Skip to content
article

our startup tried a 4 day week for 3 months. here are the numbers, not the vibes

deploys, incidents, PR cycle time, sick days, and the one thing that got worse

Mio Kobayashi@mio_rustacean

August 30, 2026 · 3 min read

#10 in Career Ladder Nonsense

Contents(5)

we're a 9 person logistics startup in sapporo (6 engineers). in may our ceo agreed to a 3 month trial of a 4 day week: same pay, fridays off, 32 hours. everyone on mastodon has opinions about this, so i challenged to collect actual numbers. here they are. small team, one trial, not a study. but numbers.

the setup

  • mon-thu, 9:00 to 18:00 with an hour lunch. fridays off for everyone, not rotating
  • one person on call on fridays, rotating, paid extra. on call means phone on, not working
  • we cut recurring meetings first: standup became async in a slack thread, planning went from 2 hours to 45 min
  • we compared june-august (trial) to february-april (before). may was transition, ignored

the numbers

metric                        before (feb-apr)   trial (jun-aug)
deploys to prod / week        11.2               10.6
median PR cycle time          26 h               19 h
incidents (sev1 + sev2)       5                  3
friday on-call pages          n/a                4 total
sick days (whole eng team)    14                 5
recurring meeting h/wk/person 6.5                2.75
"I feel rested" (1-5, avg)    2.6                4.1

deploys went down a little (about 5%). everything else went the right way. the PR cycle time surprised me most. my guess: with one less day, people stopped letting PRs sit "until tomorrow", because tomorrow might be friday.

the thing that got worse

customer support. our ops team (3 people) also got fridays off, and our customers (warehouses) work fridays. we had to bring in a rotating friday support person with a stricter escalation path. two customers complained in june. none in august. if your customers work 5 days, you need a real plan for that day, not "eh, someone will check slack".

what we didn't measure well

  • "rested" is self reported and everyone knew the trial could be cancelled. people had a reason to say good things. i know
  • three months includes summer. sapporo summer is the best time of the year and everyone is happier anyway
  • we're small. coordination costs at 50 people are different

result

our ceo made it permanent from september. i'm not saying every company should do this. i'm saying the "productivity will collapse" prediction didn't happen for us, and i'd like people who say it will to bring numbers too 🦀

(the rust part: our CI got faster in the same period because i finally split the workspace into crates and turned on sccache. totally unrelated to the 4 day week. but i'm putting it here because i'm proud)

#productivity#burnout#career#rust

Written by

Mio Kobayashi

rust backend @ a small logistics startup in sapporo. ex-ruby. they/them. snow, ramen, borrow checker 🦀❄️

View profile

Get the best of #career-ladder-nonsense weekly

One email each Monday with the most useful posts. No spam; unsubscribe in one click.

Comments

13

Sign in to join the conversation.

Sign in to comment
Emeka Nwosu@emeka_golang1 rep in Career Ladder Nonsense
numbers! respect. but 6 engineers in a startup isn't a company, it's a group chat 😂 try this with 200 people and payment SLAs then we talk. also you snuck rust in at the end like a spy
2 net votes
Mio Kobayashi@mio_rustacean#10 in Career Ladder Nonsense
fair on the size, i said that. but "try it at 200" is what people said about remote too. and the rust was not sneaked. it was announced 🦀
2 net votes
Ankit Khurana@faang_or_bust14 rep in Career Ladder Nonsense
4 day week is how you get outshipped by a team in bangalore working 6. lol
0 net votes
Bram Verhoeven@bram_on_a_bike2 rep in Career Ladder Nonsense
No. Tired engineers write incidents. The table shows it. Read it again.
2 net votes
Ankit Khurana@faang_or_bust14 rep in Career Ladder Nonsense
small sample, meaningless
0 net votes
Mio Kobayashi@mio_rustacean#10 in Career Ladder Nonsense
yes, small sample, i said that in the post. your sample is zero though
2 net votes
Elin Sjöberg@elin_pm1 rep in Career Ladder Nonsense
Swedish view: a lot of us already work something close to this in July, and nobody pretends the company dies. The customer support point is the real implementation detail people skip. Happy to be proven wrong, but I'd bet the PR effect holds at bigger sizes.
3 net votes
Danny Cho@danny_cho_la26 rep in Career Ladder Nonsense
Credit for posting the thing that got worse. That section is why I believe the rest. One question: did deploy size change? Fewer, bigger deploys would explain the PR cycle time and look fine in your table while hiding risk.
6 net votes
Mio Kobayashi@mio_rustacean#10 in Career Ladder Nonsense
good question. checked: median diff size per deploy went from 140 to 155 lines. slightly bigger, not dramatically. adding it to the table
0 net votes
Nate Calloway@nostep_nate1 rep in Career Ladder Nonsense
1. self-reported "rested" with the trial on the line is worthless, you admit it. 2. summer. 3. n=6. so what we actually have is: deploys went down. the one hard number went the wrong way. well, actually, that's your result
7 net votes
Mio Kobayashi@mio_rustacean#10 in Career Ladder Nonsense
1 and 2 and 3 are in my post already. incidents and PR cycle time are hard numbers too and they went the right way. but i like that you read the whole thing
3 net votes
Jordan Pike@jpike2 rep in Career Ladder Nonsense
nate discovering that "deploys -5%" isn't the only row in a seven-row table is very on brand
2 net votes
Kenji Tanabe@kenjitanabe#6 in Career Ladder Nonsense
The sick days number is the one I'd show a CFO. 14 to 5 on a six person team is real money. Thanks for doing this properly.
2 net votes