Inside Orbislo
Remote, written down, and everyone takes a shift.
This page is the honest description of the working week, including the parts that are inconvenient. If the on call rotation is a dealbreaker for you, it is better that you learn it here than in week three.
Remote, with a real overlap
We hire between UTC minus 5 and UTC plus 9, and every team agrees a four hour overlap window. That band is wide enough to reach most of the Americas, all of Europe and Africa, and as far east as Japan. It is narrow enough that a question does not wait a full day for an answer.
Outside the window, your hours are yours. The only people expected to respond at odd hours are the ones on call, and they are paid for it.
Remote is not a perk here, it is the operating condition. It is also genuinely worse at some things. Onboarding is harder. Casual mentoring almost never happens by accident. New people take longer to work out who to ask.
We compensate with a written handbook that a new starter can read end to end in a day, a named onboarding buddy for the first six weeks, and two in person weeks a year that the company pays for in full.
Written first, and what that costs
A decision that exists only in a meeting informs the people who were in the room. A decision in a document informs the person who joins in March and the person who has to reverse it in October.
So proposals are written, argued in comments, and closed with a short note saying what was decided and what was rejected.
The cost is speed on the first day. Writing the argument down takes an hour that a conversation would have taken ten minutes. We pay it because the alternative is arguing the same question every quarter with slightly different people.
Two recurring meetings a week survive: a 30 minute team sync and a 45 minute Thursday demo where anybody shows anything. A meeting without a written agenda can be declined by anyone, without explanation.
The rotations, in full
| Rotation | Who | How often | Paid how |
|---|---|---|---|
| Support shift | Everyone, including engineers and the leadership team | One week in eight | Normal salary, no shift loading, it is part of the job |
| Primary on call | Engineers on provisioning, payments and the API | One week in six | Flat weekly rate plus time and a half for each page acted on |
| Secondary on call | A named backup, usually a senior engineer or a lead | One week in six | Half the flat weekly rate |
| Incident commander | Rotating across leads and senior engineers | One week in five | Flat weekly rate |
| Post mortem author | Whoever held the pager, never a manager writing about somebody else | Per incident over 30 minutes | Inside working hours, always |
Nobody is on two rotations in the same week. The calendar is public inside the company and the swaps are self service.
On call, for a product people depend on abroad
Most software can wait until morning. Ours cannot. A traveller whose profile fails to provision late at night has no data with which to contact us, which is the exact failure our free rescue data exists to break. The pager is the human half of the same idea.
We publish a post mortem for any incident over 30 minutes, inside 72 hours, on the status page. Writing that in public is the discipline that keeps the rotation honest, because a team that has to explain itself to customers fixes causes rather than symptoms.
- Primary on call is one week in six, never two weeks in a row, never over a public holiday you observe
- A page acted on overnight buys the following day off automatically, with no request and no approval
- More than three overnight pages in one week ends the shift early and the secondary takes over
- Every page has a runbook link. A page with no runbook is a defect and gets a ticket before the incident is closed
- The post mortem is written by the person who held the pager, and it names systems rather than people
- Nothing is deployed on a Friday afternoon unless it is fixing something already broken
The support rotation
One week in eight, you answer customers. Not a shadow session, not a listening exercise. Real tickets, with a first response target measured in seconds rather than hours, at every hour of the day across the rotation.
Frontline agents, including whoever is on rotation that week, can approve a refund up to 200 dollars without escalation. A person arguing with us about a small refund has already cost us more than the refund.
The rotation is the reason our roadmap is boring in the right way. When the person who wrote the provisioning retry loop spends a week reading tickets about the provisioning retry loop, the retry loop improves.
In the last cycle, four shipped changes came straight out of a support week. A clearer error when a device is locked. An earlier device check on the store pages. A plain language explanation of the daily fair use threshold in the app. An automatic credit when a plan price falls inside the price protection window.
1 in 8
weeks on support, everybody
1 in 6
weeks carrying the pager, engineers
2
recurring meetings a week
72 h
to publish a post mortem over 30 minutes