ADAuditcover Direct
Guides › Security Controls

Your First 90 Days With an MDR Provider: An Onboarding Checklist for Buyers

The contract is signed, but your coverage isn't real yet. Use this phased 90-day MDR onboarding checklist to confirm sensors, log sources, response authority and SLAs match what the provider sold you.

Auditcover Direct editorial team · November 2026 · 8 min read

Turn the Signed Proposal Into Your Onboarding Baseline

Once the MDR contract is signed, the provider starts calling you a customer. Until the sensors are reporting, the logs are flowing and the analysts are allowed to act, you aren't protected. That gap is where most MDR relationships go wrong. Pricing is rarely the problem. The usual cause is that nobody on the buyer's side checked whether the service that went live matches the one that was sold.

This MDR onboarding checklist splits the first 90 days into phases with clear exit criteria. Before kickoff, pull together three documents: the statement of work, the service description or service level agreement, and any slides or emails from the sales cycle that made specific promises, such as "we'll ingest your Microsoft 365 and firewall logs" or "we'll isolate hosts 24/7 without waiting for you." Copy every concrete promise into a single tracker with an owner, a due date and a pass/fail field.

Advertisement

Don't skip this. Sales engineers sometimes promise integrations or response actions that the delivery team doesn't treat as in scope. If a promise isn't in your tracker, it won't come up at the 90-day review, and you'll lose the leverage to enforce it.

Days 1–10: Kickoff, Named Contacts and the Escalation Tree

The kickoff call should leave you with names, not just a support portal. Ask for your onboarding lead, the customer success or service delivery manager, the escalation path inside the SOC, and how commercial disputes get escalated. Record each person's hours and time zone. A provider with SOCs in only one region may handle night-time escalations differently from daytime ones, and you need to know that before an incident.

On your side, build an escalation contact list the provider can use at 3 a.m.:

  • Primary and secondary technical contacts with mobile numbers, plus a third fallback for holidays.
  • An executive decision-maker who can approve high-impact actions such as shutting down a production server or disabling a privileged account.
  • Your cyber insurer's breach hotline and your panel incident response firm, if your policy requires them. Many policies limit cover when you use a non-panel firm without consent. Your broker can confirm the terms.
  • Legal or privacy counsel. Notification deadlines vary widely by country and sector. The EU and UK GDPR set a short window for notifying the regulator, India's CERT-In directions require rapid reporting of certain incidents, and regulators in the Gulf, Singapore, Brazil and elsewhere have their own rules. Have counsel confirm which apply to you.

Exit criteria: both parties have signed off the contact list, the provider has tested at least one phone escalation to a named person, and you've agreed how often the list gets reviewed (quarterly is reasonable).

Days 5–30: MDR Deployment and Sensor Coverage

MDR deployment depends on the endpoint agent, whether it's the provider's own or a third-party EDR they manage, such as CrowdStrike, Microsoft Defender for Endpoint or SentinelOne. Ask for a coverage report against your asset inventory, and don't accept the provider's own agent count as proof. Their dashboard can only show machines that already have an agent.

Reconcile three lists: your CMDB or asset inventory, your identity provider's device list, and the MDR console. Machines that appear in the first two but not the third are your blind spots. Expect gaps in Linux servers, legacy Windows builds, VDI images, OT-adjacent workstations and contractor machines.

Set realistic targets. For most mid-size companies, 95% or more of in-scope endpoints and servers reporting by day 30 is a reasonable goal, and anything below 90% at day 45 is a red flag. Either the deployment has stalled, or the agent doesn't support part of your estate, which the provider should have raised during scoping.

Check billing at this stage too. Typical 2026 mid-market MDR pricing runs roughly USD 8–25 per endpoint per month, sometimes with an onboarding fee. Prices vary by region, contract length and volume, so confirm with your vendor. If you're billed on contracted endpoints, the provider has every reason to finish deployment quickly. If you're billed on deployed endpoints, watch for counts that creep past the number you scoped.

Days 15–45: MDR Log Source Integration by Priority

Endpoint telemetry alone misses identity attacks, email compromise and cloud misconfiguration. MDR log source integration is where scope disputes come up most often, because ingestion volume costs the provider real money. Integrate in order of what attackers commonly use against companies like yours.

PriorityLog sourceWhy it mattersTarget by
1Identity provider (Entra ID, Okta, Google Workspace)Account takeover and MFA fatigue attacks show up here firstDay 20
1Email and collaboration (Microsoft 365, Google Workspace)Business email compromise, mailbox rules, OAuth app abuseDay 20
2Firewall and VPN or ZTNARemote access abuse and exfiltration pathsDay 30
2Cloud control plane (AWS CloudTrail, Azure Activity, GCP Audit)Privilege changes, exposed storage, crypto-miningDay 35
3Critical SaaS and backup platformsAttackers often target backups before deploying ransomwareDay 45

For each source, confirm three things. Logs are arriving. Detections are enabled against them (ingesting a log isn't the same as monitoring it). And retention meets your obligations. Ask where the data is stored. Data protection rules such as GDPR, India's DPDP Act, Saudi Arabia's PDPL and Brazil's LGPD can affect whether logs may leave the country or region, so involve your privacy lead before you approve a storage location.

