A cluster is an observation, not an explanation

A spatial cluster means events occurred relatively close together. A temporal cluster means they occurred within a short period. A ship-characteristic cluster means several affected vessels share an observed attribute.

Each can generate a hypothesis. None proves causation.

Events may cluster near Hormuz because traffic is physically concentrated there. They may cluster on a busy day because more ships are exposed. Several tankers may appear in a casualty list because tankers form a large share of relevant traffic, not because every tanker is selected for its cargo.

A cluster can also result from operational conditions, reporting practices, route openings, weather, military activity or delayed confirmation. UKMTO advisories sometimes include time-late reports. Plotting them by publication date rather than occurrence date can create a false burst.

The correct first statement is descriptive: “Five confirmed incidents occurred in this area during this interval.” A causal statement such as “loaded tankers were deliberately selected” requires additional evidence.

Build a clean event dataset first

The analyst should assign one event identifier and preserve all linked source records. A master’s report, UKMTO warning, JMIC update and later IMO entry may describe the same occurrence.

Deduplication should compare:

- vessel name and IMO number; - UTC event time and report time; - position and movement direction; - event type and consequence; - warning or incident reference; - source chain; and - later corrections or reclassifications.

Names alone are insufficient because a vessel can be attacked more than once. IMO’s current list explicitly identifies some second attacks. Conversely, name withholding can cause one event to look like several unidentified casualties.

The dataset should separate occurrence date from confirmation and publication dates. It should preserve “unknown” rather than filling missing flag, cargo, load condition or direction from assumption.

Counts must state geographic and temporal scope. “Hormuz incidents” might mean the charted Strait, a wider security area or the entire Gulf region. Changing the boundary can change the apparent cluster.

Use the correct denominator

An incident count is a numerator. A targeting-rate inference also needs the number of comparable vessels exposed.

Suppose six tankers and two bulk carriers are damaged. That does not establish higher tanker risk if 300 tankers and 20 bulk carriers traversed the assessed area. Nor can those illustrative numbers be presented as real data.

A useful denominator can be vessel transits, vessel-hours in the exposure area or ship-days at specified anchorages. The choice should match the hypothesized mechanism. Boarding exposure may relate to time and speed; a strike near a route may relate more closely to passages through a geographic window.

UKMTO’s November 13 overview draws on commercial AIS, geospatial analysis, vessel and industry reporting, open sources and partners. It says its record is not complete, reporting is voluntary and some incidents cannot be independently verified. JMIC transit comparisons also identify separate datasets and exclusions. These caveats make the figures useful but not interchangeable.

AIS-derived exposure can miss ships with transmissions disabled, disrupted or absent. It can double-count tracks, misclassify activity or reflect spoofing. Analysts should disclose filters such as minimum deadweight, Iranian-port exclusions and whether facilitated and non-facilitated movements are combined.

Control for changing exposure

Traffic composition changes during a crisis. Some operators stop sailing, others wait, and some cargoes move by pipeline or ship-to-ship arrangements. A share calculated during one month may not apply in the next.

EIA estimated that total oil flow through Hormuz fell from 21.6 million barrels per day in the fourth quarter of 2025 to 5.1 million in the second quarter of 2026, then recovered to 9.1 million in the third quarter. These are quarterly flow estimates based partly on tanker tracking, not vessel counts. They nevertheless show that exposure volume changed sharply.

If incidents rise while traffic also rises, the cause may be increased opportunity, increased threat or both. If traffic falls and incident counts remain steady, the rate per exposed transit may have increased. Neither conclusion is valid until compatible numerator and denominator periods are used.

Waiting ships create another exposure pool. Anchorage congestion can concentrate stationary targets outside the transit lane. Mixing anchorage events with underway transit events can obscure the operational control being assessed.

Direction, load condition and vessel class should be analyzed only when those fields are reliably available for both affected and exposed populations.

Distinguish selection evidence from capability

A hostile actor may have the capability to observe, hail or strike many ships. That does not establish the selection rule used in a particular event.

Stronger targeting evidence can include an authenticated statement of intent, repeated communications naming a ship or association, consistent selection after controlling for exposure, forensic linkage, or an official intelligence assessment. Each still needs source and confidence.

A ship’s flag, ownership, charterer, cargo, port history or public profile may be relevant to a vessel-specific threat assessment. The analyst should phrase the relationship carefully: “may increase exposure under the current assessment,” rather than “will be attacked.”

An apparent pattern can be confounded. Large vessels may be easier to identify. AIS-visible ships may be easier to track, but switching off AIS does not make a vessel invisible and can create collision risk. Laden and ballast ships may travel in different directions or time windows.

The absence of a hit among one group is not proof of protection, especially with a small sample.

Use clusters for action without claiming causation

Operators do not need a final targeting theory to respond to a credible cluster.

A recent group of incidents near the intended route can trigger:

1. a fresh UKMTO, JMIC, flag and naval-advice check; 2. route and transit-window revalidation; 3. updated vessel-specific threat and vulnerability assessment; 4. enhanced lookouts, machinery readiness and damage-control posture; 5. revised waiting or abort criteria; 6. insurer and charter-party notifications where required; and 7. a new review time as the cluster evolves.

The action should be tied to the observed condition. “Three incidents inside the planned sector within 24 hours” is a defensible trigger. “Attackers prefer our competitor’s fleet” is not, unless supported by evidence.

Masters should receive the location, time, verified event type and uncertainty. They should not receive a speculative story that biases watchkeepers toward one threat while they miss another.

Clusters should be rechecked after corrections. A withdrawn report, changed position or duplicate can materially alter a small sample.

Protect commercial decisions from pattern overreach

Owners may be tempted to market a ship as lower risk because its flag or cargo has not recently appeared in incident reports. That claim can create unsafe confidence and may become a disclosure issue if contrary evidence exists.

Charterers should compare voyage alternatives using current route exposure, vessel controls, delay and contract terms. A supposed “safe class” is not a substitute for the actual assessment.

Underwriters can use clusters to revise accumulation zones and request information. Pricing a causal targeting factor requires stronger evidence than mapping claims. Policies and quotations should be read for their actual geographical and operational conditions.

Brokers and traders should date every pattern statement and distinguish count, rate and forecast. A rate requires a denominator. A forecast requires assumptions. Neither should be presented as a verified fact.

ZeaClub can link incident records, deduplicate sources, normalize time and location, and pair events with compatible transit data. It can show confidence intervals and missing fields. It should refuse to rank flags or ship types when the exposure denominator is unavailable.

Commercial conclusion

Attack clusters are valuable warning signals. They identify where and when the operating picture deserves immediate review.

They are weaker evidence of who or what is being selected. That conclusion requires clean event records, compatible exposure data, controlled comparisons and credible attribution evidence.

The commercially disciplined response is to act on the verified cluster while labeling the targeting theory as a hypothesis. Owners and masters gain useful caution without false certainty, and charterers, insurers, brokers and traders avoid building decisions on a pattern the data cannot yet support.

This article distinguishes verified source material from ZeaClub analysis, recommendations, forecasts and hypothetical scenarios.

Sources