Define the event
Specify actors, location, interval, type of change and decision. Avoid collecting everything matching a keyword.
GUIDE / PROTOCOL
Reconstruct a timeline of public signals
Order of appearance on a platform does not establish the order of events.
Direct answer
How do you build a verifiable timeline from open sources?
Define the change under study, then record claimed dates, observed dates and capture dates separately. Link every event to its source and uncertainty; look for earlier versions before inferring causation.
01 / Review method
Review method
Collect milestones
Record earliest found version, updates, announcements, repetitions and corrections. Note timezone and precision of each date.
Separate clocks
Distinguish alleged event, publication, indexing, viewing and archiving. A late capture does not date the event.
Test relationships
Compare sequences with other sources and competing hypotheses. Publish an annotated timeline, not an unsupported causal story.
02 / Evidence to keep
Evidence to keep
Milestone table
Event, date and time, timezone, date type, URL and verification status.
Retained versions
Copy or archive reference, capture timestamp and material differences.
Reuse chain
Who cites whom, repetition date and any correction to the first source.
Critical reading
Established events, uncertain intervals, alternative explanations and review point.
PRACTICAL CHECK
Before reaching a conclusion
A frequent misreading and the check that helps avoid it.
Premature conclusion
The date of a capture or share is not necessarily the date of the depicted event. Conflating them can reverse causality.
Decisive check
Record alleged event, earliest located publication, edits, reposts and access time separately, with time zone and confidence level.
AFTER THE REVIEW
Decide from the evidence
An archive appears after the alleged event
The archive at most establishes the page at capture time. Look for an earlier trace and leave the event undated if none exists.
A correction removes the original wording
Keep both versions, date each observation and explain which finding the correction changes.
Decision and review trail
Decision and review trail
What the check must help decide and what to retain so the choice can be reviewed.
Sufficient evidence
Record event, publication and access dates separately for every signal; note corrections and captures. Page dates alone are not a chronology of events.
Next decision
Compare plausible sequences and mark unknown intervals. Revise the timeline when a primary record moves a significant milestone.
OSINT / Official references
When dating a public signal
Event, filing, publication and capture dates mark different moments; record them separately.
BODACC
Why start hereDate publication of a French commercial notice.
Limit to rememberA notice records a published event; it does not describe the entity’s entire current activity.
Read source entryTED — Tenders Electronic Daily
Why start hereFollow published stages of a European procurement procedure.
Limit to rememberA notice does not prove contract delivery or outcome; read its stage and linked documents.
Read source entryWayback Machine
Why start hereLocate wording in an archived web-page version.
Limit to rememberA capture can be incomplete or combine resources from different dates. Archiving does not validate page claims; absence from replay does not automatically mean an element did not exist at the time.
Read source entryDetailed inspection cases
Typical situations for preparing a check. They do not describe completed assignments or actual observations.
An update date is mistaken for the event date
What to inspect
Identify the described event, first publication, revision and consultation time separately. Check time zones when closely timed sources appear contradictory.
Decision and action
Create separate fields and order the timeline according to the question studied. A late revision does not automatically move the event it describes.
Record to retain
Keep each date with its type, precision, time zone where known and supporting document.
A retrospective correction erases an initial claim
What to inspect
Compare available versions and read correction notices. Distinguish what was published on a given date from what the current version states.
Decision and action
Add a correction event and limit conclusions to the relevant version. Absence of the old wording on the current page does not prove it never existed.
Record to retain
Keep dated versions, observed changes and periods during which each wording was documented.
Publication is mistaken for the event
What to inspect
Separate event, publication and collection dates.
Decision and action
Show temporal uncertainty instead of manufacturing a definite order.
Record to retain
Three dates, known time zone and ordering rationale.
Review before deciding
These questions complement the method. They help establish what is actually supported while retaining qualifications in the record.
Distinct dates
Does each row separate event, publication, capture and observation, with known precision and time zone? A page update must not become a new event date.
View a worked decision
Fictional example: the state applies only to the criterion and scope described.
- Fictional observation
- The page was updated in 2024; the timeline dates an explicitly 2021 event to 2024.
