In an on-call schedule, what is an override, and why is agreeing in a chat message to cover for a teammate not the same thing?

## Two mechanisms, one principle An **override** replaces whoever is on call for a specific window — an afternoon, three hours, a single evening — without changing the underlying rotation. After the window ends, the schedule reverts to normal. A **swap** exchanges whole shifts between two people in the rotation, so the schedule itself changes for both. Both exist because of one principle: the paging system routes to whoever the schedule says is on call at the moment the alert fires. It has no knowledge of conversations, good intentions or team chat. ## What actually happens when the agreement is informal Suppose you are on call and you will be on a three-hour flight. A teammate says in chat that they will watch things. At minute forty of the flight, an alert fires: 1. The paging system looks at the schedule, sees your name, and calls your phone. You are in the air. 2. The acknowledgement timeout expires and escalation proceeds down **your** chain — your secondary, your manager — not the chain of the person who volunteered. 3. The volunteer, who is not being paged, has no idea an incident started. 4. When someone eventually asks who was on call, the record says you were. The cost is measured in the minutes escalation spends trying to reach an unreachable person, at exactly the time when minutes matter. Nothing about the informal agreement was wrong except that it never reached the only system that acts on it. ## The practices that follow **The schedule is the single source of truth.** If a coverage change is not in the schedule, it did not happen. This is worth stating explicitly as a team norm, because the failure is quiet — it only shows up during an incident. **Use overrides for short planned absences.** Appointments, travel, school runs, anything where you know in advance you cannot respond. Enter it with the covering person's agreement and a start and end time. **Use swaps for whole shifts.** A trade of entire weeks or days belongs in the schedule so that everyone's future load is visible and nobody quietly ends up carrying two shifts in a row. **Review the coming weeks for gaps.** Someone — usually the rotation owner — should look a few weeks ahead, especially around public holidays and end-of-year periods, and confirm every hour has a named, agreed owner. Holidays are exactly when coverage is thinnest and change freezes make lingering issues more likely to matter. **Verify the covering person can be reached.** New joiners, changed phone numbers and notification settings silently muted by a phone upgrade are all common. A page that routes correctly to a device set to Do Not Disturb has the same effect as no override at all. ## Overrides as a health signal Occasional overrides are normal and healthy — they are the mechanism that makes on-call compatible with having a life, and a team where nobody ever takes one is usually a team where people feel they cannot ask. But chronic overrides say something structural: a rotation too small for people's actual lives, shifts that land on the same person's difficult days, or one person quietly absorbing everyone else's coverage. Reading the pattern over a quarter is more informative than reacting to any single request. ## What a good answer sounds like Name the mechanism, name the reason it exists — routing follows the schedule, not the conversation — and give the concrete failure: a page rings the wrong phone while the person who agreed to help sleeps through an incident they never heard about.