When the Night Shift Is Invisible: A SOC Ops Diagnosis

When the Night Shift Is Invisible: A SOC Ops Diagnosis

Picture a small security team whose weekly report always came out green. Median handling time looked respectable, the ticket count trended down, and nobody lost sleep over the dashboard. Then a new manager asked a simple question: what do these numbers look like at four in the morning? That question is where this story about SOC Ops begins, and the answer was not comfortable.

The symptom: a calm chart and a nervous feeling

Teams in this position usually describe the same feeling. The charts say things are fine, yet analysts mention that Monday mornings feel like cleaning up after a storm. Alerts that arrived over the weekend are sitting there, aged and unexplained. Nobody disputes the numbers; they simply do not match the lived experience of the people doing the work.

The mismatch is rarely dishonesty. It comes from how the number was built. A single average across every hour of every day blends a busy, staffed afternoon with a thin, nearly unwatched night, and the afternoon wins by volume.

The diagnosis: split by shift and look at the tail

A write-up of SOC Ops in the Ops Atlas gives a lab illustration of exactly this. Overall, the median time to start triaging an alert was seven minutes, which sounds excellent. Restricting the view to after-hours alerts, the median jumped to 512 minutes, over eight hours. Both figures were true. Only one of them described what an attacker working at night would experience.

The Atlas’s answer is to publish two figures, the typical case and the slow tenth, cut by shift and by severity. MITRE’s cautionary tale fits: one email left untouched for two days is enough to flatten the meaning of a mean, so distributions beat lone averages.

In our imagined team, splitting by shift took an afternoon of work and produced one uncomfortable chart. Day shift looked healthy. The overnight band looked like a different organisation.

The root causes behind a quiet SOC Ops night

Several causes tend to stack up:

  • Staffing arithmetic nobody did. Keeping one seat occupied around the clock takes about five people once leave, training and days off are counted. MITRE puts the figure at roughly five, and the Atlas’s own staffing calculation lands at just over five. A team of three cannot cover nights without gaps.
  • A queue that is never fully drained. If alerts arrive faster than the team can handle them, waiting time grows without limit, no matter how hard people work.
  • Automation counted as attention. When a tool tags or auto-assigns an alert, the timestamp moves, but no human has looked yet. The Atlas insists the triage clock should stop at the first human action.

The fix: smaller, specific changes

None of the repairs are dramatic. The first is to define a priority matrix and publish it. In the Atlas example, a high-severity alert on a top-tier asset is a priority one: triage within 15 minutes, a decision within four hours and any pre-approved containment within 30 minutes. A matrix removes the argument about what to do first at 3 a.m.

The second is a minimum of two analysts on shift at all times, per MITRE’s guidance, so that nobody triages a serious alert alone. The third is to give every alert a recorded outcome, chosen from four values (genuine, legitimate-but-flagged, misfire or unresolved) with a reason attached. That last step is what lets the team learn from nights it was not awake for.

What the team said when they saw the chart

The reaction in rooms like this is usually quieter than managers expect. Analysts rarely argue with the overnight chart; several say they already knew. One common comment is that the night pile gets skimmed rather than worked, because the person on duty is also covering a second job. Another is that escalating at night feels pointless when nobody senior will pick up the context until morning.

Those comments are the real data. They explain why the 512-minute figure existed and they point to cheap improvements: a named escalation contact for the overnight band, a short template that forces context into every escalation, and a standing review of anything older than a set age at shift start. None of this needs a new platform.

Handing the night over properly

Even a well-staffed team can lose information at the boundary between shifts. The Atlas describes a pass-down log that both the outgoing analysts and the incoming lead sign. Anything the incoming lead has not accepted stays with the outgoing owner. The ritual feels fussy right up to the morning a half-finished investigation arrives on the next desk with its notes still attached.

Measuring the repair

After a few weeks, our imagined team would expect to see the after-hours median fall toward the daytime figure and the 90th percentile tighten. They would also watch the reopen rate, since the Atlas warns that pressure to close quickly makes reopened cases climb. If the night number improves while reopens rise, the fix is cosmetic.

The wider lesson applies beyond security. A business is easy to run when you only look at the busy hours. Our own planning notes on how a growth roadmap ranks what to do first argue for the same habit of checking the parts of a process that do not show up in headline results, and the discussion of walking a new client through a services roadmap makes a similar point about handovers.

If your dashboard has been green for months, try one experiment this week: split every metric by hour and by severity, then read the slowest tenth, not the middle. The story you find there is usually the real one, and it is the first honest step in running SOC Ops as a service rather than a report.