A small product firm is full of one-person dependencies. One engineer knows the billing job. One customer success manager holds the relationship with the largest renewal. One support lead decides what goes to engineering at night. When any of them is ill or on holiday, the company discovers just how much of its calm depended on one person.
How do you stop leave and sickness becoming emergencies? Begin by listing duties instead of people.
Begin with a map of duties that cannot wait
Walk through a normal month and write down every task that causes a customer or a colleague pain if it slips for three days. Most teams find the same short list.
- The pager for production alerts and the first response to an outage.
- The release and deployment window, including the rollback decision.
- The support queue, split by language, tier and priority customers.
- Renewal and upsell calls on accounts with a date attached.
- Monthly inputs that finance and HR need from the team, such as attendance and variable pay data.
- Access and admin duties: who can add a user to the cloud console or reset a licence.
Put a second name beside each. A blank is your risk.
Make the second person real
A name on a sheet does not amount to cover. The backup needs access, context and one live rehearsal. For the pager, that means a shadow week where the backup is paged alongside the primary. For a renewal, a joint call before the leave starts, so the customer has heard the second voice. For deployment, a supervised release on a quiet day.
- 1
Shadow before the leave, not during it
The primary invites the backup to two real instances of the duty and narrates decisions.
- 2
Write the two-page runbook
Credentials location, the usual failures, whom to wake, what never to touch. Keep it in the repository wiki or the ticket.
- 3
Grant access in advance
Check logins to the cloud console, the helpdesk, the CRM and the billing tool a week before.
- 4
Mark the handover in the calendar
A fifteen-minute call on the last working day, with a note posted in the team channel.
- 5
Review after the return
A quick ten minutes on what broke, what confused the stand-in and what the runbook still lacks.
Planned leave and the six-week window
Most clashes can be seen weeks ahead. Each month the lead scans the shared calendar six weeks out. Where two people from the same duty group are out together, or leave overlaps a release, a quarter-end or a festival peak for the support queue, raise it early. Early objections are accepted gracefully. A refusal two days before a flight is not.
Sickness and sudden absence
Unplanned absence is where the runbook proves its worth. Agree a rule for the first hour: who tells the team, who checks the pager, who reassigns the queue. Keep it simple enough for a groggy Monday morning. Where a key account is affected, the second owner phones the customer the same day.
Cover for the leaders themselves
Founders and function heads are the last to get a reserve. A head of engineering on a long flight cannot approve a hotfix. A founder who alone holds the payment gateway keys cannot be reached for a failed subscription run. Set up a delegation note: who may approve spending, who may sign off a release, who may speak to the largest customers. Circulate it to the team and refresh it whenever someone senior will be away for a stretch.
Documentation that gets read
Runbooks fail when they sit in a folder nobody opens. Store them where the work happens: next to the code, inside the alert message, pinned in the support channel. Keep each to a page. Put the date of the last review at the top and ask the person who used it last to correct it. A runbook that is tested by a new joiner each quarter stays honest.
Where this meets wider planning
Cover planning is workforce planning in miniature. The HR strategy and workforce planning service maps one-person dependencies across the firm, while a tidy attendance, shift and leave process keeps the calendar truthful. Sector-wide patterns sit on the SaaS and product companies hub.
Budget a little slack
Cover has a price. A reserve engineer spends part of a sprint shadowing instead of closing tickets, and a buddy seller gives up a demo slot. Treat that as an investment and leave about a tenth of capacity free in planning, so the cost is visible and nobody feels punished for helping. Teams that plan to the last story point discover that the first illness breaks the sprint.
Signs the plan is working
- Nobody cancels a holiday because the pager has no backup.
- Planned leave is approved weeks ahead, not days.
- A new joiner can read the runbook and answer a basic alert.
- Customers rarely notice who was away.
