Skip to content

Security

MSP vs MSSP: should your IT provider also do your security?

One vendor is fine until somebody outside the company asks you to prove something. After that, the test needs its own contract, scope and report, and, for an estate the same firm runs, a different firm.

Computing America IT. Help desk support, backups and network infrastructure.

Send the tier sheet
Updated 5 min readComputing America
A safe deposit box with two keyholes and a different key in each.

In short

  • A team assessing work it configured inherits the assumptions it made while configuring it. That blind spot involves no dishonesty, and no amount of skill removes it.
  • Security work has no per-seat shape: most months are close to nothing, then one alert takes several people and a day or more. When the fee is fixed per seat, the incentive is to deliver the part that fits the price, and triage doesn't fit.
  • In a real incident someone has to be able to say "stop, this is worse than it looks, take it offline." A firm has a conflict when the system in question is its own work.
  • Patching, endpoint protection, tested restores, MFA and joiners-movers-leavers are operational hygiene and belong to whoever runs the estate. Adversarial testing is a separate practice with its own contract and scope, and for an estate one firm runs, it comes from a different firm.
  • When the test is a separate engagement, its findings go to you and not to the team that did the work, and each becomes a ticket with an owner and a date. A finding forwarded by email lands in a queue with no owner.

The trigger is the first outside request for proof: a customer's questionnaire with contractual weight, an insurer asking about your controls, or a contract that flows a requirement down to you. From then on, whoever runs the estate shouldn't be the only one grading it. Before then, one vendor is usually enough.

An MSP, a managed service provider, runs your IT estate: the helpdesk, patching, endpoint protection, backups and the administrative credentials. An MSSP, a managed security service provider, is paid for security work: reading what the tools flag, answering alerts and incidents, and testing the estate as an attacker would.

The bundling pitch is fair: the MSP already holds the administrative credentials, and a second vendor means paying someone to learn what it knows. Once proof is requested, we still advise against buying the test that way, for three reasons. None is a claim about competence. All three are about where the team stands when it does the work.

One: the assessor is the assessed

The team that segmented your network decided where the boundaries go. Asked six months later whether the segmentation holds, they test the boundaries they drew. Assumptions made while building a thing are invisible from inside it.

Someone reading the estate cold has no boundary to defend. They look at the firewall rule that predates the contract, the service account whose password was set years ago, and the integration orphaned when its requester left.

Two: the pricing decides what gets delivered

When security is priced per seat per month, like the helpdesk, the price fits support, where work tracks headcount. Security work has no such shape. Most months it's close to nothing. Then an alert looks wrong, and answering it properly takes several people, a day or more and calls to vendors. With a fixed fee, the incentive is to deliver the part that fits the price.

Fits the priceDoesn't fit, and is the point
An EDR agent on every endpointSomeone reading what it flagged, the day it flagged it
A monthly posture report from a consoleJudgment about which of its findings matter here
An annual vulnerability scanChaining medium findings into the one that would work
A policy pack adapted from a templateEvidence that the policies describe what the estate does
What survives a per-seat bundle, and what you were actually buying

Three: somebody has to be able to say stop

The moment that decides what a security arrangement was worth is short and verbal. Something is wrong, the picture is incomplete, and the responsible advice is to take a production system off the network now and work out why afterwards. NIST SP 800-61r3 (opens in a new tab) says the response policy should state "which roles have the authority to confiscate, disconnect, or shut down technology assets." The plan has to name who can call it.

Suppose that decision sits inside a bundled contract. The advice is expensive and disruptive, it might be wrong, and it concerns an estate the same firm configured. Giving it means saying the firm's own work may be at fault, at the cost of a day of production. The same conflict applies to us wherever we run your estate and also test it.

Where the line sits

None of this means an IT provider should carry no security duty. Most of what keeps an ordinary business safe is operational hygiene, and a specialist who doesn't hold the keys or watch the queue would do it worse.

Ours, because we run the estateChecked separately, because assurance is the product
Patching on a schedule you can seeTesting whether it happened everywhere
Endpoint protection deployed, updated and monitoredAdversary emulation that tries to get past it
Backups running, restores tested and evidencedWhether the backups survive an attacker holding your credentials
MFA, least privilege, joiners, movers, leaversIdentity attack paths: what the third account reaches through the second
Answering the questionnaire about how the estate is runSomeone who didn't do the work confirming the answers are true
The seam: each operational duty, and the separate check on it

A tested restore proves reliability, not resistance to an attacker. We run the left column.

What the separation is, and what it is not

We sell both managed IT and security testing, and a second contract does not make us a second firm. A separate engagement gives you a written scope and a report addressed to you, not to the IT team. It does not remove reason One. When we run your estate, a test by our cybersecurity division is a review, not independent proof. For independent proof, use another firm. Where another firm runs your estate, our cybersecurity division can be that firm.

Two firms also mean a coordination problem. The failure to prevent: a finding is delivered to you, you forward it, and it lands in a queue with no owner or date. Ask for the report in your name and for each finding to become a ticket before you accept it.

Fig. 1
  1. Group: Bundled

    Runs the estate

    Drew the boundaries

    • To Tests it
  2. Group: Bundled

    Tests it

    Tests the boundaries it drew

    • To Reports to IT
  3. Group: Bundled

    Reports to IT

    Not to you

  4. Group: Separate

    Test by another firm

    Own contract, written scope, reads the estate cold

    • To You · report
  5. Group: Separate

    You

    Report in your name

    • To Runs the estate · ticket
  6. Group: Separate

    Runs the estate

    One ticket per finding, a named owner and a date

In a bundle the team that drew the boundaries tests them and reports to IT; a separate test reports to you, who turn each finding into a ticket.

SourceThe handoff this article recommends; no measured data.

When one vendor is the right answer

A thirty-person company with no regulated customers and no contractual security requirement needs patching, MFA, backups that restore and somebody competent answering the phone. A second firm to assess an estate that simple spends assurance money before there's anything to assure.

An incident is the other point to separate the test, because afterward you need a defensible account of what happened more than a fix.

Three questions for a provider offering both

  1. 1.Are the people doing the security work the same people who administer our estate? If yes, what stops an assessment from testing only the parts they built?
  2. 2.Show me a finding you raised about your own configuration. Not a vulnerability in a vendor's product: something you got wrong and reported anyway.
  3. 3.In an incident, who decides to take a production system offline, and is that authority written down before the incident?

Sources

  1. 1.SP 800-61r3: Incident Response Recommendations and Considerations for Cybersecurity Risk Management (opens in a new tab), NIST

Next step

Send us the tier sheet.

Send the page where your provider lists what each tier includes. We will mark which lines are support work, which are security claims, and which ones describe a result nobody is named against.

Reply
A person replies, not a sequence: within one business day, from someone who would be on the engagement.

Packages

Start with a package

Four packages, each with its price on the page. Pick one and buy it online.

Billing