At a board meeting someone puts up a screenshot from the Microsoft Defender portal: the company’s Secure Score sits well below the average for organisations of a similar size. The question for IT sounds simple, what do we need to click to make it higher, and that is exactly the question that leads you astray. A few “resolved through third party” statuses will raise the score while nothing in the tenant changes.
Secure Score is a good tool for managing Microsoft 365 cybersecurity, provided you know what it measures, because knowing how to improve Microsoft Secure Score starts with what the score counts. It counts Microsoft’s recommendations you have deployed rather than the likelihood of a breach, it has its own points arithmetic and a few statuses that change the score without changing protection. Once you know these rules, you can order the work by risk and show the board a trend instead of a single number.
What is Microsoft Secure Score and what does it actually measure?
Microsoft Secure Score is a measurement of an organisation’s security posture, with a higher number indicating more recommended actions taken. That is how the Microsoft Secure Score documentation defines it, and you will find the score itself in the Microsoft Defender portal. A single dashboard there shows the security of your Microsoft 365 identities, apps and devices.
Microsoft is explicit about what the score is not. Secure Score is a numerical summary based on system configurations, user behaviour and other security-related measurements, but it is not an absolute measurement of how likely a breach is. It shows the extent to which you use security controls that can help offset risk, and it is no guarantee against a breach.
That distinction shapes everything else. A high score says the tenant is configured in line with Microsoft’s recommendations. It does not say that nobody has taken over an account yet.
Key takeaways:
- Secure Score counts Microsoft’s recommendations you have deployed, not the risk of a breach.
- You see the score in the Microsoft Defender portal as a percentage and a number of points.
- Some points depend on the share of protected users and devices.
- Recommendation statuses can change the score without any configuration change.
How does Microsoft count points for recommendations?
According to the scoring description in Microsoft’s documentation, each recommended action is worth 10 points or less, and most are scored in a binary fashion: implemented earns the full value, not implemented earns nothing. Some recommendations are scored in proportion to how widely they are deployed. In Microsoft’s example, a recommendation worth 10 points for protecting all users with multifactor authentication gives 5 points when 50 out of 100 accounts are protected.
Points are awarded for three kinds of action: configuring recommended security features, doing security-related tasks, and addressing a recommendation with a non-Microsoft application or an alternate mitigation. That third kind comes back below, under statuses, because it earns points without any check by Microsoft.
The score updates in real time, and once a day Secure Score syncs data on the points achieved. On top of that, the state of Microsoft Teams recommendations is refreshed once a month and Microsoft Entra recommendations once a week. According to the description of assessing your security posture, after you complete an action it can take between 24 and 48 hours for the change to be reflected in the score.
How do you read the percentage and the comparison with other companies?
Read the score against what your company can achieve, not against the full scale. The “Your secure score” tile shows a percentage and the points achieved out of the total possible. The “Include” button adds three views: the planned score once planned actions are completed, the score achievable with your current licences, and the score achievable with your current licences and accepted risk.
The difference matters because Microsoft shows the full set of recommendations for a product regardless of licence edition, subscription or plan. According to the documentation, your absolute security posture, represented by Secure Score, stays the same no matter what licences you own. Some of the missing points may therefore come from features your company has not bought, and that is a budget decision, not an administrator’s oversight.
The page on score history and trends describes the comparisons: the average for organisations with a similar seat count, a custom comparison and score zones that you define yourself as good, okay and bad. The comparison data is anonymised. Even so, Microsoft advises in the Identity Secure Score description that instead of a specific number you should focus on the high-importance recommendations relevant to your organisation.
Which areas make up the score and what is the Identity Secure Score?
Recommendations are grouped into four areas: identity (Microsoft Entra accounts and roles), devices (Microsoft Defender for Endpoint), apps (email and cloud apps, including Office 365 and Defender for Cloud Apps) and data (Microsoft Information Protection). The overview page shows how points are split between these groups and how many are still available.
The product list goes beyond Microsoft 365. Besides Entra ID, Exchange Online, SharePoint Online, Teams and the Defender family, it also covers third-party services, including GitHub, Okta, Salesforce, ServiceNow and Zoom. Microsoft notes that the recommendations do not cover all the attack surfaces of each product, but they are a good baseline.
