When software ships
When does software actually get released, and does the rule against shipping on Friday show up in the data?
11.9%
of releases land on a Saturday or Sunday. Two days in seven is 28.6%.
20.0%
on Wednesday, the busiest day of the week, from 33,100 releases.
14.4%
on Friday, despite its reputation as a day to avoid. An even week would give it 14.3%.
A release may be a technical event, but its timing is distinctly human. Across 33,100 releases published in the last three years, the dates follow office hours, the working week and the holiday season closely enough to reveal how teams plan, coordinate and pause.
Teams protect the weekend more than Friday
“Never ship on a Friday” may be software’s most repeated release convention. In this catalog, the more visible habit is keeping releases inside the working week. Friday carries 14.4% of all releases, against the 14.3% it would carry if nobody thought about it at all. Monday sits in almost the same place, at 14.9%. Both are ordinary days. The week forms a clear rhythm: activity builds toward a midweek peak, eases at both ends of the working week, and leaves the weekend largely protected.
The same figures as a table
| Day | Share | Releases |
|---|---|---|
| Monday | 14.9% | 4,937 |
| Tuesday | 19.4% | 6,424 |
| Wednesday | 20.0% | 6,626 |
| Thursday | 19.4% | 6,415 |
| Friday | 14.4% | 4,766 |
| Saturday | 6.2% | 2,045 |
| Sunday | 5.7% | 1,887 |
Read against each other, Wednesday and Friday are 1.4× apart. Read against an even week, the standout is the midweek peak—not a Friday drop. The practical pattern is a preference for the middle of the week, with more working-week runway on either side.
Releases happen during working hours
The clock says the same thing the calendar does. Publication peaks at 17:00 UTC , which is 13:00 in New York and 19:00 in Berlin, and bottoms out at 05:00 UTC — 01:00 and 07:00. The busiest hour runs at 3.6× the volume of the quietest. The climb between them traces a working day across the two regions most of this catalog is published from.
The same figures as a table
| Hour (UTC) | Share | Releases |
|---|---|---|
| 00:00 | 2.9% | 875 |
| 01:00 | 2.6% | 786 |
| 02:00 | 2.4% | 704 |
| 03:00 | 2.3% | 700 |
| 04:00 | 2.0% | 586 |
| 05:00 | 1.9% | 571 |
| 06:00 | 2.4% | 730 |
| 07:00 | 3.3% | 978 |
| 08:00 | 3.8% | 1,126 |
| 09:00 | 4.5% | 1,351 |
| 10:00 | 4.2% | 1,271 |
| 11:00 | 4.1% | 1,236 |
| 12:00 | 5.0% | 1,496 |
| 13:00 | 5.6% | 1,665 |
| 14:00 | 6.2% | 1,855 |
| 15:00 | 6.3% | 1,884 |
| 16:00 | 6.6% | 1,965 |
| 17:00 | 6.8% | 2,033 |
| 18:00 | 6.0% | 1,805 |
| 19:00 | 5.2% | 1,563 |
| 20:00 | 4.5% | 1,345 |
| 21:00 | 4.2% | 1,268 |
| 22:00 | 3.8% | 1,127 |
| 23:00 | 3.5% | 1,035 |
This is a shape worth reading carefully, because it is also the measurement’s weakness. Everything here is bucketed in UTC, and the corpus is mostly published from North America and Europe. The working day that shows up in the middle of this chart is real; the exact hour is a property of where the publishers are, not of software.
Weekend releases reveal different team rhythms
Weekend activity is not evenly distributed. Between Firmware at the top and Infrastructure & DevOps at the bottom there is a spread of about 5×. Games produces more releases here than any other category and comes low on this measure, at 5.8%. The consistent pattern underneath the ordering is organizational: products with formal release coverage cluster toward the bottom, while independently maintained projects appear more often near the top.
