Port Congestion Trend

Method & caveats

• Source. Built entirely from the AIS-derived daily snapshot the platform writes per active port (vessels at anchor, average elapsed wait-so-far, anchorage count, berth occupancy). The per-port chart reads that history directly; the verdicts summarise each port's recent window against the prior baseline week. No third-party congestion feed is used.
• The metric. For each port, the centre of the recent window's waiting time is compared against the prior baseline window. A positive change (longer waits / bigger queue) is worsening; a negative change is easing; movement inside a noise band is stable.
• Average wait is elapsed-so-far, not completed. The waiting-days figure is the mean time-at-anchor so far for vessels currently queued (elapsed since each was first geofenced at anchor), refreshed each AIS cycle: a live lower bound on the eventual completed wait, not a backward-looking average of finished queue times. It reads low relative to a port's true completed dwell.
• Basis & fallback. Where a port's queue could be timed, the trend is in waiting-days; where the queue was present but untimed, it falls back to vessels at anchor (labelled accordingly). The two bases are never mixed inside one port's comparison.
• Honest about young history. A verdict is only given when BOTH windows have enough distinct AIS observation days; otherwise the port is shown as “building history” with its recent median but no comparison. Withheld, never fabricated. The observation-day counts are shown per port so the confidence of every verdict is inspectable, and thin windows are flagged low-confidence.
• This is the trend companion to the latest-snapshot Port Congestion desk.