- Declared state
- Observed discrepancy
- Decision and next check
- Restore event date and retain publication, update and observation separately.
Justified interval
Is the first recovered trace presented as an observation rather than a certain origin? Retain gaps and check the date of each destination when navigating an archive.
View a worked decision
Fictional example: the state applies only to the criterion and scope described.
- Fictional observation
- The earliest found capture is from 2018; no documented earlier search is available.
- Declared state
- Not checked
- Decision and next check
- Write “trace found in 2018”, without inferring creation in 2018. Document search limits.
Partial capture
Are missing media and links leaving the archive identified? A captured page does not establish that all its resources were preserved.
Documentation consulted : Internet Archive · Wayback Machine General Information ·
View a worked decision
Fictional example: the state applies only to the criterion and scope described.
- Fictional observation
- The record separates the archived page, one missing image and a link opening the live web.
- Declared state
- Verified within this scope
- Decision and next check
- The partial capture is explicit. Do not date current destination content to the archive date.
TOOLS / REGISTERS
Information to record
Distinguish event, publication and observation dates. A later correction creates a new row version; it should not erase what could be observed previously.
- Access date and time zone
- Exact URL
- Claim to verify
- Event, publication and observation dates
- Version / revision
- Limit / uncertainty
- Decision / next action
- Review owner
- Next review
03 / FAQ
Frequently asked questions
Does an archive prove creation date?
No. It shows a capture at one time, not necessarily the first publication.
What about approximate dates?
Keep the interval or stated precision instead of inventing an exact time.
Keep every repost?
Only those needed to understand spread or a changed message.
When revise the timeline?
When an earlier source or documented correction changes the sequence or interpretation.
04 / Official references
Official references
Describe a context
DOCUMENTED REFERENCE
Time windows: state the bounds to prevent double counting
The [start, end) convention includes the start and excludes the end. Two windows can meet without assigning their shared boundary instant to both. An extraction must state this choice; it does not automatically make an API’s filters exclusive.
| Reference | Value or form | What it identifies |
|---|---|---|
| [a,b) | a ≤ t < b | Instant b belongs to [b,c), but not [a,b). A corresponding selection condition is t ≥ start AND t < end. |
| [a,b) ∩ [b,c) | ∅ | Two non-empty windows meeting at a boundary have no shared duration. Windows closed at both ends would share instant b. |
| intersection | max(0, min(end) − max(start)) | On the same scale, shared duration is max(0, min(ends) − max(starts)). A positive duration indicates overlap. |
| [a,a) | empty | Under this convention [a,a) is empty. It is not a point event; the tool flags it separately and does not report a gap between two non-empty windows. |
| PostgreSQL 18 | daterange / tsrange / tstzrange | PostgreSQL distinguishes daterange, tsrange and tstzrange. The canonical [) form for discrete dates does not mean all timestamp ranges are automatically rewritten that way. |
Compare two [start, end) windows
Format YYYY-MM-DDTHH:MM:SS[.fraction]Z or ±HH:MM; t and z accepted. No added spaces or inferred zone. Years and UTC result 0001–9999, 0–9 fractional digits, at most 256 characters per field. 24:00 is not accepted in this profile.
Local calculation without transmitting inputs. Durations use days of 86,400 seconds without adding leap seconds; they do not measure clock drift.
Example calculated; initial data remain available without JavaScript.
- Duration A · s
- 86400
- Duration B · s
- 86400
- Shared duration · s
- 0
- Union duration · s
- 172800
- Gap between non-empty windows · s
- 0
- Window relationship
- Adjacent windows
- Instant in A
- No
- Instant in B
- Yes
Initial example data
{
"aStart": "2025-01-01T00:00:00Z",
"aEnd": "2025-01-02T00:00:00Z",
"bStart": "2025-01-02T00:00:00Z",
"bEnd": "2025-01-03T00:00:00Z",
"probe": "2025-01-02T00:00:00Z"
}Implication for the record
This reader imposes [start, end), finite bounds and instants with explicit offsets. It compares a chronological model without leap seconds, not a remote service’s conventions. Check the API’s precision and inclusivity before moving a bound; arbitrarily subtracting a millisecond can lose finer observations.
