International shipments: track milestones and delays

How to assemble a shipment from individual segments, preserve status sources, and promptly notify responsible parties of deviations.

By Artem Korsakov ·

Theoretical example. A container with equipment departed from Shanghai on 3 October, due in Rotterdam on the 10th. The shipping line shifts the vessel's port call to the 14th; the notice reaches the agent and sits in their inbox. The train planned for the container departs without it, with the next one leaving five days later. The client learns of the one-week delay only when calling independently. Throughout this time, the card displays 'in transit'.

International logistics automation is needed so that a delay in one segment immediately shows which subsequent connections are affected and who needs to be notified.

Editorial illustration: International logistics automation: milestones and deviations
Editorial generated illustration for the article topic. Not a photograph of an actual project.

Why international shipments require a different accounting model

Different companies and systems participate in the international chain. The shipper confirms cargo readiness, the forwarder organizes an agreed set of services, the carrier executes their segment, the terminal processes the cargo, the customs broker acts within the mandate and applicable regulations. The importer makes commercial decisions and provides necessary information. Specific distribution of obligations is determined by contracts, authorities, and applicable law.

Therefore, you cannot assign a single 'responsible for transport' and consider the issue closed. For each stage, you need an executor's contact person, a person within your team, and a clear escalation procedure. If the agent has not confirmed booking, this is the logistician's task. If additional costs are not agreed upon, a decision may be required from the client or manager. Different tasks must not enter a single queue without prioritization.

One shipment, multiple segments and objects

Start with a general card, but do not attempt to place the route in a single text field. A shipment may include road pickup, sea segment, transshipment, and warehouse delivery. On each segment, the executor, transport document number, point of origin, and control event differ. The container, cargo batch, and vehicle do not always correspond one-to-one.

For the first version, it is useful to preserve:

  • shipment internal number and links to the customer order;
  • cargo units, weight, volume, and handling features;
  • containers and their link to batches;
  • route segments, carriers, and agents;
  • planned dates, forecasts, and confirmed events;
  • documents, tasks, contacts, and agreed costs.

If cargo is split or combined, the link must remain visible. Otherwise, updating one container would incorrectly change the status of the entire batch. A full consolidation model can be postponed, but limitations for the first version should be defined before launch. For a pilot, it is reasonable to choose shipments with the same route scheme and without complex batch splitting.

Plan, forecast, and fact are stored separately

A date in the schedule, a fresh arrival forecast, and confirmed arrival solve different tasks. ETA (Estimated Time of Arrival) means the expected time of arrival. When a new ETA replaces the previous one, history must be preserved. The event 'unloaded' appears only after confirmation from an agreed source. An expected date does not create it.

Next to the event, display the time of the event itself, the message receipt time, timezone, and source. For a human, local port or warehouse date is convenient; for comparison, a unified temporal basis is needed. Lack of update does not mean the cargo has stopped. It means the system has not yet received confirmation.

Подпись: Illustrative example. The ship's new ETA does not match the planned train departure. The system assigns a task to the logistician to check the connection, making no promises to the client.

Segment data comes from various places: shipping line emails, port agents, rail wagon operators, terminal, and last-mile carriers. Each source has its own update delay. Therefore, next to each event, the system shows where it came from and when.

Connection control is more important than a pretty map

Multimodal transport requires control over the transfer between participants. The ship's arrival does not mean the container is available for pickup. After unloading, there must be clearance; after clearance, a truck and warehouse slot are needed. The screen should show the next necessary action and its start condition.

Hypothetical example: The container is unloaded, but confirmation of availability for pickup is missing. The client's portal shows the fact of unloading and the expected next stage. In the internal interface, the logistician sees a task to clarify availability with the agent. The system does not tell the client 'delivery tomorrow' until this is confirmed.

Separately agree on control deadlines: submission of instructions, documents, and weight information, cargo availability, end of agreed free periods. The set depends on the route, transport, and contract. A warning must contain the deadline, source, task owner, and action. A single red icon is not enough.

Deviation queue and notification rules

A good notification answers four questions: what changed, what it affects, who must react, and by when. For a shipment delay, this might be checking the next connection and agreeing on a new plan. For a missing document, you need to contact a specific participant. Sending the signal to all employees is useless.

