Selecting a managed detection and response (MDR) provider is not simply a decision about who will monitor your security alerts.
It is a decision about who will investigate suspicious activity, determine what matters, communicate what they know, and help your organization respond before an incident becomes a business disruption.
Those responsibilities are not interchangeable.
Two providers may offer 24/7 monitoring, threat intelligence, endpoint visibility, and incident response support. Yet their ability to investigate an attack, establish confidence in their findings, and support timely containment can be very different.
The challenge for security leaders is separating a provider's list of capabilities from the operational outcomes those capabilities actually produce.
Here are 12 questions worth asking before signing an MDR agreement.
1. What happens after an alert is generated?
Most MDR conversations begin with detection capabilities. Fewer explain what happens between an alert and a defensible security decision.
Ask the provider to walk through an actual investigation workflow.
Who receives the alert? What context is collected? How is suspicious activity validated? What determines whether an investigation is escalated?
A provider should be able to explain the process without relying entirely on product demonstrations or dashboard screenshots.
What to look for: A defined investigative workflow with clear ownership, evidence collection, escalation criteria, and documented decisions.
2. How do you distinguish a real threat from routine activity?
An alert indicates that something matched a detection condition. It does not necessarily establish malicious intent.
Investigators need to understand the surrounding activity, affected assets, identities, behavioral patterns, and available supporting evidence.
Ask how analysts distinguish malicious activity from legitimate administrative behavior, expected system changes, and other benign events.
What to look for: Contextual investigation and documented reasoning rather than alert severity alone.
3. What evidence supports your conclusions?
Security teams need more than a notification that something suspicious happened.
They need to understand what was observed, why it matters, what systems may be affected, and what remains uncertain.
Ask to see a sanitized investigation report.
Does it contain a timeline? Relevant telemetry? Affected identities and assets? Analyst conclusions? Recommended actions?
What to look for: Evidence that your team can independently review and use to support an operational decision.
4. How do you investigate activity across different security tools?
Attackers do not organize their activity around the boundaries of your technology stack.
An investigation may require endpoint, identity, network, cloud, and other relevant telemetry.
Ask which sources the provider can investigate, how information is correlated, and what happens when a required source is unavailable.
A provider should also explain which integrations are supported today and which require additional configuration, services, or development.
What to look for: Demonstrable cross-source investigation and transparent visibility limitations.
5. What does threat hunting actually mean in your service?
Threat hunting is frequently included in MDR marketing, but the term does not always describe the same operational activity.
Ask whether hunting is hypothesis-driven, how hypotheses are developed, which data sources are used, and how findings are documented.
Is hunting performed on a defined schedule, continuously where appropriate, or only after a detection triggers an investigation?
How are hunt results communicated when no malicious activity is found?
What to look for: A clearly defined hunting methodology, scope, frequency, and reporting process.
6. Who has authority to contain a threat?
Detecting malicious activity and stopping it are separate responsibilities.
A provider may identify a compromised endpoint but lack permission to isolate it. Another may have predefined authority to execute specific containment actions.
Neither model is automatically right for every organization.
What matters is whether authority, approval requirements, operational safeguards, and escalation paths are established before an incident occurs.
What to look for: A documented responsibility and authorization model identifying who can act, under what conditions, and with what safeguards.
7. How do you measure detection, investigation, and containment performance?
A single response-time metric rarely tells the whole story.
The time required to acknowledge an alert is not the same as the time required to investigate it, determine the necessary action, or contain a confirmed threat.
Ask the provider to define each metric, identify its starting and stopping points, and explain exclusions.
Also ask whether performance reporting distinguishes provider-controlled time from delays caused by customer approvals or missing telemetry.
What to look for: Consistent definitions, meaningful operational measurements, and transparent reporting.
8. What happens when your analysts cannot reach our team?
Incidents do not wait for convenient business hours.
Ask what happens when the designated customer contact does not respond, when an escalation fails, or when an authorized decision-maker is unavailable.
Does the provider have a documented escalation hierarchy? Are there predefined actions it may take? How are unresolved decisions recorded?
What to look for: Tested escalation procedures and explicit rules for situations requiring urgent action.
9. How do you verify that containment worked?
Issuing a containment command does not necessarily prove that the threat has been contained.
The action may fail, apply to only part of the affected environment, or leave another attack path available.
Ask how the provider verifies the result of an action and determines whether additional investigation is necessary.
What to look for: Confirmation evidence, defined success criteria, and a process for identifying incomplete or unsuccessful containment.
10. What will our internal team still be responsible for?
MDR does not eliminate the organization's responsibility for security decisions.
Internal teams may retain responsibility for asset ownership, identity changes, business-impact decisions, recovery, remediation, and regulatory obligations.
Ask for a responsibility matrix that covers normal operations and incident scenarios.
What to look for: Clear boundaries between provider responsibilities, customer responsibilities, and shared activities.
11. How will you demonstrate improvement over time?
An MDR relationship should produce more than a monthly count of alerts and incidents.
Ask how the provider identifies recurring investigation gaps, telemetry weaknesses, detection deficiencies, and opportunities to improve response procedures.
Can the provider show how findings lead to recommendations and how those recommendations are tracked?
What to look for: Reporting that connects investigative findings to measurable improvements in security operations.
12. Can you demonstrate the complete process using a realistic scenario?
A polished dashboard can demonstrate software functionality. It cannot, by itself, demonstrate operational effectiveness.
Ask the provider to walk through a realistic incident from initial observation to investigation, decision, authorized action, and verification.
Include complications: incomplete telemetry, ambiguous evidence, an unavailable customer contact, or a containment action that does not succeed on the first attempt.
What to look for: A credible end-to-end demonstration that exposes assumptions, dependencies, and operational limitations.
The decision is about outcomes, not feature counts
An MDR provider's value is not determined by the number of alerts it processes, the number of tools it integrates, or the number of features listed in a proposal.
Those capabilities matter only when they contribute to effective security operations.
Security leaders should understand whether the provider can establish what happened, determine what needs attention, support an authorized response, and produce evidence that the response achieved its intended result.
That is the difference between purchasing monitoring coverage and establishing an accountable detection-and-response capability.
Before choosing an MDR provider, ask for proof of the process—not just a description of the technology.
The most important question is not simply whether the provider can detect a threat. It is whether the organization can make and execute a defensible decision when that threat is detected.