Primary documentation
Documentation consulted on .
Related checks
- Read an archive at record level
- Europe/Paris: two instants, the same displayed time
- BODACC: choose the edition before interpreting results
- Memento: separate the requested date from the state returned
- RFC 3161 timestamps: identify what the token actually dates
- Timestamps: check the calendar before comparing strings
- Partial dates: preserve the source’s precision
DEEPER READING / PRACTICE
Distinguish an event from its publication
Documents appear on several dates and sometimes describe the same event. The chronology must distinguish the time of the fact, the time of its publication and the time of his consultation.
Updated on .
Priority checks
Release the different dates with their role and accuracy. Keep the fuses when they are useful and report the deducted or uncertain dates.
Interpretation pitfalls
Classing only by date of publication can reverse the order of events. A updated version can recover an older fact without changing its timing.
Working record to retain
The chronology associates event, date of the fact, date of publication, source and reserve. Uncertain intervals remain visible rather than replaced by an arbitrary date.
PRACTICAL CASE / WORKED EXAMPLE
A late publication describes an old event
The example below is pedagogical. It shows how to move from a document to a qualified proposal, preserving the context, the contradictory elements and the information that the research does not yet allow to establish.
Example context
A recent article refers to a decision made several weeks earlier. The ranking by publication places the event after the consequences that already arise from it.
Suggested approach
Identify the date of the fact, publication and consultation. Examine passages that specify time and keep the dates unknown. Reclass the events according to their role rather than their sole public appearance.
Conclusion and next step
The chronology shows the gap between the event and public knowledge. The conclusions on the order of the facts remain related to the established dates and uncertain intervals.
Questions for deeper practice
How do we define an approximate date?
Keep the accuracy available: month, period or interval. Do not choose an arbitrary day only to facilitate the ranking; the visualization and the comment must show that limit.
DETAILED WORKSHOP / APPLICATION
Preserving the different timelines of an event
01 / Prepare the necessary information
Gathering facts, publications and consultations. Identify passages that give a exact date, period or relative order between events.
02 / Carry out the approach
Attribute a role to each date, and then build the chronology. Check the useful fuses and versions. Keep the intervals uncertain and distinguish public knowledge and occurred from the fact.
03 / Examine an alternative situation
If the date is deducted, report the deduction elements. The classification should not give it a greater accuracy than that of the sources.
04 / Decide on the next step
The chronology is exploitable when the dates and their accuracy are understandable. The conclusions on the order of the facts comply with the intervals that prevail.
05 / Hand over the verification records
Transmit events, date roles, sources and reserves. A new piece can specify time without deleting the limits of previous observations.
ANNOTATED WORKING DOCUMENT / REVIEW GRID
A table of events with role and date accuracy
The table distinguishes date of the fact, publication and consultation. It retains the approximate periods and sources that make it possible to understand the order of events and their public visibility.
On a small screen, scroll the grid horizontally to read all three columns.
| Observation | Supported interpretation | Next step |
|---|---|---|
| The date of the incident is explicitly documented. | The event can be located with the accuracy provided. | Keep the passage and fuse when they influence reading. |
| Only the date of publication is known. | The event may be previous and remains to be clarified. | Do not replace the date role and look for a supplementary document. |
| Uncertain intervals are overflowing. | The exact order cannot be established from these pieces. | To present uncertainty rather than impose an arbitrary sequence. |
Reviewing the working document
The final verification verifies that a normalized date does not receive artificial accuracy. The conclusions on succession and causality must respect the distinction between the event and public knowledge of that event.
REPORTING / EXAMPLE WORDING
Presentation of results: Distinguish event and publication
Point to clarify
The order of publication may differ from the order of events.
Before handing over
Check the role and accuracy of each date before a time conclusion.
Example of a qualified conclusion
The chronology distinguishes occurring from the fact, publication and consultation. Approximative periods are preserved and unestablished successions remain explicitly uncertain.