Managed services: what 24/7 really means on a Sunday

Sunday, 14:40 in New Delhi, 05:10 in Mississauga. An alert fires. What happens over the next ten minutes has almost nothing to do with the alerting stack and almost everything to do with decisions taken months earlier, most of which were staffing decisions.

A rota is an arithmetic problem first

Continuous cover for one role is 168 hours a week. Staff it with three people and the rota has a resignation built into it. Honest cover needs about five, before holiday, illness, training and the certainty that somebody will eventually leave. Suppliers sell round-the-clock support against a team of three and then meet the arithmetic in month seven, usually during an incident.

Our two centres sit in Mississauga and New Delhi, nine and a half hours apart in summer and ten and a half in winter. That gap is what makes daylight cover on both sides possible without putting most of the rota on permanent nights. It also creates a handover twice a day, which is where the real operating cost of the model lives.

Handover is a written artefact

A verbal handover call is pleasant and does not survive contact with a Sunday. What survives is a written record with a fixed shape: what is open and its current state, what has been tried and ruled out, what is scheduled to run in the next twelve hours, which client contacts already know, and what the incoming shift should deliberately ignore because it is known and tracked.

The line most often missing is the ruled-out list. Without it the incoming engineer spends their first hour repeating the outgoing engineer's last hour, which is how a two-hour incident becomes a six-hour one with a regional boundary in the middle of it.

Escalation fails at the third name

Escalation paths get tested at the first name and almost never at the third. The first name is on the rota and expecting the call. The third name is a senior engineer who changed mobile number in March, or a client-side approver whose deputy has since left. On a Sunday afternoon the third name is exactly who is needed, because the first two have concluded that this requires a decision nobody on shift is permitted to make.

We test the path rather than the pager. Quarterly, out of hours, against the list as written, including the client-side names. It is a mildly irritating exercise for everyone involved and it has never yet been a waste of time.

Cover and capability are separate purchases

A weekend responder can restore service. Explaining why it broke usually requires whoever built the integration. Clients reading a managed service proposal tend to assume they have bought both, and the gap shows up as a Sunday restoration followed by a Thursday recurrence.

That is why we keep people from the build inside the operate phase and inside the rota. A support model staffed entirely by people who inherited the system reads well on a price sheet and holds up for roughly a year, at which point the recurring incidents start winning.

Communication has its own rota

Half of what a client remembers about a weekend incident is when they were told and how often. Set the cadence in the contract: who sends the first notification and within how long, the update interval while the incident is open, who is allowed to say the words that carry commercial weight, and what goes out when there is nothing new to report. Silence on a Sunday reads as absence, even when four people are working the problem.

The best possible Sunday shift is one where nothing happens and several people are paid to be available anyway. That standing cost is the product. Anyone quoting continuous cover cheaply is quoting for a phone that gets answered, which is a different thing, and clients find out which one they bought at about five in the morning.