← All articles
Microsoft Sentinel · Technical field note

Finding underutilised Microsoft Sentinel log tables without weakening detection coverage

A defensible method for comparing log ingestion cost with analytics rule, hunting and investigation dependencies.

Finding underutilised Microsoft Sentinel log tables without weakening detection coverage

Security telemetry is expensive, but removing a table simply because it does not appear in an analytics rule can create blind spots. During a July 2025 investigation, I explored a workspace-wide inventory of Microsoft Sentinel tables and compared it against analytics rule queries. The goal was a defensible cost and coverage review.

Separate inventory, ingestion and detection dependencies

The initial exploration included the .show tables command, which is not supported as a Log Analytics workspace query; it should not be treated as a valid Sentinel inventory command. The Usage table also cannot establish a complete inventory of configured tables or the last time an analyst queried a table. These are three distinct questions.

  1. Which tables exist?
  2. Which tables incur ingestion volume?
  3. Which detection rules and other workloads depend on those tables?

Use the Log Analytics workspace Tables view or the Azure Resource Manager tables API to enumerate configured tables. Follow pagination and retain the plan and retention settings. Retrieve the Sentinel analytics rule inventory using the alertRules API; keep disabled rules separately and inspect each page.

Measure ingestion using KQL

kusto
Usage
| where TimeGenerated > ago(32d)
| where StartTime >= startofday(ago(30d))
    and EndTime <= startofday(now())
| where IsBillable == true
| summarize BillableGB = sum(Quantity) / 1000.0,
            LastBillableUsageBucket = max(EndTime) by DataType
| order by BillableGB desc

This reports volume, not invoice totals or analyst usage. Join the result to the full table inventory so that tables with no ingestion remain visible.

Compare the detection dependencies

Match known table identifiers against analytics rule query text, recording enabled and disabled rules separately. This is a first-pass heuristic, not a KQL parser. Stored functions, wildcard unions, cross-workspace queries, string literals and comments can cause incorrect results. Investigate dependencies before making decisions.

If query auditing was already enabled and retained, LAQueryLogs may give additional insight:

kusto
LAQueryLogs
| where TimeGenerated > ago(30d)
| where QueryText has "SigninLogs"
| summarize LastMatchingAuditedQuery = max(TimeGenerated),
            MatchingQueries = count()

The result reflects matching audited queries only. Certain Sentinel query experiences are excluded from this audit stream. No result is not evidence that a table is unused.

Remediation decision register

Use columns for table name, retention, plan, ingestion volume, direct enabled-rule references, indirect consumers, compliance owner, proposed change and validation result.

Before changing collection, assess workbooks, hunting, functions, playbooks, incident response, retention requirements and cross-workspace consumers. Obtain data-owner approval, test coverage and establish a rollback method.

Conclusion

A missing direct rule reference is a reason for investigation, not a reason to delete security telemetry.

References