The Entra admin centre also shows the identity part as the Identity Secure Score. According to Microsoft, it is the same part of the score with the same recommendations, and the Identity Secure Score is available to both free and paid customers. Entra recalculates it every 24 hours, after comparing your configuration with recommended best practices.
How does Microsoft Secure Score differ from Secure Score in Defender for Cloud?
They are two different scores with similar names. Microsoft Secure Score covers Microsoft 365 and services connected to the Defender portal, while Secure Score in Microsoft Defender for Cloud assesses resources in Azure, AWS and Google Cloud against the Microsoft cloud security benchmark (MCSB).
The Defender for Cloud score is calculated differently: only built-in MCSB recommendations affect it, and Defender for Cloud calculates each control every eight hours for each Azure subscription or AWS and GCP connector. Microsoft also describes two models of that score: the new risk-based Cloud Secure Score in the Defender portal and the classic one in the Azure portal, with different calculations and values.
Secure Score for Azure is covered in the guide to Azure security best practices, and the article on securing AWS, Azure and GCP gives the context of the three public clouds. Record the distinction in board reporting, so that two percentages with the same name do not end up in one table.
Which recommendations should you deploy first?
Start with those that protect administrator accounts and sign-in for all users, and only then move to the rest. The portal ranks the list by points left to achieve, implementation difficulty, user impact and complexity, so actions with many points and little friction land at the top. That order suits the score, but not necessarily the risk.
Microsoft offers a hint here: it is more beneficial to have a high score on high-importance recommendations than a high score on low-importance ones. Three identity recommendations also come with a stated value. According to the description of security defaults, turning them on in Entra ID awards full points for requiring MFA for administrative roles (10 points), ensuring all users can complete MFA (9 points) and blocking legacy authentication (7 points).
The list of identity recommendations also includes items worth doing early: more than one Global Administrator, least-privileged administrative roles, not letting users grant consent to unreliable applications and managing local administrator passwords with Microsoft LAPS, described in the article on Windows LAPS. The What’s new in Microsoft Secure Score page shows that in January 2024 a recommendation to require phishing-resistant MFA strength for administrators was added.
How do you mark recommendations you will not deploy?
With the status that reflects the real decision, not the one that earns points. The portal offers these statuses: to address, planned, risk accepted, resolved through third party, resolved through alternate mitigation and completed. Microsoft confirms “completed” with its own data, and you cannot change that status manually.
Accepting the risk earns no points and stays in the history, where you can undo it. Both “resolved” statuses work the other way round: they award the full points, although Microsoft notes that it then has no visibility into the completeness of implementation. This is the only place where the score rises on an administrator’s word, without any check by Microsoft: the other statuses either earn no points or are confirmed by Microsoft’s own data.
That is why every such status deserves evidence: the tool, the scope and the date of verification in the recommendation’s notes, and for accepted risk, the risk owner and a review date. Microsoft itself suggests one justified case: with security defaults turned on, it recommends marking the sign-in risk policy and user risk policy recommendations as resolved through alternate mitigation instead of adding those policies.
Recommendations in the device group work differently. You do not choose their status in Secure Score but in Microsoft Defender Vulnerability Management: a global exception updates the status with the justification, while an exception per device group leaves the recommendation as “to address”.
Why does your score drop when nobody turned anything off?
Because some points depend on users and devices that change, and the list of recommendations grows. The Metrics & trends tab has a separate regression trend, that is, a timeline of points lost because of configuration, user or device changes. Next to it are charts of score changes, risk acceptance and comparisons with others.
The mechanism follows from proportional scoring. If a recommendation counts the share of protected users, every new account without the required configuration lowers that share and the points, even though nobody turned anything off. A new group of workstations without the required setting has the same effect.
The second cause is new recommendations. The What’s new in Microsoft Secure Score page lists items added month by month, for example in January 2024 for Entra, Microsoft Forms and SharePoint. A score shown as a percentage of the total possible points reacts to every new item your company has not yet deployed.
So before a drop goes into a report as a regression, check the History tab. Below the chart is a list of the actions taken in the selected period with their points and category, and each item opens together with the full history of the recommendation.
