Solo On-Call: The Hidden Cost of One-Engineer Rotations
Updated May 2026. Sources: Google SRE Book Chapter 11 (Being On-Call), incident.io 2024 State of On-Call Report, AASM shift-work research, SHRM replacement cost formulas.
The Default Pattern at Small Startups
Solo on-call is rarely a designed rotation pattern. It is the default outcome of being small enough that there is exactly one engineer (or one founder-engineer) who knows enough about the production system to respond credibly to a page at 3 AM. Other people on the team may technically be able to take a page, but only one person has the system context to actually mitigate most incidents without escalation, and so the pager defaults to that person permanently.
At pre-seed and seed-stage companies the pattern is almost universal. The founding-engineer-as-permanent-on-call is rarely scrutinised because the engineer does not have time or political capital to complain, and because the perceived alternatives (hiring a second engineer with operational depth, outsourcing first-response) cost real money that the company does not have. The 168 hours per week of pager exposure becomes invisible because it has always been there and because the engineer has internalised it as part of the role.
The pattern persists longer than it should at growing companies because each successive engineering hire is brought in for product work rather than operational depth. By the time the company is at 10 to 15 engineers, the founding engineer often still holds the pager because no one else has been deliberately backfilled into operational responsibility. This is where the structural cost becomes critical.
The 168-Hour Math
Solo on-call means 168 hours per week of pager-eligible time. The engineer is not actively responding for most of that, but the anticipatory cognitive load is continuous. AASM and broader sleep-medicine research on shift workers, physicians on permanent call, and emergency-services workers consistently shows that being permanently on call (as opposed to in a finite shift rotation) carries cognitive and physiological costs that compound over months. Sleep architecture degrades even on nights without active pages because the autonomic system maintains low-grade vigilance. Recovery during pager-free vacation takes 2 to 4 weeks to fully restore, which means anything less than a full month of leave annually leaves the engineer in a persistent partial-recovery state.
Active page response is the visible part of the burden. At a startup median of roughly 15 to 30 pages per week (small startups have fewer pages than the incident.io 2024 median of 42 per week, simply because the system is smaller), the active page-handling time at 35 minutes per page (12 minute MTTA plus 23 minute Gloria Mark refocus) is 9 to 18 hours per week. At a founder-engineer fully-loaded cost of $250,000 annually, that is $52,000 to $104,000 per year of direct active-page time. This is before night-page premium, before the recovery cost, and before the contingent attrition cost.
The compounded total cost of a solo on-call founder-engineer for one year is realistically $150,000 to $300,000 in direct time, recovery, and operational impact, before any catastrophic outcome (attrition, mental-health event, missed product window because the founder is too tired to think strategically). It is one of the most expensive engineering practices that small startups maintain, precisely because it is invisible and unmeasured.
Attrition Probability
The incident.io 2024 State of On-Call Report found that 41 percent of on-call engineers across all rotation patterns had considered leaving specifically because of alert volume. Solo on-call sits at the high end of the distribution because the burden is continuous and unrelenting. Adjusting the headline 41 percent for solo-on-call selection (no rotation relief, no rotation-based recovery, no peer support during pages), the realistic attrition-consideration rate is closer to 60 to 75 percent over a 12-month window.
The conversion rate from considering to actual departure is lower (maybe 25 to 35 percent of those who consider leaving actually do, on a 12-month horizon). For a solo on-call engineer, that gives a realistic annual departure probability in the 15 to 25 percent range. Multiplied by the SHRM 1.0 to 1.5x salary replacement cost for a senior technical hire, the expected annual attrition liability is $30,000 to $75,000 for a $200,000-salary engineer, before the gap-cost premium of having no on-call coverage during the recruitment window.
For a founder-engineer the calculation changes shape: the founder rarely literally leaves the company, but the equivalent failure mode is the founder becoming too burnt out to do the strategic work that only the founder can do. This is harder to quantify but more expensive when it happens; many small-company failure stories trace back to a founder-engineer who was too operationally overloaded to make a critical product, hiring, or fundraising decision in time.
The Realistic Alternatives
At a 2-to-3-engineer team, the only structural alternative to solo on-call that costs less than zero in productivity is the alternating week-on-week-off rotation. Two engineers each take a full week of primary on-call, with the off-week engineer truly off-pager. This is worse than a 4-person weekly rotation but materially better than solo because the off-week gives the off-call engineer time to recover and to do focused engineering work without pager interruption. Pair this with aggressive alert hygiene targeting under 5 pages per engineer per week (per /alert-tuning), and the practice becomes sustainable for 12 to 18 months at a stretch.
At a 4-to-6-engineer team, a standard weekly rotation across all four-to-six engineers becomes structurally viable, with each engineer carrying the pager roughly every fourth-to-sixth week. This is the cheapest sustainable model and is the default at most engineering-led startups by the time they reach Series A. The investment required is not financial; it is the willingness to spread operational responsibility across a wider set of engineers, which often requires explicit onboarding into the system and runbook creation for the most common alert classes.
For teams that cannot yet support a multi-engineer rotation, outsourced first-response (PagerDuty on-call-as-a-service or comparable) at roughly $200 to $400 per month per shift provides a safety net that handles routine runbook execution and escalates only when genuine engineering work is needed. This works well when more than half of pages can be runbooked; it works badly when pages typically require deep system knowledge. Evaluate by looking at the last quarter of pages and counting how many could realistically have been handled by a first-responder with a written runbook. If more than 50 percent, outsourced first-response is a sound bridge to a multi-engineer rotation.