.75 h
525119038	product	10/12, 02:33 AM UTC	4.96 h	7.69 h
563096800	product	10/12, 02:08 AM UTC	4.59 h	-1.90 h
563220900	container	10/12, 02:00 AM UTC	6.56 h	-2411.00 h
249035000	container	10/12, 01:59 AM UTC	7.10 h	-2463.99 h
565102000	product	10/12, 01:45 AM UTC	8.71 h	—
564688000	product	10/12, 01:41 AM UTC	4.42 h	-5.69 h
566373000	chemical	10/12, 01:38 AM UTC	5.64 h	—
538010579	product	10/12, 01:36 AM UTC	4.63 h	-2458.60 h
566530000	product	10/12, 01:16 AM UTC	-7.51 h	-966.28 h
563133100	container	10/12, 01:07 AM UTC	2.56 h	-455.12 h
565173000	product	10/12, 12:52 AM UTC	3.68 h	-1416.88 h
566540000	product	10/12, 12:50 AM UTC	-37.81 h	-33.74 h
538012001	container	10/12, 12:28 AM UTC	5.50 h	-2273.47 h
563038700	product	10/12, 12:18 AM UTC	0.07 h	-0.72 h
563086610	product	10/12, 12:15 AM UTC	0.48 h	-11.31 h
566538000	chemical	10/12, 12:10 AM UTC	-1.44 h	-6.51 h
636013530	container	10/11, 11:59 PM UTC	8.49 h	-261.99 h
563312700	product	10/11, 11:57 PM UTC	7.57 h	-215.96 h
566799000	product	10/11, 11:50 PM UTC	-23.02 h	-137.01 h
563042900	product	10/11, 11:21 PM UTC	3.82 h	—
563291200	container	10/11, 11:19 PM UTC	-40.35 h	-654.33 h
563132900	product	10/11, 11:16 PM UTC	4.00 h	—
566601000	product	10/11, 11:14 PM UTC	-1.31 h	-177.23 h
564616000	product	10/11, 11:13 PM UTC	5.96 h	-140.72 h
247437600	container	10/11, 10:57 PM UTC	8.89 h	-631.97 h
219039000	container	10/11, 10:42 PM UTC	15.42 h	—
352002539	general-cargo	10/11, 10:40 PM UTC	3.59 h	—
431578000	container	10/11, 10:23 PM UTC	4.72 h	-2561.39 h
563056400	product	10/11, 10:23 PM UTC	-4.05 h	-167.38 h
563212500	chemical	10/11, 10:07 PM UTC	4.72 h	-65.13 h
564167000	chemical	10/11, 09:53 PM UTC	4.63 h	6859.10 h
219039000	container	10/11, 09:51 PM UTC	4.98 h	-2082.86 h
563212500	product	10/11, 09:32 PM UTC	4.98 h	-1844.54 h
563085320	chemical	10/11, 09:24 PM UTC	-1.75 h	-53.41 h
566116000	product	10/11, 09:24 PM UTC	-3.59 h	-428.41 h
256267000	container	10/11, 09:21 PM UTC	6.17 h	-308.28 h
636018769	container	10/11, 09:12 PM UTC	5.44 h	—
636018769	container	10/11, 08:21 PM UTC	2.28 h	-34.85 h
563056100	product	10/11, 08:06 PM UTC	6.59 h	-7.12 h
538011015	general-cargo	10/11, 08:02 PM UTC	5.69 h	-2.04 h
564938000	container	10/11, 07:59 PM UTC	2.78 h	-765.49 h
538003172	product	10/11, 07:40 PM UTC	-2.18 h	-2830.68 h
636022117	general-cargo	10/11, 07:39 PM UTC	2.84 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.