| Firmware | 30.1%46/153 | |
|---|---|---|
| Media | 28.0%207/739 | |
| Gaming Tools | 27.0%195/723 | |
| System Utilities | 22.7%253/1,115 | |
| Productivity | 17.1%342/1,995 | |
| Operating Systems | 16.7%105/628 | |
| Creative Tools | 16.3%120/738 | |
| Frameworks & Libraries | 14.8%584/3,941 | |
| AI | 13.3%563/4,233 | |
| Lifestyle & Apps | 12.1%57/471 | |
| Developer Tools | 11.3%542/4,791 | |
| Security & Privacy | 10.8%126/1,168 | |
| Drivers | 10.1%12/119 | |
| Databases & Data | 8.9%115/1,298 | |
| Browsers | 8.1%56/688 | |
| Communication | 7.1%85/1,191 | |
| Languages & Runtimes | 6.2%26/418 | |
| Games | 5.8%352/6,088 | |
| Infrastructure & DevOps | 5.6%146/2,603 |
Weekday consistency across 3 years
Products with at least 20 releases in the window and not a single one on a Saturday or Sunday.
- GitHub219 releases
- Deepgram156 releases
- Apple Platforms116 releases
- n8n115 releases
- Google Workspace107 releases
- Firefox77 releases
- League of Legends73 releases
- VALORANT69 releases
Weekend-friendly schedules
Products whose releases land at the weekend most often in this catalog.
- exelban-stats90% · 18/20
- ARK: Survival Ascended76% · 32/42
- Inquirer.js55% · 16/29
- OpenTTD52% · 12/23
- Motrix50% · 15/30
- Commander.js50% · 12/24
- tsx42% · 10/24
- Symfony42% · 10/24
Together, the lists show two release rhythms consistent with different project structures: formal weekday coverage on one side and more flexible maintainer schedules on the other. Weekend activity in a release history is therefore a signal about how a project may be organized, not simply how its engineering works.
The holiday slowdown is visible in the data
The same human rhythm appears at a larger scale. In every complete December this window covers, the fortnight from the 18th ran between 20% and 46% below the four weeks before it, measured as releases per day so the two spans are comparable.
| Season | Nov 24 – Dec 17 | Dec 18 – Dec 31 | Change |
|---|---|---|---|
| 2023 | 4.5/day | 3.6/day | −20% |
| 2024 | 9.7/day | 5.9/day | −39% |
| 2025 | 26.3/day | 14.2/day | −46% |
The drop grows with the size of the catalog, which is a caution rather than a trend: the earliest season here is measured over a few hundred releases and the latest over several thousand, so the recent figures are the more reliable ones.
How this is measured, and how to read it
- This is publication, not work. Every figure comes from the date a vendor published a release, as their own feed reported it. It is not a commit time, a merge, or a deploy. The familiar arguments about shipping on a Friday are about deploys, which is a different question: a release published on Tuesday was very often built on the Friday before.
- Every bucket is UTC, and that moves releases between days. A release published on Friday evening in California is recorded on Saturday. 29.3% of Saturday’s timestamped releases land before 07:00 UTC — the previous afternoon and evening across the Americas — against 14.4% Monday to Thursday. So the real weekend floor is lower than the one on this page and Friday’s share is higher. The bias runs with the argument rather than against it, which is the reason to put a number on it instead of a disclaimer.
- About one release in ten carries no time of day. 29,955 of 33,100 releases have one. The rest were published as a bare date — some vendors, and some of the pages we read, never carried a clock. Those releases count towards the day of the week and are left out of the hourly chart entirely, rather than being piled onto midnight.
- Category is partly a proxy for channel. Releases tracked from GitHub sit at 16.1% at the weekend and releases tracked from Steam at 6.1%, so a category made mostly of one or the other inherits some of that gap. The category table is a real difference between kinds of software and a real difference between where they publish, and this measurement cannot separate the two.
- The window is the last 3 years, and it slides. Everything here is measured from 2023-10-10 forward. The catalog has grown a great deal over that span, so recent releases dominate every total — though the weekly shape has held steady year by year since well before the catalog reached its current size.
- The catalog is not the world. These are the products What’s New tracks, which skews towards software that publishes a changelog at all, and towards the parts of the industry that publish in English on the open web.
Recomputed every week. Every figure comes from the catalog at build time rather than being typed in.
The companion to this page is the release note clarity gap, which asks what those releases actually say when they arrive. Declarative Changelogs turns the patterns in both reports into a practical proposal for clearer, more useful release communication.