Repository navigation
Portal Workload Reports
Web Portal → Workload reports
Use this guide to choose the right Insights page for a question, then check that its evidence is available before exporting or making a decision. The portal groups its pages by workload: Copilot, Agents, Apps & licences, SharePoint & OneDrive and Teams; the same underlying page can be reached from a workload or a direct route.
Screenshots elsewhere in these guides use synthetic data. They illustrate controls and layouts, not import completion, consent or external-service acceptance.
| Question | Released route | Source or next guide |
|---|---|---|
| How much evidence has arrived? | /insights/overview |
Counts/newest timestamps; investigate Service health separately |
| Which reports are available? | /insights/reports |
Report landing page; each workload requires its own enabled import |
| How is Copilot used? | /insights/reports/copilot |
Audit-based Copilot reports, not billing |
| What derived prompt statistics/categories exist? | /insights/reports/copilot-prompts |
History can make this reachable without audit; Prompt Categories still requires separate opt-in |
| Which agents ran? | /insights/reports/copilot-agents |
Audit agent evidence, counted turns and maker-testing exclusions; Copilot Agent Evidence |
| What do workload usage reports say? | /insights/reports/usage |
Imported Microsoft 365 usage reports, GraphUsageReports
|
| Which Office apps/platforms are used? | /insights/reports/office-apps |
Imported usage reports; not per-user Teams app-install tracking, which was removed |
| What file activity was audited? | /insights/reports/spo-audit |
SharePoint audit import (ActivityLog), not browser page views |
| What web traffic was tracked? | /insights/reports/web-traffic |
AITracker and AppInsightsImporter; use Web activity for exploration |
| What call records or sent-email evidence exists? |
/insights/reports/calls, /insights/reports/emails
|
Independent Calls and SentEmails opt-ins, permissions and privacy policies |
| Where should Copilot seats/enablement go? | /insights/copilot-adoption |
Copilot Adoption User Guide |
| What does Copilot cost, and what does the evidence of use support? |
/insights/copilot-investment (since Stable 1857) |
Copilot investment; modelled money from saved assumptions, never an invoice or a measured saving |
| Who consumed Copilot credits, from which source? |
/insights/copilot-spend (since Stable 1857) |
Copilot spend; billed Copilot Studio credits and preview consumed-credit readings, kept apart, with no prices or assumptions |
| Which licence holders have observed activity? | /insights/licence-activity |
Licence Activity; its dates select activity, not historical seat membership |
| How has weekly activity changed, and in which departments? | /insights/activity-analysis |
Activity analysis; weeks compiled by the profiling runbooks, not a live query |
| What was billed? | /insights/agent-costs |
Agent Costs; delegated consumption and optional resource-scoped Azure costs |
| Which policies/signals accompanied Copilot use? | /insights/dlp |
Copilot DLP and Governance; embedded and tenant-wide feeds stay separate |
Copilot investment and Copilot spend were added in Stable 1857 (PR #787).
Standard report routes accept the page's 1, 3 or 6 month selection (?months=1, 3 or 6). That selector is not the same as Copilot Adoption's rolling/past range, an explorer's day window, or Activity analysis's compiled weeks. Record the actual dates/source and filter, not just “last quarter”.
Since Stable 1859 (PR #816): on
/insights/reports/usage, Weekly active users by workload counts everyone active on any day of each week. Microsoft's daily usage reports list only the people active that day, and builds before 1859 count only the people active on the last imported day of each week, so expect much higher figures after upgrading, past weeks included. A week is charted once its last day is at least three days old (UTC). A week missing a day of reports is drawn from the days it has and flagged as incomplete: read its value as a floor. A week with no reports at all is a gap, not a zero. On a small tenant, a quiet workload such as Viva Engage at a weekend can flag a fully imported week as incomplete, because a day when nobody was active looks the same as a missed import.
Since Stable build 1860 (PR #866, PR #879): for readers without See PII, Copilot agents shows only agents that 5 or more people used in the period, and searching by name follows the same rule; Office apps leaves departments and email domains of fewer than 5 people out of its two matrices; and Prompt insights hides prompt phrases and languages used by fewer than 5 people. See Portal permissions → Small groups withheld from readers without See PII.
Open Teams explorer (/insights/teams).
- Read the source-availability panel. Teams metadata, usage reports, meetings/calls and deep conversation analysis are separate inputs. A configured permission is not proof of a completed import.
- Choose 7, 28, 90, 180 or 365 days. Start with Overview, then use Adoption, Meetings, Collaboration, Conversations or People for the question at hand. Adoption can group by organisational dimensions such as department.
- Keep the source/coverage explanation beside the result. A day with no imported call records does not prove no meetings occurred; a Team without delegated deep-analytics authorisation cannot supply channel-message evidence.
- Use People only when authorised with See PII. Aggregate access does not authorise named lists or person-level exports.
- Export the intended section/window and retain its method notes. Changing a period while a panel loads is not an additional independent observation.
Since Stable 1859 (PR #816): Dormant people, in the engagement mix and the Dormant users list, are enabled people who hold a licence that includes Teams but did nothing in Teams in the period. Teams Rooms and other shared-device licences don't count. Microsoft's daily usage reports list only the people active that day, so in builds before 1859 the Dormant list is usually empty and the Dormant share close to 0%. Dormant people are identified only when every day of the window has a Teams report imported, the usage import isn't limited to selected groups, the reports don't conceal names, and someone in scope holds a Teams licence. Otherwise the page says why it can't identify them, and leaves licensed people with no activity out of the mix. Graph keeps only about 28 days of reports, so the 90-, 180- and 365-day windows can identify dormant people only once the import has run that long. Reach and meetings per active user stop counting people who have since been disabled.