In the deviation registry, preserve the reason, consequences, decision, approver, and date of next review. Not every ETA update needs to be sent to the client. Notifications can be sent for an agreed deviation from the promised window, route change, or need for a decision. The threshold is set by business and should not be passed off as an industry norm.

Provide for exchange failures. A repeated message does not create a duplicate shipment; temporary unavailability of the source is visible to the user; manual correction is signed by an employee's name. If a later import overwrites a fix without warning, trust in the registry will quickly disappear. Disputed data is better placed in a reconciliation queue.

Where AI helps: collect notifications into a verifiable chronology

A narrow useful scenario: agent emails and transport notifications enter a queue; the model extracts the shipment number, segment, event type, date, timezone, and change formulation. The output is a table of candidates for the event with links to sources. The logistician confirms the link to the shipment, distinguishes between forecast and fact, and only then updates the operational account.

Hypothetical input: 'ETA moved from 12 to 15 November; connection subject to confirmation'. Result: new expected date 15 November, year and timezone require clarification, connection not confirmed. Next action: request confirmation and check impact on delivery. AI does not draw missing segments or send promises to the client.

What is already in your software

First, check the used TMS (Transport Management System). Check whether your TMS supports splitting a multimodal order into legs. If the necessary accounting is already working, we keep it. Email and carrier portals remain sources, and a small AI tool prepares verifiable updates. Transfer to the existing system is possible via verified exchange or agreed upload; interface availability is checked separately. There is no need to buy a new TMS for a single exception queue.

If sources contradict each other

A later received message does not necessarily contain fresher information. For example, today an agent forwarded a letter with yesterday's ETA, while the shipping line portal already shows a new forecast. Save the event date, source message date, and receipt time separately. Agree on priority rules for each type of event, and output conflicts to the logistician.

A fix might also cancel a previous fact: a container number was linked to the wrong shipment, or confirmation refers to another segment. Records must never be deleted without an audit trail. The history shows which record was withdrawn, by whom, and why; the client summary is built only on current permitted events. This scenario is good for testing AI: it explains the conflict and does not choose a convenient date.

How to launch a small working product

For a pilot, choose one direction, one team, and a limited range of sources. Initially, you can enter some events manually and load agreed tables. Then connect available exchange interfaces. It is important to check access, usage terms, costs, and data completeness in advance. The promise 'we will connect all lines' without such verification is not suitable for project evaluation.

Define roles as necessary. Customers have access to their shipments and agreed comments. Agents can be given specific segments. Internal procurement rates and margins must not appear in customer exports. Administrators need a change log, backup, and recovery procedure after erroneous updates.

Acceptance criteria are checked on scenarios: connection delay, late event, duplicate, loss of source access, and manual fix. Measure response preparation time to the client, number of shipments without next action, and duration of open deviations. Delivery time depends on external participants, so attributing its improvement to a single portal is impossible.

Which events change a specific shipment

For each shipment, determine in advance the events that change exactly it. For example: connection delay, storage extension, price change for a separate segment, or payment delay. For each, record the signal, affected stages, possible costs, and the person making the decision. Probabilities are not assigned without grounds.

AI can parse notifications and prepare a list of changes with sources. After logistician confirmation, standard rules update dates, tasks, and budgets. Thus, information about external instability turns into verifiable work with deviations. The portal itself does not manage geopolitics or guarantee a route, but it helps not to lose the received warning.

Next step

If you need control over international shipments, start with one route scheme: collect stages, sources, and exceptions. After this, you can evaluate the registry or portal with clear boundaries of the first version.

Related reading: International shipment budget: expense forecast and margin in transit; Documents for international shipment: completeness and versions.

Частые вопросы ### Can I start without integration with carriers?

Yes. A registry, roles, manual confirmation of stages, and file loading can be a useful first version. The screen must honestly show the source and freshness of information. Automatic connections are added after access verification.

Will the portal replace a forwarder?

The portal helps collect information and make decisions. It does not assume the responsibilities of a forwarder nor confirm legal facts on behalf of authorized participants.