A feed announces an old publication
Check first-publication date and content. Separate renewed promotion from documentary novelty. If a real revision exists, describe the difference rather than repeat the title as a new fact.
In-depth practical guides
Choose sources, recognise real novelty and retain history: build monitoring that produces useful information rather than a pile of notifications.
Updated :

AI-generated illustration
Monitoring starts with a decision need, not a feed count. Define topics, actors, territories and expected documents to distinguish a relevant signal from republication. Preserve provenance and dates for each observation. An absence of updates is meaningful only when consulted sources and unavailable sources are known.
Write the questions, exclusions and intended recipient. Define what triggers a note: new document, material change, correction or disappearance. Use a short output containing the change, supporting evidence and relevance to the decision.
List accessible official pages, publications, registers and feeds. Record issuer, coverage, language, access method and chosen checking frequency. Fill coverage gaps rather than multiplying interfaces that repeat the same publication.
Record event, publication, announced modification and observation in separate fields. Atom feeds distinguish identifiers and update dates; an update date should not automatically be interpreted as the date of the described event.
Compare document reference, stable identifier where available, final URL and useful content. A changed title alone does not create a new document. Link a revision to its predecessor and describe the actual change.
Open the complete document, read qualifications and verify scope. Attribute statements clearly. Media coverage can help discover a publication, but keep the document link that actually supports the conclusion.
Retain findings, access failures and classification decisions. Schedule checks according to actual change frequency and the cost of missing information. Reassess inactive sources and alerts that never yield useful signals.
Reading workshop
File DEMO-W08 separates coverage, observation and material change at an unchanged URL.
Fictional data and records for learning: none of these results comes from a live service query.
Document v1 was retained before the outage. On resumption the feed window starts at 06:00, and v2 changes the case count without changing the event date.
Open each proposition to compare its diagnosis. More checks may still be needed.
No: the outage leaves an uncovered interval and v2 contains a correction.
No: 3 May is the observation date; the event remains dated 30 April.
Appropriate here. The gap stays open until other evidence resolves it.
File: DEMO-W08 2 May: 503; uncovered period 3 May 08:00: resumed; window since 06:00 URL: unchanged between v1 and v2 Event: 30 April; value 50 → 47 Mirror: v1 repost; linked provenance Outcome: correction to follow; gap retained
Cite the version actually reviewed and separate event, publication and observation dates. Keep the uncovered period and the mirror’s provenance link in the log.
Check first-publication date and content. Separate renewed promotion from documentary novelty. If a real revision exists, describe the difference rather than repeat the title as a new fact.
Link notifications to the initial publication. Keep republications adding an identifiable contribution; group others so notification volume is not presented as independent confirmation.
Record last successful access and current failure. Look for a new official address or equivalent channel. Do not translate access failure into no change; mark coverage as incomplete.
Worked exercise
Fictional example: the values and documents teach the method and do not describe a real organisation.
Eight alerts concern these three records. Repeated notifications do not change their content.
If the table exceeds the screen, scroll horizontally; arrow keys work when the region is focused.
| Record | Alerts received | Record status |
|---|---|---|
| A | 4 | Already known |
| B | 2 | Correction identified |
| C | 2 | New record |
There are three distinct documents. File A, check affected conclusions for B and read C. The useful outcome is one new item and one correction, not eight confirmations. Deduplicating notifications does not replace comparing versions.
If B's alerts concern different versions, separate them before grouping.
Keep identifier, previous version, observed change, action and follow-up status.
Match frequency to the source, need and acceptable delay. Annual publications and frequently edited pages need different schedules. Reassess based on changes actually observed.
It identifies a lead. Open the document, confirm version and context and retain its reference. If inaccessible, mark the signal unverified instead of treating the notification as complete evidence.
No. They may add translation, context or independent verification. Link them to the original and describe their contribution; copies adding nothing can be grouped.
Track qualified signals, decisions supported, coverage gaps and processing time. Raw alert count measures activity, not by itself the documentary value produced.
Compare the information used in the folder: values, definitions, conditions and conclusions. A visual change may be minor, but it must be examined if it affects the readability or location of evidence.
DEEPER READING / PRACTICE
A documentary watch collects new links but does not notice changes to a already known document. Tracking must include versions, replacement and context changes.
Updated on .
Defining followed documents, identificators, frequency and change criteria. Compare versions and distinguish form modification, new data and background correction.
An unchanged URL can be used as a modified document. A new publication can also be a copy without additional information.
The watch journal retains the previous version, the observed version, the change and the consequence for the conclusions. The warnings correspond to a concrete effect on the case.