Being on-call is a restriction on liberty. An engineer on-call cannot drink alcohol at a dinner party, cannot go camping without cell coverage, cannot take a long bath without their phone nearby. This restriction has real value, and the company is consuming it. They should pay for it.
Model 1: "It's in the Salary" (The Problematic Default)
Method: No extra pay. On-call is described as "part of the job" in the offer letter.
Problem: This model creates a perverse incentive for management. If on-call costs the company nothing extra, there is no financial pressure to fix noisy alerts or invest in system stability. The pain is externalized entirely onto engineers. This is the fastest path to attrition.
Model 2: Flat Stipend + Hourly Incident Pay (The Hybrid)
Method: Engineers receive a flat weekly stipend simply for carrying the pager — compensating for the restriction of availability. If they are actually paged and must work, that time is logged and paid at an overtime rate (commonly 1.5x base hourly).
Pros: Fair and transparent. It acknowledges two distinct costs: the cost of being available (stipend) and the cost of actually working (hourly). It also gives finance a real number that makes alert noise visible as a budget line item.
Example structure:
- Availability stipend: $300/week for carrying the pager
- Incident pay: 1.5x hourly rate for any time spent actively working an incident, logged in 30-minute increments
Model 3: Time in Lieu (TOIL)
Method: For every hour spent working an incident outside business hours, the engineer accrues equivalent vacation time to use later.
Pros: Directly addresses the recovery need. If an engineer worked from 2 AM to 5 AM, they should sleep until noon the next day — and TOIL makes that official policy rather than a favor.
Cons: In chronically understaffed or high-incident teams, accrued TOIL never gets used. Engineers accumulate "debt" that the company never actually pays back. This creates resentment. TOIL only works if the team has enough slack capacity to actually take the time off.
The Pain-Sharing Principle
The single most effective incentive for improving system stability is making on-call pain visible to leadership. If directors and VPs join the rotation — even in a Secondary or observer capacity — the organizational priority of fixing noisy alerts changes overnight. This is not punitive; it is alignment. Leaders who experience the pager firsthand make better investment decisions about reliability work.
Key Point: Compensation models are not just about fairness — they are feedback mechanisms. A model that makes on-call pain financially visible to the organization creates pressure to reduce that pain through engineering investment. A model that hides the cost removes that pressure entirely.