How to build platform timelines that survive review

A timeline looks confident and is one of the hardest things to get right. It puts events on a line, in an order, with a spacing.

Each of those three decisions smuggles in a claim. Most published platform timelines copy an earlier one, which is why they agree with each other and disagree with the record.

This page is about how to build one that holds up, and about the specific reasons this subject resists the format.

What to take away

Four date types used in platform timelines

The reported date is the one that travels, because it is the one attached to a citable document. That produces a systematic bias: timelines assembled from press coverage place everything later than it happened, and they place things that got early coverage earlier relative to things that did not.

Fixing this does not require better sources. It requires labeling. An entry that says which of the four it is remains useful even when it is the wrong one, because a reader can correct for it. An unlabeled date cannot be corrected by anybody.

Why sources agree without being independent

When five outlets carry the same date, the natural reading is confirmation. Usually it is repetition. One account is written, others summarize it, later writers cite the summaries, and eventually the original is out of print or offline while the copies remain. Nothing in the resulting set is independent, but it looks like consensus.

The way out is to trace rather than to count. Follow each source back until it either produces a contemporary document or produces another summary. Most chains end at one place, and a surprising number end nowhere, at a sentence with no source that everyone has been quoting.

Three practical tells that a chain is short:

The material itself is disappearing

Even careful work runs into the fact that the primary record for this subject is unusually fragile.

Pages move and vanish, and the citations pointing at them stop resolving. Link rot is the background condition here, not an occasional annoyance.

It hits hardest the material that matters most: announcements, help pages, terms of service and developer notes. All are edited in place, with no version history a reader can see.

Web archiving recovers some of it. Its coverage is uneven in ways worth understanding before you rely on it.

The Internet Archive’s Wayback Machine lets you look for saved versions of public pages. Search for the page URL, open an available capture, and record its capture timestamp separately from any date printed on the page. If no capture appears, that shows only that the service has no available capture for that URL—not that the page never existed.

Public pages are captured better than pages behind a login. Popular pages are captured more often than obscure ones. Content assembled by scripts in the reader's browser may be archived as a shell with no content. Each capture has its own date, not the date the page was published.

So a timeline built today from archived material is not a record of what happened. It is a record of what was captured, filtered by what was public, popular and simple enough to save.

Building an entry that survives review

Example: a YouTube timeline entry

This entry supports a claim about a video upload, not a claim about YouTube’s launch. If another account gives a different date, keep that account’s claim and source separate rather than averaging the dates.

What a timeline cannot show

Spacing implies pace, and a line implies causation. Both are usually wrong.

Events plotted at even intervals suggest steady development. Real change in this area is lumpy, and long quiet periods do most of the work. Two entries next to each other on a line invite the reader to connect them, and readers do, whether or not there is any relationship.

The other systematic distortion is survivorship. A timeline is built from things that lasted long enough to be recorded. Systems that were widely used and then closed leave a thinner trail than smaller ones that persisted, so the line quietly describes durability rather than importance.

Where those effects matter, a table beats a line. It carries the same information without the visual argument, and it has room for the date type and the source, which is the information a reader actually needs.

How this connects to the rest of the site

The definitional problems that make chronology hard appear in more detail under early social networks. There, what counts as a beginning decides the entire order of events.

That page and this one are two halves of the same difficulty. One asks what you are counting; this one asks when you say it happened.

Common questions

Why not just use the company's own history page?

Use it, and label it. A company's account of itself is a primary source about what the company says, which is genuinely useful, and it is edited to serve the present. It is the right source for a founding claim and the wrong source for how something was received.

Is an approximate date better than no date?

Yes, if it is marked approximate. "Some time in this period, from a retrospective account" is an honest entry. The same value written as a precise date is a fabrication with good manners.

How do I handle something that launched in stages?

Record the stages. A staged rollout has no single date, and forcing one produces an entry that is wrong for most readers. This is the most common single error in published platform timelines.

What should I do when a cited page has disappeared?

Note that it is gone, keep what you quoted, and look for a capture. If nothing survives, the entry drops from documented to reported, and it should be relabeled rather than quietly kept at its old confidence.