Few decisions in security operations are as consequential, or as often made by habit, as how to staff the monitoring function. Build a team, buy a service, or blend the two? The honest answer depends on what an organisation can sustain at three in the morning, not on what looks neat on a slide. This comparison weighs the three models for SOC Ops along four axes: coverage, cost shape, control and accountability.
Start SOC Ops planning with the arithmetic of coverage
Every comparison should begin with the same sum. MITRE offers a planning figure: keeping a single desk occupied twenty-four hours a day takes about five employees, after leave, sickness and training are counted. The Atlas works through a queueing calculation and lands at 5.03 per seat, which agrees. It adds a floor of two people on duty in every hour. For a company that wants continuous coverage with two seats, that is already about ten people before any specialist or manager is counted.
The Ops Atlas discussion of SOC Ops lists build-versus-buy among the core scope questions, alongside tiers, playbooks and metrics. It treats the choice as a question of fit, and the NCSC, which it quotes, puts the point plainly: there is no one-size-fits-all SOC.
Model one: an internal team
Strengths. Analysts learn the business. They know which unusual login is a regular contractor and which is not, so more alerts close quickly as benign. Escalations travel short distances, and knowledge stays in the organisation.
Weaknesses. Continuous coverage is expensive and hard to staff. Recruiting, retaining and rotating analysts is a permanent job, and the Atlas lists burnout as a core operating concern. A team too small for night shifts will quietly leave them thin.
Suits: Large organisations with steady alert volume, or those whose data cannot leave their boundary.
Model two: an outsourced provider
Strengths. Coverage comes with the contract. A provider spreads its staff across many customers, which makes round-the-clock service far cheaper than building it alone. Onboarding can be quick.
Weaknesses. Context is thinner. A provider sees your alerts but not your calendar, so it escalates more and closes less confidently. The Atlas also warns about contracts that measure providers on how fast tickets close: speed targets reward early closure and weak investigation, and reopen rates climb as the target bites. And outsourcing does not move responsibility. MITRE is explicit that system owners and executives stay accountable.
Suits: Smaller organisations that need continuous cover but cannot staff it, provided the contract measures quality, not only speed.
Model three: the hybrid
Strengths. A common split keeps daytime triage and investigation in-house while a provider covers nights and weekends, or a provider handles first-line triage and escalates to an internal specialist. The Atlas lists both managed security services and managed detection and response as the outside options.
Weaknesses. The seams are where things fall. Two parties need one queue, one definition of severity and one handover record. Without a shared pass-down log, the hybrid model becomes two partial teams who each assume the other has the incident.
Suits: Mid-sized organisations that want local knowledge by day and economical coverage by night.
The comparison at a glance
- Coverage: internal is the hardest to guarantee; outsourced is the easiest; hybrid depends on the seam.
- Cost shape: internal is mostly fixed salary; outsourced is a recurring fee; hybrid mixes both.
- Control and context: internal is highest; outsourced is lowest; hybrid sits between.
- Accountability: identical in all three. It stays with the owner.
One practical way to compare them is to price the worst hour, not the average one. Ask what each model does when two serious alerts land together on a Sunday night: who is awake, who decides, and who tells the executives. An internal team of five may have one person reachable. A provider may have many analysts but no authority to disable an account. A hybrid may have both, if the escalation path is written down. The answer to that single scenario usually reveals more than a price list.
Questions to settle before signing anything
Whatever the model, a few requirements belong in writing. Every alert must carry a recorded outcome and a reason. Median and 90th-percentile triage times should be reported by shift and severity, not as one average. The wait should be timed until a person first acts, because automated tagging can flatter the figures. And there should be a stated escalation path with context attached, so no one gets a bare ticket at 3 a.m.
Tooling sequence matters too. Government guidance quoted by the Atlas says automation is usually unsuitable for immature environments, so the central log platform comes first, with automated steps added later. Ask any provider how it handles that order, and where your data sits. The Atlas also calls for keeping monitoring infrastructure in a protected enclave, since leaked monitoring data helps intruders hide their tracks.
Reading the choice like an infrastructure decision
This is not unlike sizing capacity for a busy site. Our article on handling surges of web visitors on cloud infrastructure argues that you plan for the peak and the odd hour, not the mean. And the checklist in the full engineering blueprint for web project requirements shows why a written requirement list beats good intentions when someone else does part of the work.
The verdict: pick the model whose weak points you can compensate for. Internal teams must solve staffing; providers must be steered with the right metrics; hybrids must nail the handover. In all three, the organisation keeps the accountability and the obligation to ask for evidence, not reassurance.

