The best time to post on Mastodon (and why we did not measure it)

We ran a real study on Bluesky and found posting time barely mattered. Here is what that implies for Mastodon, and why we did not build the same dataset here.

This article is reasoning, not measurement, and it is worth saying that before anything else rather than in a footnote at the bottom.

We collected a real dataset for Bluesky and found that posting time barely mattered. We deliberately did not build the equivalent for Mastodon. What follows is what the Bluesky result implies here, which parts of the argument get stronger, and the one place on Mastodon where timing might genuinely matter.

What we actually measured, on Bluesky

We sampled Bluesky's public firehose and analysed 8,373 English posts from 344 accounts. Pooled across accounts, the data showed a 2.9x spread between the best and worst hour — the sort of number these articles usually print.

That number turned out to be an artefact of which accounts post at which hours. Scoring every post against its own account's median instead removed the confounder, and the effect very nearly vanished: of 31 hour and weekday buckets, exactly one differed from baseline, which is what testing 31 buckets at 95% confidence produces by chance. The full study is here, method included.

Why the argument is stronger on Mastodon

The mechanism behind that null result is the chronological feed. On a ranked network, timing compounds: the algorithm notices early engagement and shows the post to more people, so a good slot gets amplified into a large difference. Bluesky's default Following feed does not do that, and the effect duly failed to appear.

Mastodon's home timeline is chronological too — and unlike Bluesky, there is no algorithmic alternative sitting alongside it. Bluesky has custom feeds, which are a real discovery surface even if they are opt-in. Mastodon's home timeline has no ranking at all.

"If timing did not measurably matter on the network that has some amplification available, it is unlikely to matter more on the one that has less."

That is the whole inference, and it is worth being clear about its strength: it is an argument from mechanism, not a measurement. It could be wrong.

The one place timing might genuinely matter here

Mastodon is not entirely unranked. The Explore tab surfaces trending posts, hashtags and links, and trending is computed from boosts over a recent window.

That is a cascade mechanism, and cascades are exactly where timing effects live. A post that picks up boosts quickly can reach Explore, and reaching Explore produces more boosts. If posting when more of your followers are awake makes early boosts more likely, that could matter in a way it does not on the home timeline.

We have not measured this and we are not going to assert it. It is the honest place to look if someone does run the study — and it is a much narrower claim than the ones usually made.

What actually decides who sees a Mastodon post

Four things, none of which is the hour on the clock:

• Boosts. The only real amplification path. A post that gets boosted by someone with a different audience reaches people you could never reach directly, and that is essentially the entire growth mechanism.
• How busy your instance's local timeline is. On a large server, posts scroll past quickly. This is timing-adjacent, but it is about which server you are on rather than which hour you post.
• Whether you show up otherwise. Replies and boosts of other people's work put you in front of their followers. On a network with no algorithm, being present is the distribution.
• Content warnings and alt text. Not engagement tricks — conventions. A post ignoring them gets fewer boosts here in a way it would not elsewhere.

Why we did not build a Mastodon dataset

We could have. Public timeline endpoints exist and collecting from them is technically trivial. We decided not to, and the reasons are specific rather than squeamish.

Many instances prohibit it in their terms. Servers are independently governed, and a good number explicitly forbid collecting posts for research or analysis. A crawler that ignores instance policy is not gathering public data — it is breaking the rules of the community it is studying.

The ethical norm here is higher than "it was public". In fediverse research the expected standard is consent from instance administrators, even for publicly visible posts. That is a stricter bar than most social-media research applies, and it is the bar this community has set for itself.

There is precedent. A study analysing content warnings on Mastodon drew an open letter from the community and its dataset was withdrawn from the Harvard Dataverse. Researchers with more standing than a scheduling company have got this wrong and paid for it.

Set against that: we would have produced a blog post. Bluesky's firehose is a deliberate, documented public broadcast with different norms around it, which is why that study exists and this one does not.

What to do instead

• Do not optimise the hour. On the available evidence it is among the lowest-value things you could spend attention on.
• Post consistently. A post that exists at a mediocre time beats a perfect slot you never filled.
• Spread posts out. Not because certain hours are magic, but because a chronological timeline gives your followers one pass and they are not all reading at once.
• Reply and boost. On a network with no algorithm, that is the distribution mechanism. It is not a supplement to posting; it is most of the job.
• Write content warnings and alt text. They affect whether people boost you, which is the thing that actually determines reach here.

Checking it for your own account

Our finding is about 344 Bluesky accounts. Yours is one account with one audience, and the only data that can answer the question for you is your own.

Vary your posting times deliberately for a few weeks, record what each post got, and compare posts against your own typical post rather than against an industry benchmark. That last part is the entire methodological trick from the Bluesky study, and it is what stops you concluding that Tuesdays are magic because you happened to post something good on one.

If you would rather set the rhythm once and stop thinking about it, that is what a posting queue is for — and the honest pitch is consistency, not a secret about what time it is.