CII and Vessel Performance Data: Building the Reporting Backbone Behind the Rating

A Carbon Intensity Indicator rating is a single letter that a charterer, a lender and a flag administration will all read as a fact about your ship. It is only ever as trustworthy as the fuel and distance data underneath it, and most of that data still arrives as a typed noon report.

The Carbon Intensity Indicator has been in force since 1 January 2023, and the letter it produces now travels well beyond the technical department. Charterers write CII expectations into fixtures. Lenders track it against Poseidon Principles portfolios. A D three years running, or a single E, forces a corrective action plan into the ship’s SEEMP before a Statement of Compliance can be issued. That makes the rating a governed number, and a governed number needs a data pipeline you can defend under question, not a spreadsheet that averages whatever the vessels sent in.

What the rating actually consumes

Strip away the presentation and the attained annual operational CII is one ratio. For most cargo ships it is the Annual Efficiency Ratio: total CO2 emitted in the calendar year, divided by the ship’s capacity multiplied by the total distance sailed. CO2 is not measured directly; it is derived from fuel oil consumption by fuel type, each type carrying its own carbon conversion factor. Capacity is deadweight for bulkers, tankers, container and gas carriers, and gross tonnage for cruise and ro-ro passenger ships. Distance is the total sailed over the year.

Two of those three inputs are the same figures a ship already reports into the IMO Data Collection System: annual fuel consumption per fuel type and distance travelled. The attained ratio is then measured against a required value, a 2019-based reference line reduced by an annual factor that steps up each year: 5 per cent in 2023, 7 in 2024, 9 in 2025 and 11 in 2026, with the factors beyond 2026 still to be set by the IMO. The gap between attained and required places the ship into one of five bands, A through E. The reference line moves every year whether the ship changes or not, which is the point most operators miss: standing still means sliding down the bands.

So the whole edifice rests on two quantities per ship per year — fuel burned and miles sailed — and on getting the fuel-to-CO2 conversion right. Everything a supervisor, charterer or auditor eventually challenges reduces to whether those numbers are the true ones.

Where the numbers come from, and where they drift

The honest starting position is that most fleets still assemble the annual fuel and distance figures from daily noon reports: a crew member enters consumption, remaining-on-board, distance run and hours steaming into a form once a day. Noon reports are useful, but they are hand-keyed operational documents, not instrument readings. Digits get transposed. A remaining-on-board figure carries an error forward until the next bunkering resets it. Distance sometimes records the great-circle plan rather than the miles actually run. None of that is fraud; it is what happens when a governed number is sourced from a fatigue-prone manual process.

The corroborating sources rarely agree cleanly. Bunker delivery notes give you fuel loaded, and the sum of daily consumption plus closing ROB should reconcile to opening ROB plus deliveries — but tank soundings, density assumptions and stripping losses mean it seldom balances to the litre. Mass flow meters and automated data-logging give a second, higher-frequency consumption signal that frequently diverges from the noon figure. The task of the reporting backbone is not to pick a favourite source. It is to ingest all of them, reconcile them against each other, and record which number won and why. A rating built by silently averaging conflicting sources is a rating you cannot explain when someone asks how a specific voyage was counted.

This is the same discipline that the EU ETS and FuelEU Maritime emissions pipeline now demands for money rather than a letter grade — once emissions data drives an allowance surrender or a penalty, the tolerance for unreconciled inputs falls to zero. CII is the gentler introduction to the same problem.

Deriving the rating with the outliers left visible

The instinct when a leg looks wrong is to quietly drop it or smooth it into the annual average. Both destroy defensibility. A better pattern computes the ratio from clean legs, flags the suspect ones with a reason, and quarantines them for reconciliation rather than deleting them — so the annual figure is finalised only once every flag has been resolved against a bunker note or a sensor trace. The check below runs the flagging and the derivation in one pass.

-- Attained annual operational CII (AER) from voyage-leg fuel and distance
-- records, with data-quality flagging so a single bad noon report cannot
-- silently move the rating. PostgreSQL.
WITH legs AS (
    SELECT
        imo_number,
        report_date,
        period_hours,
        distance_nm,
        fuel_hfo_mt,
        fuel_mgo_mt,
        -- IMO DCS carbon factors (Cf): tonnes CO2 per tonne of fuel
        fuel_hfo_mt * 3.114 + fuel_mgo_mt * 3.206 AS co2_mt
    FROM voyage_leg
    WHERE report_year = 2025
),
flagged AS (
    SELECT
        *,
        -- implied speed sanity check against the reported steaming hours
        distance_nm / NULLIF(period_hours, 0) AS avg_speed_kn,
        CASE
            WHEN distance_nm <= 0                       THEN 'zero_distance'
            WHEN co2_mt <= 0                            THEN 'zero_fuel'
            WHEN distance_nm / NULLIF(period_hours,0) > 26
                                                        THEN 'implausible_speed'
            ELSE 'ok'
        END AS quality_flag
    FROM legs
)
SELECT
    v.imo_number,
    v.capacity_dwt,
    SUM(f.co2_mt) * 1e6                                   AS co2_grams,
    SUM(f.distance_nm)                                    AS distance_nm,
    -- AER: grams CO2 per dwt-nautical mile, on reconciled legs only
    (SUM(f.co2_mt) * 1e6)
        / (v.capacity_dwt * SUM(f.distance_nm))          AS attained_aer,
    -- how much of the year is still awaiting reconciliation
    COUNT(*) FILTER (WHERE f.quality_flag <> 'ok')       AS legs_quarantined
FROM flagged f
JOIN vessel v USING (imo_number)
WHERE f.quality_flag = 'ok'         -- flagged legs are held for review, not dropped
GROUP BY v.imo_number, v.capacity_dwt;

The carbon factors are the standard IMO conversion values for heavy fuel oil and marine gas oil; a real pipeline carries the full fuel table. The important design choice is the last two lines: the attained ratio is reported alongside a count of legs still in quarantine, so nobody can sign a rating while a meaningful slice of the year is unresolved. The rating is provisional until legs_quarantined reaches zero.

Lineage is the part that makes it defensible

A clean number is not the same as a defensible one. Defensibility is being able to take the published rating and walk it backwards: this attained AER came from these annual totals, which came from these legs, each leg sourced from this noon report or this sensor value, reconciled against this bunker delivery note, corrected on this date by this person for this reason. That chain is lineage, and it is the artefact a verifier actually tests. The same expectation now runs through structured regulatory reporting generally — the metadata and format demands behind portals like ESAP exist precisely so a reported figure can be traced and validated rather than taken on trust.

Concretely, the backbone should store every raw report immutably on arrival, keep every correction as a new versioned record rather than an overwrite, stamp each derived figure with the input identifiers and rule versions that produced it, and log who resolved each flag. When a charterer disputes a rating, or a flag state asks how a borderline C-to-D case was calculated, the answer is a query, not a fortnight of email archaeology. Fleets that already run digital documentation flows — the kind moving towards electronic bills of lading and connected trade systems — have most of the plumbing to do this; the gap is usually that emissions data was treated as reporting exhaust rather than as a governed record.

The reduction factor tightens every year and will tighten again beyond 2026 once the IMO sets the next steps. Ships that sit near a band boundary will cross it on the strength of a few tonnes of fuel and a few hundred miles — exactly the quantities most exposed to a mistyped noon report. Build the pipeline so the letter is auditable back to its source, and the rating survives scrutiny. Leave it as a spreadsheet average, and the first serious question about a single voyage is the one you cannot answer.