Voyage Performance Surveillance - Applied AI Foundation

Voyage Performance Surveillance pipeline reading position, weather, and operational data into a CP-aware leg-by-leg performance view

A voyage isn’t a single thing. It’s a sequence of legs — sea passage, port approach, port stay, cargo operation, departure, sea passage again. Each leg has its own performance shape, its own commercial implications, and its own way of going wrong. The vessel that arrives at Singapore three days late may have been below CP speed, or above CP speed but routed through bad weather, or stuck in a charterer-induced port delay — and the appropriate conversation is different for each. The Voyage Performance Surveillance pipeline reads the noon-report stream, the position track, and the Charter Party terms together, and produces a leg-by-leg performance view that resolves the ambiguity.

Where the data comes from

Data	Source	Cadence
Position	AIS (Automatic Identification System) + noon report	Continuous + daily
Speed	Noon report	Daily noon-to-noon
FOC by grade	Noon report	Daily
Weather (wind, sea, current)	Noon report + Stormglass + NAVTOR routing feed	Daily
Routing analysis	NAVTOR (route optimisation)	Per voyage
Sea time / port time	ERP voyage log	Per voyage
Charter Party terms	CP file in vessel records	Per charter
ETA vs laycan	Voyage plan in ERP	Continuous
Port call data	ERP port module	Per port call

Position and weather come from external feeds (AIS, Stormglass, NAVTOR); operational data (FOC, sea/port time, voyage log) comes from the ERP and the vessel’s noon report stream. The Charter Party file lives in the vessel record and gets re-read on every analysis run so changes mid-voyage are reflected immediately. Two map views keep the position context honest:

Vessel detail
Fleet view

Single vessel with current position pinned and past 7 days of track history overlaid. Used when reviewing one voyage in detail.

Charter Party compliance — the central question

Most voyage analysis questions reduce to “is the vessel performing within Charter Party?” The CP defines a “good weather” envelope and a warranted speed-and-consumption profile within that envelope:

GOOD WEATHER ENVELOPE (typical CP)
   Wind:    Beaufort scale ≤ 4
   Sea:     Douglas sea state ≤ 3
   Current: ≤ ±0.5 knot favourable / adverse
   Swell:   ≤ 2 m

WARRANTED PERFORMANCE (within envelope)
   Speed:           14.0 kn (laden) / 14.5 kn (ballast)
   FOC at sea:      28.5 MT/day VLSFO
   FOC in port:      3.2 MT/day MGO

The pipeline filters every voyage leg to its good-weather subset, then computes the per-leg deviation:

\Delta V = V_\text{actual} - V_\text{CP}

\Delta \text{FOC} = \text{FOC}_\text{actual} - \text{FOC}_\text{CP}

A negative speed deviation under good weather with a positive consumption deviation is the textbook under-performance pattern that drives owner-vs-charterer disputes. A negative speed deviation in bad weather isn’t a CP issue — the warranted speed doesn’t apply.

Why weather filtering matters

Naively averaging speed across an entire voyage produces meaningless numbers. A vessel that hit a Beaufort 8 storm for two days and ran at 9 knots will look like a CP underperformer even if it ran at 14.2 knots through every good-weather leg. The pipeline applies the filter first, then runs the maths:

def good_weather_legs(legs, cp_envelope):
    """Filter legs to those within CP good-weather definition."""
    out = []
    for leg in legs:
        if leg["wind_bf"]   > cp_envelope["wind_max"]:    continue
        if leg["sea_ds"]    > cp_envelope["sea_max"]:     continue
        if abs(leg["current_kn"]) > cp_envelope["current_max"]: continue
        if leg["swell_m"]   > cp_envelope["swell_max"]:   continue
        out.append(leg)
    return out

def cp_performance(good_legs, cp):
    """Compute mean deviation of speed and FOC under good-weather conditions."""
    if not good_legs:
        return {"verdict": "Insufficient good-weather data", "coverage": 0.0}
    n = len(good_legs)
    delta_v   = sum(leg["speed_kn"]  - cp["warranted_speed"]   for leg in good_legs) / n
    delta_foc = sum(leg["foc_mt"]    - cp["warranted_foc"]     for leg in good_legs) / n
    return {
        "delta_v":     delta_v,
        "delta_foc":   delta_foc,
        "good_legs":   n,
        "coverage":    n / len(legs),
    }

Coverage matters because a voyage with only 10% good-weather legs has too small a sample to support a CP claim either way. The pipeline reports coverage alongside the deviation — a TSI sees both.

ETA vs laycan

Beyond CP performance, the timeliness question:

Metric	Calculation
Projected ETA	Current position + remaining distance / projected speed
Time to laycan start	$t_\text{laycan\_start} - t_\text{ETA}$
Time to laycan end	$t_\text{laycan\_end} - t_\text{ETA}$

Verdict:

Condition	Verdict
ETA before laycan start with margin	Early — risk of demurrage if charterer not ready
ETA inside laycan	On-window
ETA after laycan end	Late — penalty exposure

The pipeline updates the ETA every noon report so a TSI sees drift as it happens, not at port-call time.