What will Secure Score not tell you about your tenant’s security?
It will not tell you whether someone is already inside. The score assesses configuration and deployed controls, and Microsoft itself states that it does not measure breach likelihood and that the recommendations do not cover all attack surfaces. A device registered by an attacker, an MFA method added to a compromised account or a mailbox rule hiding correspondence is not a missing recommendation but a trace you have to look for in the tenant’s logs and objects.
The second limit is the perspective of a single vendor. One external reference point is the CISA SCuBA project: the US agency CISA established it in 2022 and publishes secure configuration baselines for Microsoft 365. In 2023 CISA introduced ScubaGear, a no-cost tool that checks a tenant’s configuration against those baselines.
Two independent lists give a fuller picture than one number. Secure Score is well suited to tracking progress, while an external baseline and a review for signs of compromise answer questions the score does not ask.
How do you put the recommendations in the right order?
Set the order by what an attacker gains by taking over a given element, and treat the points as a measure of progress.
| Step | Recommendation in the portal | Points stated by Microsoft | What to watch out for | Evidence of closure |
|---|---|---|---|---|
| 1 | Require MFA for administrative roles | 10 | All privileged roles, not just Global Administrator | “Completed” status confirmed by Microsoft data |
| 2 | Ensure all users can complete multifactor authentication | 9 | Methods registered before the requirement is switched on | “Completed” status |
| 3 | Enable policy to block legacy authentication | 7 | Old mail clients and devices that send email | “Completed” status and no sign-ins over legacy protocols |
| 4 | Designate more than one Global Administrator, Use least privileged administrative roles | value in the recommendation panel | Permanent role assignments instead of just-in-time access | List of roles with owners |
| 5 | Do not allow users to grant consent to unreliable applications | value in the recommendation panel | Admin consent process for new applications | “Completed” status and review of existing consents |
| 6 | Protect and manage local admin passwords with Microsoft LAPS | value in the recommendation panel | Unmanaged devices | “Completed” status |
| 7 | Device group recommendations | value in the recommendation panel | An exception per device group does not change the status | Decision in Defender Vulnerability Management |
Microsoft states the points in the third column outright only for the recommendations covered by security defaults; you can see the value of the others in the recommendation details panel in your tenant. Steps 1–3 close the shortest route to account takeover, and steps 4–6 limit the impact of a takeover that has already happened.
How can nFlo help you assess your Microsoft 365 tenant?
The Microsoft 365 and Entra ID Security Review answers the question Secure Score does not ask: whether any traces of account takeover remain in the tenant. It is a one-off, remote assessment of a single tenant using an account with read-only permissions. We check privileged roles, consents granted to applications, devices in Entra ID, MFA methods, mailbox rules, Conditional Access and traces of token replay after MFA, and we compare the configuration with the CISA SCuBA secure configuration baselines and Microsoft playbooks.
The result is a report with remediation priorities and a presentation for your team. The review does not cover on-premises Active Directory and does not promise any particular Secure Score. You will find more articles on protecting cloud environments in the cloud security knowledge base. If you would like to agree the scope of the review, contact us.
What else is worth knowing about Microsoft Secure Score?
Does a high Secure Score mean your company is secure?
No. Microsoft states that the score is not an absolute measure of how likely a breach is, nor a guarantee of protection. It shows the extent to which you use the recommended security controls.
Does Secure Score depend on the licences you own?
The list of recommendations does not: Microsoft shows the full set for a product regardless of licence. The “Current license score” view shows what you can achieve with your current licences.
How long does it take for the score to reflect a deployed recommendation?
According to Microsoft, the change may only show up in the score after a day or two. Teams and Entra recommendations also follow their own refresh cycle.
Who can change the status of recommendations?
Write access comes with roles such as Security Administrator, Exchange Administrator and SharePoint Administrator, while Global Reader and Security Reader are read-only. You can also use Microsoft Defender Unified RBAC custom roles with the Exposure Management (manage) permission.
Is the “Resolved through third party” status safe?
It gives full points, but Microsoft then has no visibility into whether the recommendation is really implemented. Record the tool, scope and date of verification in the notes.