Days 20–45: Write the MDR Response Authorization Matrix

Detection without action is an expensive alerting service. The MDR response authorization matrix sets out which actions the provider can take on its own, which need your approval, and how long it will wait for that approval before falling back to a default. Get it written and signed. Verbal agreements from the kickoff call won't hold up during an incident.

ActionTypical defaultConsider changing if
Isolate a standard workstationPre-authorized, notify afterRarely; this is the main value of MDR
Isolate a serverApproval requiredPre-authorize non-production servers by tag
Disable or reset a user accountPre-authorized for standard usersExclude executives and service accounts, or require a call first
Revoke sessions and tokensPre-authorizedKeep pre-authorized; low business impact
Block a hash, domain or IPPre-authorizedRequire approval for blocks on critical vendor domains
Kill process or quarantine filePre-authorizedExclude line-of-business application servers

For any action that needs approval, add a timeout rule: "If no named contact responds within 30 minutes, the provider may proceed with isolation." Without one, a ransomware intrusion at 2 a.m. can sit waiting for someone to pick up the phone. Tag crown-jewel systems in the EDR console so the matrix applies to groups of hosts instead of listing them one by one.

Days 30–75: Alert Tuning Without Hiding Real Threats

The first month usually produces too many alerts. Admin tools, vulnerability scanners, backup agents and IT scripts all look suspicious to a detection engine that doesn't know your environment. Tuning is necessary, but every exclusion is a place an attacker could hide.

Set rules for how tuning happens:

  1. Every exclusion is logged with who requested it, why, what it suppresses and when it gets reviewed.
  2. Exclusions are as narrow as possible. Exclude a specific signed binary on specific hosts, never a whole directory or a tool across the estate.
  3. Remote access tools get special scrutiny. Attackers use the same AnyDesk, ScreenConnect or PowerShell activity your helpdesk does.
  4. The provider explains volume trends. A sharp drop in escalations can mean tuning worked, or that coverage quietly broke. Ask which.

Say a 300-person company sees around 40 escalations in month one and about 8 in month two. That looks like progress. But if the drop happened the same week the firewall integration stopped sending logs, it isn't. Check source health before you celebrate.

Days 75–90: The Formal MDR Service Level Review

Book the MDR service level review at kickoff so it doesn't depend on the provider's availability. Ask for the data a week in advance. Bring your security lead, a finance or procurement representative, and ideally your broker if MDR affects your cyber insurance terms.

Measure against the contract, not the provider's standard QBR deck:

  • Time to acknowledge and time to escalate for critical alerts, checked against the SLA using your own records as well as theirs.
  • Time to contain for pre-authorized actions. Ask for the median and the worst case, not just the average.
  • Coverage: percentage of endpoints reporting and the status of each contracted log source.
  • Tracker closure: every sales-cycle promise from day one, marked pass or fail.
  • Quality of investigations: pick three escalations and read the write-ups. Do they say what happened, what was done and what you need to do next?

Also ask the provider to run a tabletop or live response test. A benign simulation, such as a known-safe test file or an atomic test technique agreed in advance, shows whether the matrix and the escalation tree work. A report can't tell you that.

Red Flags That Justify a Commercial Escalation

Some gaps are normal friction. Others mean you're paying for a service you aren't getting. Escalate in writing to the provider's account leadership if any of these are still true at day 90:

  • Endpoint coverage is below 90% with no remediation plan or dates.
  • A contracted log source isn't ingesting, or is ingesting with no detections enabled.
  • The response authorization matrix still isn't signed, or the SOC hasn't followed it.
  • SLA metrics are reported only by the provider and can't be checked against ticket timestamps.
  • Exclusions were added without your approval or a record.

Use this decision rule. If two or more flags are open at the review, ask for service credits or a fee hold on the affected scope, plus a 30-day remediation plan with a re-review date. If the same flags are still open at that re-review, start planning your exit. Check notice periods and data return terms with your legal advisor now, while you still have time to act on them.

Frequently asked questions

How long should MDR onboarding take before we're actually protected?

For a mid-size company, endpoint agents and basic monitoring are often live within two to four weeks. Full log source integration and a signed response matrix commonly take 45 to 60 days, though timelines vary by provider and estate size. You should be partly protected from the first weeks, so agree response authorization for standard workstations early instead of waiting for everything else to finish.

Should we let the MDR provider isolate machines without asking us first?

For standard workstations, yes. Fast isolation is a big part of what you're paying for, and waiting for approval at night defeats the point. For production servers, executive accounts and line-of-business systems, most buyers require approval but add a timeout so the provider can act if no one responds within a set window.

Does using an MDR provider reduce our cyber insurance premium?

It can help. Many insurers ask about 24/7 monitoring and EDR in their applications, and having these controls may improve your terms or eligibility. Results vary by insurer, country and your overall risk profile. Ask your broker how your carrier treats MDR, and check whether your policy requires a specific incident response firm during a claim.

Written by the Auditcover Direct editorial team. Prices, features and rules change often - confirm details with the vendor or a licensed professional before you decide.