Claim probability

A composite that combines the deviation magnitude, the good-weather coverage, and CP wording sensitivity:

P_\text{claim} = w_1 \cdot \overline{|\Delta V|} + w_2 \cdot \overline{|\Delta \text{FOC}|} + w_3 \cdot (1 - C_\text{good-wx})

The third term is what protects against weak claims. Low good-weather coverage means the data is insufficient to support the deviation either way — the score reflects that.

$P_\text{claim}$ band	Action
Low	Routine — log and continue
Medium	Owner brief — document for next charter
High	Active claim or counter-claim risk — engage commercial
Critical	Significant exposure — legal advisory

Slow-steaming and bad-weather routing

Two patterns the pipeline flags specifically:

Slow-steaming

A vessel deliberately running below CP speed (typically owner-instructed for fuel saving). Looks like underperformance unless the CP allows slow-steaming. The pipeline checks the CP terms and the master’s voyage instructions — if both agree, the leg is excluded from the deviation calculation.

Bad-weather routing

When a master chose a longer route to avoid weather. Shows up as a longer voyage at lower average speed. Cross-references the fuel oil pipeline for FOC over the routed leg vs the rhumb-line projection — sometimes a longer route is the lower-cost option even though it looks slower on the speed metric.

Worked example: a CP-claim case on `MV POSUN`

Voyage: Singapore to Rotterdam, 21 days at sea, laden.

Leg type	Days	Avg speed	Avg FOC	Notes
Good weather	14	13.2 kn	30.6 MT	Below CP (14.0 kn / 28.5 MT)
Bad weather	5	11.4 kn	32.8 MT	Excluded from CP calculation
Slow-steaming (instructed)	2	11.0 kn	27.4 MT	Excluded

Filtered to good-weather legs only:

Metric	Value
Good-weather coverage	67%
$\overline{\Delta V}$	-0.8 kn
$\overline{\Delta \text{FOC}}$	+2.1 MT/day

Verdict: Sustained underperformance under good weather — claim probability HIGH. The pipeline:

Generates a CP-claim brief with the per-leg numbers, weather classification, and the deviation maths.
Routes to the Commercial Operator (claim handling) and Technical Superintendent (root cause — fouled hull? Engine issue?).
Cross-references with ME performance for engine condition signals and fuel oil for fuel-quality signals.

What the senior review contains

Voyage map — current position, 7-day track, ETA marker.
Leg-by-leg table — distance, speed, FOC, weather classification per leg.
CP compliance — good-weather coverage, $\Delta V$ , $\Delta \text{FOC}$ , verdict.
ETA vs laycan — projected arrival, demurrage / despatch implications.
Claim probability — composite score with contributing factors.
Bad-weather routing assessment — if a routing decision occurred, the cost-benefit analysis.
Cross-domain notes — engine, fuel, hull-fouling signals that touch performance.
Recommendations — operational, owner-communication, commercial actions.
Escalation decision — to whom, and why.

Escalation triggers

Trigger	Severity
Speed deficit >5% under good weather sustained 3+ legs	HIGH
Consumption excess >5% under good weather sustained 3+ legs	HIGH
ETA threatening laycan end	HIGH
Claim probability score = HIGH or CRITICAL	CRITICAL
Pattern across 3+ consecutive voyages	CRITICAL
Weather routing decision adding >10% to passage time	per case

Why the pipeline reads CP terms first

A voyage analysis without CP context misses the entire commercial dimension. A 0.5-knot deficit may or may not be a claim depending on CP wording — some CPs allow that as natural variation; others penalise it as breach of warranted speed. The pipeline reads the CP first, computes against the CP-defined envelope, and produces verdicts that map to the wording the lawyers will eventually argue over.

Voyage analysis is one of the few places where the same numbers are read by completely different audiences — owner, charterer, commercial operator, master, TSI. The pipeline keeps all of them in one document so when the dispute starts, everyone is reading from the same numbers.

References

Templates: eta-voyage

ETA-voyage suite — single-vessel position with 7-day track and the fleet-wide map view.

Related: Fuel oil

Consumption deviation traces back to fuel quality or quantity in many cases.

Related: ME performance

Sustained underperformance under good weather often correlates with cylinder asymmetry or turbocharger fouling.

Related: Emissions

CII attained value is voyage-driven; ECA exposure depends on the route.

Documentation Index

​Where the data comes from

​Charter Party compliance — the central question

​Why weather filtering matters

​ETA vs laycan

​Claim probability

​Slow-steaming and bad-weather routing

​Slow-steaming

​Bad-weather routing

​Worked example: a CP-claim case on MV POSUN

​What the senior review contains

​Escalation triggers

​Why the pipeline reads CP terms first

​References

Templates: eta-voyage

Related: Fuel oil

Related: ME performance

Related: Emissions

Where the data comes from

Charter Party compliance — the central question

Why weather filtering matters

ETA vs laycan

Claim probability

Slow-steaming and bad-weather routing

Slow-steaming

Bad-weather routing

Worked example: a CP-claim case on `MV POSUN`

What the senior review contains

Escalation triggers

Why the pipeline reads CP terms first

References