When Ship Counts Stop Measuring Access: Visibility, Cargo Flow and Commercial Reliability
By Lim Wei Sheng, Head of Maritime & Logistics Research
Three kinds of number describe the Strait of Hormuz, and they disagree by an order of magnitude. AIS-visible transits print at roughly 5 per cent of their June-2023 reference, while Kpler's closed-window flow accounting read near normal for the 17 June–17 August window. They are not rival estimates of one quantity but counts of different populations — never to be added, averaged or spliced.
That arithmetic sets the decision rule for any organisation exposed to the Gulf: do not trigger rerouting, force majeure or a contract action on a first-print transit count. Run the six-field check first — observation date, publication date, population, units, revision status, confidence — because first prints on these series are floors that firm over several days.
What Can Be Observed: Three Populations, Not One Series
The September 2026 Gateway Access Monitor runs three Hormuz series side by side and states plainly that "the series are never spliced":
- IMF PortWatch AIS transits — an August mean of 4.7 vessels a day, roughly 5 per cent of the 102-a-day June-2023 reference — explicitly a floor, with industry estimates putting most residual movement AIS-dark.
- Kpler visible-commodity counts — two to nine a day over 5–8 September, counting identifiable cargoes.
- JMIC-relayed US NCAGS counts — near 30 US-facilitated transits a day over 3–5 September, a third population.
Our Contested Baseline analysis covered inputs versus outputs in official statistics; this is the live-tracker counterpart — population arithmetic.
First Prints Are Floors That Firm
The Gateway Access Monitor's dated updates record the 9 September Hormuz observation revised from seven transits to eleven — "the third consecutive daily observation revised upward" — with its rule that "first-print counts on this series are floors that firm over several days". The 16 September update adds a Monday series-low first print of four transits, against a Kpler 10-day average reading ten to twelve by vintage. At Bab el-Mandeb, an intraday capture of six on 10 September settled at twenty-six for the full day, recorded in the 11–12 September updates. That is a population change — partial-day counting maturing into full-day counting — not a revision. Al Jazeera's reporting corroborates the divergence, with Kpler dailies of five, six and eleven — corroboration of the divergence at a different vintage, not a reading to splice with the above. Monthly benchmark indices are the mirror image: they trade timeliness for population stability.
What Actually Moves: Flow-Normal Is Not Access-Normal
Kpler's closed-window flow accounting, in the same edition, put all-liquids-plus-LPG movement at roughly 8.8 million barrels a day for the 17 June–17 August window, later updated to roughly 8.6 million (31 August, press-reported) with dark transits and ship-to-ship transfers included. No comparable closed-window estimate yet exists for any September window. Flows near normal to mid-August while observed transits read roughly 5 per cent: both are true, because they count different populations. In our reading, movement can coexist with unreliable access — a flow estimate says whether oil moves in aggregate, not whether your vessel can secure insured passage.
Who Can Secure Passage, at What Cost
War-risk cover is quoted at 7.5–12.5 per cent of hull value per transit — the upper edge a press-reported indication in the Gateway Access Monitor's 8 September dated update, with public reporting corroborating 7.5–10 per cent against a pre-conflict norm near 0.25 per cent. Iranian transit fees run US$150,000–2 million on the edition's reported band, and payment is OFAC-sanctionable while simultaneously voiding Lloyd's war-risk cover under the market's model transit-fee clause — compliance with one gatekeeper breaches the other. Several carriers have suspended transits outright.
Nor is the overland bypass a bypass this month. The East-West pipeline to Yanbu shut on 10 September after drone strikes on pumping stations. Officials briefed on the damage put repair at three to five weeks; traders put export cover from Yanbu stocks at roughly five to seven days. Both are attributed estimates from the Gateway Access Monitor's dated updates, and the buffer depends on withdrawal assumptions. By 16 September, Yanbu loadings were suspended and some September cargoes cancelled.
The Six-Field Check, Worked
On 16 September the Gateway Access Monitor announced methodology v1.1 for its 1 October edition — an evidence-confidence label per fully assessed gateway, plus an annex documenting the revision behaviour of high-frequency sources — and moves to publishing on the first of each month. The table applies the check those labels formalise to this month's data.
| Data point | Observed → published | Population | Units | Revision status | Confidence |
|---|---|---|---|---|---|
| Hormuz transits: 7 → 11 | 9 Sept obs → revised in 12 Sept update | Kpler visible-commodity daily counts, press-relayed, single-day vintage | vessels/day | Third consecutive upward revision | Low at print; medium firmed |
| Hormuz series low: 4 | 14 Sept → 16 Sept update | Kpler visible-commodity daily counts, press-relayed, single-day vintage | vessels/day | First print; expected to firm | Low; not decision-grade alone |
| Bab el-Mandeb: 6 → 26 | 10 Sept obs, intraday → settled full-day (11–12 Sept updates) | Partial-day vs full-day transits | vessels/day | Population change, not a revision | Very low intraday; medium settled |
| Kpler flows: ~8.8 → ~8.6M | 17 Jun–17 Aug window → Sept edition; updated 31 Aug, press-reported | All liquids + LPG incl. dark, ship-to-ship | million b/d | Closed window; revised once | Medium-high |
| War-risk cover: 7.5–12.5% | As at 8 Sept → 8 Sept dated update | Quoted indications, not bound rates | % of hull per transit | Upper edge single-sourced; 7.5–10 corroborated | Medium; weaker at upper edge |
Revision behaviour is itself assessable: as of 18 September the corrections log recorded 16 entries — seven corrections, nine updates — across the Gateway Access Monitor's two editions, under the rule that changes caused by documents issued after publication are updates, not corrections. Distinguishing one's own errors from the world changing is the mark to look for in any provider — us included.
Update (20 September 2026): Two developments since publication test the framework. First, on 19 September the commander of the United States Central Command (CENTCOM), the American military command for the Middle East, claimed a cumulative milestone. More than one billion barrels of crude, he said, have moved out through the Strait, with over 2,000 commercial transits assisted. Run the six-field check. The window is stated only as "the last couple of months". The population is the command's own facilitation ledger — nearest kin to the JMIC-relayed NCAGS counts above, and distinct from the PortWatch and Kpler series. The underlying records are not independently verifiable. Aggregate movement, however large, still does not answer whether an eligible shipper can obtain repeatable passage at an acceptable cost. Second, the table's series low of four has since revised to seven, exactly as its "expected to firm" cell anticipated. The Monitor's 20 September update adds that no daily count below four has survived seasoning.
What This Means for Decision-Makers
- Run the six-field check before any count-triggered action. A first print on these series is a floor — the last three daily Hormuz observations all revised upward — and no rerouting, force-majeure or contract decision should key off one.
- Never splice the series in internal dashboards. PortWatch, Kpler and NCAGS counts answer different questions over different populations; the Gateway Access Monitor itself refuses to splice them.
- Price access, not throughput. At up to 12.5 per cent of hull per transit, with a fee regime both sanctionable and cover-voiding and carriers suspending service, "flows near normal" is compatible with your cargo having no insurable path.
- Assess providers by revision behaviour. Prefer sources that date updates, keep a corrections log and distinguish corrections from updates — then hold internal estimates to the same standard.
Evidence and data cut-off: 18 September 2026; updated 20 September 2026 with the 19 September CENTCOM statement and subsequent Gateway Access Monitor revisions. Third-party figures are as published by the cited sources; Gateway Access Monitor readings are the World Research Institute's own. See our Research Standards.
The World Research Institute provides maritime and logistics intelligence, chokepoint risk monitoring, and evidence-led competitive analysis. Contact our team to commission tailored research.