Trade-fair directories look like simple lists of events. Operationally, they are changing datasets: dates move, venues reopen under new names, editions are cancelled, and subsidy rules are revised after an event has already been announced. A directory can be technically online and still be wrong in ways that are expensive for users.

The usual response is to add a “last updated” date. That date is useful, but it is not a freshness model. It does not tell you which source changed, which records conflict, or which pages deserve a human review first. A better approach is to treat staleness as a queue with explicit signals and evidence.

This article describes a four-signal refresh queue that can sit on top of an existing exhibition directory. The same pattern works for event calendars, venue databases and policy-reference pages.

Start with a small set of freshness states

Before scoring anything, define what “verified” means. A practical set is:

  • Confirmed — the current date, venue and edition are supported by a primary source with a retrieval timestamp.
  • Inherited — the record came from a trusted import, but the current edition has not been checked directly.
  • Unverified — a source exists, but its date is old or its scope is unclear.
  • Conflicting — two sources disagree on a material field such as date, venue or edition.
  • Retired — the event or page is no longer valid and should redirect or return a clear status.

Do not collapse these into one boolean. A page can have a fresh “last checked” timestamp and still be conflicting if two official pages disagree. The state should describe the evidence, not the confidence of the person who edited the record.

The four signals

1. Primary-source mutation

The strongest signal is a change at the source that controls the fact: the organizer’s event page, the venue calendar, or the official policy notice. A change detector does not need to scrape every page every day. It can hash the relevant fields, compare a normalized snapshot, and create a review task only when the hash changes.

The queue should record which field changed, not merely that a page changed. A navigation update is not the same as a moved event date.

2. Expiration horizon

Facts age differently. A venue address may remain stable for years. A show date may be provisional until the organizer publishes the next edition. A subsidy deadline may expire after a single application window.

Assign each field a review horizon. For example, an upcoming event date might be reviewed every 30 days, while a historical edition can be checked once and then frozen. The queue should sort by the next review date, not by the page’s publication date.

3. Cross-source conflict

When an organizer page, a venue calendar and a national pavilion notice disagree, do not silently choose the most convenient value. Open a conflict record with:

  • the field in dispute;
  • each candidate value;
  • the source URL;
  • the retrieval timestamp;
  • the scope of the source (global show, national pavilion, local venue, reseller page).

A conflict is not an error to hide. It is a signal that the record needs a human decision and a note explaining the chosen source.

4. User-visible impact

A stale record on a low-traffic archive page is different from a stale record on an event that starts next month. Weight the queue by impact:

  • days until the event;
  • whether the page is linked from navigation or a country/industry hub;
  • whether the record is used in a subsidy or budget workflow;
  • whether the page has meaningful impressions or internal link equity.

This keeps the queue focused on records where a wrong answer changes a business decision.

A simple score is enough

A weighted score can be deliberately transparent:

refresh_priority = 4 * primary_source_changed + 3 * date_or_venue_conflict + 2 * upcoming_event + 2 * user_visible_page + 1 * review_horizon_expired 

The exact weights matter less than consistency. Store the component values and the scoring version with each task so a reviewer can understand why a record rose to the top. If the score changes because of a code update, do not rewrite history; create a new evaluation.

A minimal record shape

A refresh queue needs more than a URL. A useful record contains:

{ "record_id": "event-2027-example", "field": "start_date", "current_value": "2027-04-12", "candidate_value": "2027-04-14", "state": "conflicting", "primary_source": "organizer event page", "source_retrieved_at": "2026-11-22T04:30:00+08:00", "review_due_at": "2026-12-22", "impact_score": 9, "evidence_note": "organizer page and venue calendar disagree" } 

The important part is the audit trail: current value, candidate value, source, timestamp and reviewer decision. A single “verified” flag loses too much information, especially when the same page is revisited months later.

Make the queue operationally boring

A refresh queue should produce a small number of clear actions:

  1. Confirm — the current value is supported; update the verification timestamp.
  2. Correct — replace the value and record the source that supports the change.
  3. Merge — two records describe the same event; preserve one canonical URL.
  4. Retire — the event is cancelled or the page is obsolete; define the redirect or status.
  5. Defer — the source is unavailable; record the reason and a new review date.

Every action should be repeatable and should leave the public page in a valid state. If the source cannot be retrieved, do not invent a date. Mark the record unverified and explain what is missing.

Apply it to a public directory

Global Expo’s international exhibition directory is a useful example of why source-linked records matter: users compare dates, venues and editions before committing to travel or booth costs. The 2027 exhibition calendar adds a planning view, while the overseas-exhibition subsidy dataset shows why a date or eligibility change can affect a budget decision.

A public site should expose enough context for a reader to evaluate a record: the source, the retrieval date and the scope of the claim. That is better for users and easier to audit than a decorative “updated daily” label.

Metrics that reveal whether the queue works

Track four numbers, not just page count:

  • Median age of confirmed records — how fresh is the verified core?
  • Open conflicts per 1,000 records — are sources diverging faster than they are reconciled?
  • Time to resolve a primary-source change — how long does a known change remain unpublished?
  • Share of upcoming events confirmed within the review horizon — is the highest-impact segment actually being checked?

A queue that is always growing is not a success. A healthy system closes tasks, records why a value was accepted, and keeps the highest-impact records inside their review window.

A practical rollout

Start with one country, one industry and one field. Choose the field that most affects users, such as start date. Add source hashes and retrieval timestamps. Run the queue manually for two weeks and compare its decisions with the current editorial process. Then automate the scoring, not the judgment.

The goal is not to prove that every record is permanently correct. It is to make uncertainty visible, prioritize the records that matter and leave a reliable evidence trail when a fact changes.

Global Expo maintains source-linked international exhibition and subsidy resources. Verify operational details with the organizer or responsible authority before making a booking or application decision.