Since Stable 1859 (PR #805): for readers without See PII, Adoption combines every group with fewer than 5 people, or fewer than 5 active people, into one smaller groups, combined row, on screen and in its export. Power users per department combines departments with fewer than 5 power users in the same way.
Since Stable build 1860 (PR #866, PR #880): readers without See PII see more withheld, on screen and in the exports. Overview and Adoption count only people whose account is enabled, and show a dash for any week, segment, lifecycle count, platform or headline figure about one to four people, so their totals can differ slightly from a See PII reader's; the tab says so. Meetings leaves blank a week that fewer than 5 people attended, together with the quietest other week. Collaboration and Conversations leave out teams of fewer than 5 members and their channels, together with the smallest other team, and hide the apps-and-tabs list, key phrases and languages; sentiment is shown only for teams of 5 or more members. A team's size is its membership as imported in the last seven days: if the Teams import hasn't refreshed a team in that time, its channels and sentiment stay hidden until the import catches up. See Portal permissions → Small groups withheld from readers without See PII.
For missing conversations, check Teams permissions, delegated consent and the enabled imports before changing SQL or clearing checkpoints. See Teams and calls troubleshooting.
Open Web activity (/insights/web-activity). This explores browser tracking, not the SharePoint audit feed.
- Confirm the intended sites have AITracker activated, the tracker can reach Application Insights, and AppInsightsImporter has completed. Use web tracking setup and Verify.
- Select 7, 28, 90, 180 or 365 days and read Overview before drilling into Visits, Pages, Journeys, Geography, Search or Technology.
- Treat page views, sessions and tracked time as their own measurements. They are not file edits, proof that content was read, or an employee-performance measure. AITracker's previous-page time updates and metadata caching have documented boundaries.
- Keep browser/network/cookie/privacy restrictions in mind when comparing cohorts. Missing client telemetry is not equivalent to no SharePoint use.
- Export only the permitted scope. Check the administrator's global filter and local period before sharing a chart or workbook.
Since Stable 1859 (PR #817): page load times on Web activity, in the SQL views and in the Power BI reports are stored in real seconds. Builds before 1859 store a tenth of the real time, so load times jump about tenfold for page views imported after the upgrade: that jump is the fix, not a slowdown. Older page views keep the old values until a DBA runs the one-off repair: see Repairing page load times stored by older builds.
Since Stable build 1860 (PR #876, PR #881): for readers without See PII, search terms and dead-end search terms are hidden, and every list leaves out rows of fewer than 5 people: sites, countries, regions and cities, browsers, operating systems and devices, the page lists, entry, exit and bounce pages, page-to-page steps, start-to-end flows, clicked elements and visits by page. The page tables no longer show unique page views, entries, exits or bounce rate: a page's visits are on the Visits tab, and its entries and exits on Journeys. The quiet-pages list is always empty, with a note saying why. CSV exports carry the same rows as the screen, and the note of how many rows were left out counts over the whole list, up to 500 rows. See Portal permissions → Small groups withheld from readers without See PII.
If the audit report is populated but web activity is empty, diagnose the tracker/Application Insights path rather than adding audit API permissions. Conversely, page tracking does not fill the audit report.
Open Activity analysis (/insights/activity-analysis) for the compiled week-by-week view. The full guide, including definitions and who sees what, is Activity analysis.
- Wait for the availability check. Available metrics come from the compiled profiling data, not a fresh query to Microsoft services.
- Choose the period and metrics. Filters - people, licences and activity ranges - need See PII; without it, the page covers everyone with recorded activity in the period. A printout records the period, metrics and filters with the figures.
- If no metrics or weeks are available, an administrator should inspect Administration → Profiling and the runbook schedules. A report reader should request that check rather than rebuilding data.
- A repaired audit record or updated SQL procedure does not rewrite already compiled weeks. Use Rebuilding Profiling Data only as an approved maintenance operation, then refresh the report.
- Match periods, definitions, filters and retention before comparing its figures with other reports. Physical audit rows, counted Copilot turns and compiled weekly measures need not be numerically identical.
The administrator's Report Filter is server-enforced across Insights, on top of the reader's choices. Only Administration plus See PII may edit/bypass it; neither bypass nor See PII overrides small-group suppression. A source error, suppressed cohort, unknown evidence and genuine zero are different results.
Before using a report in a decision, confirm:
- the intended import is enabled and has completed, with the required consent;
- dates and scope match the question, including publication delays and retained history;
- warnings, caps, incomplete results and missing sources are disclosed;
- the export matches the current view and is handled as organisational/personal data where applicable;
- no causal, productivity, security-effectiveness or billing claim is inferred from a usage chart alone.
The adoption-specific Power BI version is deprecated in favour of the Copilot Adoption portal (#261). This decision does not deprecate these workload pages or unrelated Activity & Usage Analytics PBIT reports.
- Copilot analytics overview
- Copilot Adoption user guide
- Copilot Adoption setup (admins)
- Work Trend Index
- Copilot investment
- Copilot spend
- Agent costs
- Prompt categories
- Copilot credit history (preview)
- Agent 365 package catalog
- The Copilot data in depth
- Prerequisites
- App registrations setup
- Install with the installer
- Verify the deployment
- Security & permissions model
- Private endpoints (optional)
- Certificate authentication (optional)
- Manual installation
- Activity & usage analytics (Power BI)
- SharePoint page tracking