Historical Twitter data, or how to measure what already ended
Historical Twitter data goes back to 2006, so a campaign from years ago can be measured today. How to bound the period and what deleted posts do to your method.

The request usually arrives late, and historical Twitter data is what saves it. Someone needs numbers for a campaign that ran eighteen months ago and nobody was collecting.
It is more recoverable than people assume.
How far back the archive goes
X holds public posts going back to 2006. A query against the full archive can cover any period since then, and it is not restricted to recent weeks the way live endpoints are.
That means a campaign, an event or a crisis from years ago can be measured today with the same figures it would have produced then.
Tell us the period you need and we will run it against the full archive, within the Twitter compliance rules.
Why measuring after the fact still works
The posts and their engagement are attached to the posts themselves. They do not expire because nobody looked at them. What you get back is the same dataset you would have collected live, every matching post with its author, its timestamp and its figures.
Impressions are the exception. X only began returning them per post in 2023, so a report on anything earlier carries every other metric and not that one. It is worth saying out loud before somebody asks where the reach column went.
The one thing that changes is anything deleted since. A post removed by its author is gone, and a sound methodology says so rather than pretending the archive is complete.
What a historical query is good for
Three jobs come up constantly. Benchmarking, because this year's campaign means nothing without last year's numbers. Retrospective crisis analysis, where the useful question is how fast it spread and who carried it. And academic work, where a dataset has to cover a period that ended before the research started.
The third is the one where the archive matters most. A thesis on an election cannot wait for the next election.
How to bound the period properly
The period does not live inside the query. It travels beside it, as the start and end dates of the search, so the same query can be run over any stretch of the archive without a word of it changing.
For anything longer than a few months, run it in slices rather than in one go. Month by month gives you the comparison for free, and a mistake in one slice stays in that slice instead of poisoning the whole set.
What limits the size, and what does not
The period is not the constraint. The number of posts is. A quiet keyword over five years can be fewer posts than a hashtag over one afternoon. What you are buying is volume analysed, so a long historical query costs exactly what the same number of posts would cost over a single day.
That is worth knowing before scoping a project, because people routinely assume the opposite.
Doing it within the rules
Historical access to public posts is available through the official API on the terms it sets. Everything we deliver is gathered Twitter compliant, which is what keeps a historical dataset usable for client work and for publication.
Choose a plan and the report arrives with the dataset behind it, so the period and the query can both be verified.