How TenantWard's shared numbers are made
TenantWard pools a small set of settings and events from the companies that share data, so everyone can see how they compare and which problems are spreading. This page explains what goes in, what never does, and how to opt out. It is published at tenantward.com/methodology and updated whenever a new kind of number is added.
Who is counted
- A company is one business TenantWard checks: a company account, or one client of an IT provider.
- Companies in accounts that turned off Settings › AI and data › Shared, anonymized data are left out completely, from the next nightly run.
- Sample data, closed accounts and TenantWard's own test accounts are always left out.
The rule of five
Every number covers at least 5 companies from at least 3 different accounts. A group smaller than that is not written down at all; there is no "fewer than 5" row to find. The database refuses to store a smaller group (a check on the table), so no report, screen or partner feed can show one.
Groups (cohorts) are: everyone, company size (by people with a license), and industry where the company set one. A group that is too small simply doesn't appear; the everyone group usually still covers it.
What is read
Only settings and events, never the contents of email, files, chats, documents or AI conversations:
| Source | What is used |
|---|---|
| Checks | Which check failed or passed, when it was first seen and when it was fixed. Never the issue's text or evidence. |
| One-click changes | Which kind of change ran, whether it was undone, and whether the issue stayed fixed. |
| Vulnerabilities (Microsoft Defender) | CVE numbers, how many devices, whether it is known to be exploited, its EPSS score; app names and device counts. Never device names, IDs or users. |
| Settings changed outside TenantWard | Which setting changed and whether the change was kept. Never who changed it. |
| Alerts | The kind of alert, whether someone acted on it, and whether it turned out real or harmless. |
| Scores | Scores over time after connecting. |
| Practice phishing emails | Which template was sent and whether each person clicked or finished the lesson. Never names or addresses. |
| Incidents | That a company had one, and when. Never what happened or who was involved. |
| The device agent | Whether disk encryption, the firewall and automatic updates are on, by operating system. Never hostnames, users or serial numbers. |
| Apps connected to Microsoft 365 or Google | App identifiers and names, only for apps found in 5 or more companies. |
The exact list of tables and columns is kept in the database (ops.intel_sources) and checked by a test on every change.
The numbers (rollups)
Computed every night at 03:10 UTC. History is kept so trends can be shown.
intel_issue_prevalence: for each check, the share of companies where it is failing now, the share fixed within 30 days, and the median days an open issue has been open.intel_fix_outcomes: for each one-click change, the share that were undone and the share where the issue stayed fixed.intel_cve_watch: for each CVE, how many companies and devices have it, whether it is exploited and its EPSS score; for each app, devices, CVEs and exploited CVEs.intel_drift: for each setting, how often it is changed outside TenantWard per company and the share of those changes that were kept.intel_alert_noise: for each kind of alert, the share acted on and the share that turned out real or harmless.intel_score_trajectory: the median score 30, 90 and 180 days after the first check, by starting score.intel_app_reputation: for each app, how many companies use it and the share put up for review or removed.intel_agent_fleet: the share of devices with encryption, a firewall and automatic updates on, by system.intel_benchmarks: the 25th, 50th and 75th percentile of current scores, by group.
intel_phish: for each practice-email template, the share of people who clicked in the last 180 days and the share of those who then finished the short lesson. Never who clicked. Practice emails sent through Google Workspace are left out.intel_incident_precursors: for each check, how much more often it was failing in the 30 days before a company had an incident than at companies without one (an odds ratio with a 95% interval, kept only when the interval is clearly above or below 1). Used only inside TenantWard, to rank which fixes matter most; never published.
Google Workspace
Numbers that come from Google Workspace connections are used only inside TenantWard, to make suggestions for companies like yours. They are never published or given to a partner.
Who sees them
- You, inside TenantWard: how you compare with companies like yours.
- TenantWard staff, to see what is spreading and what to build or fix first.
- Published reports, reviewed by a person before they go out.
- Partners such as insurers and brokers, only after they sign a data processing agreement, through a key that can be revoked, limited to the published numbers above, with every request logged. Partners never get a list of companies or anything about one company. A company's own report goes to someone else only when that company approves it.
Opting out
Settings › AI and data › Shared, anonymized data. When it is off, nothing from your account goes into any rollup from the next nightly run, and you still see warnings built from what others share.
Questions: hello@tenantward.com.
The report these numbers go into: State of small-business security.