Hoegh Osaka: test the inputs before trusting the stability model

365 ALTERAM analysis

A correct calculation can describe the wrong condition

A technical model can be internally consistent while answering the wrong factual question. In a maritime casualty review, the question is not simply whether software calculates correctly. It is whether the cargo, tank contents and configuration entered into it represent the vessel at the relevant moment. That distinction matters when a calculation becomes the basis for an operational decision or a later account of responsibility.

Independent review should therefore separate three tasks: checking the calculation system, establishing the provenance of its inputs and assessing what the resulting model can actually demonstrate. The Hoegh Osaka investigation provides a public example of why those tasks cannot be collapsed into a single software test.

Public record

The casualty and the limits of the reconstruction

On 3 January 2015, Hoegh Osaka developed a severe list while departing Southampton and turning near Bramble Bank, then grounded. The MAIB’s safety flyer attributes the heavy heel to inadequate stability that had not been identified before departure. It describes discrepancies in cargo figures, omitted cargo height allowances and a practice of adjusting computer ballast quantities to match observed draughts. [2]

The full report records that tests confirmed the validity of the loading computer and its Loadstar program. However, the reported pre-departure calculated condition was not saved. Actual departure ballast distribution could not be confirmed; the investigators used an iterative reconstruction to identify a plausible stability condition. A reconstructed estimate must therefore remain distinguishable from a contemporaneous saved calculation or direct measurement. [1]

The report’s stated purpose was accident prevention, not determining liability or apportioning blame. It also contains an express restriction on its use in proceedings aimed at attributing liability or blame. These safety findings are not findings of fraud or criminal guilt; counsel must assess any proposed legal use. [1]

365 ALTERAM analysis

Create an input register with an evidence trail

For a new authorised review, list each material input alongside its unit, source, creation time and status. A measured weight, a shipper’s declaration, a default value and an investigator’s later estimate should have visibly different labels. Retain the underlying documents and original electronic files so that the register does not become another unsupported summary.

Reconcile cargo at unit level before relying on deck totals. Establish which units were actually loaded, where they were placed and whether amendments reached the person preparing the calculation. A difference in total weight is only one possible discrepancy. Distribution and the assumptions about the height of cargo also need specialist assessment. Avoid treating a final tally as definitive merely because its filename says final.

Tank information needs an equally explicit chain. Identify whether a figure came from an instrument, manual observation, transfer record or adjustment made within software. Ask which records can independently support it. When readings disagree, document instrument condition, timing and subsequent operations before choosing a preferred value. Agreement achieved by adjusting an input is not independent confirmation of that input.

365 ALTERAM analysis

Keep reconstruction separate from observation

Preserve the original model before creating analytical copies. Record software versions, vessel configuration, saved conditions and the source of each replacement value. A qualified naval architect should direct stability calculations; digital specialists can help establish file history and whether the recovered data represent the relevant calculation session.

Test a justified range of uncertain inputs rather than presenting one reconstructed scenario as uniquely established. Explain which conclusions persist across that range and which depend on a narrow assumption. A model that reproduces an observed movement can support a hypothesis, but the reviewer should ask whether materially different input combinations could produce a similar result.

Post-incident measurements require particular care. Recovery operations, transfers and damage may change the condition being measured. Record those interventions and the uncertainty they introduce. Where departure conditions cannot be recovered reliably, a bounded conclusion is more useful than a precise number whose evidential basis cannot be defended.

365 ALTERAM analysis

Report both technical significance and individual knowledge

Finally, distinguish a consequential input error from evidence of who knew about it. Compare document receipt, revisions, instructions and available checking time. An incorrect value may reflect estimation, an uncommunicated change, a processing mistake or deliberate conduct; the evidence must decide between those possibilities. A review may confirm the original assessment, qualify it or remain inconclusive.

365 ALTERAM can reconcile operational records, examine digital provenance and coordinate appropriately qualified maritime specialists. The useful deliverable links each disputed input to its source, quantifies its significance where defensible and identifies unresolved questions. Counsel retains legal strategy. Neither a discrepancy nor an alternative model guarantees a favourable outcome.