1. Source Files
| Type | Files (11 per type) | Periods covered | Rows (May 2026) |
|---|
| Users |
Omada_A_S_Enterprise_users_export__YYYY-MM-DD_-_YYYY-MM-DD_.csv |
Jul 2025 – May 2026 (one per month) |
176 rows (150 enabled, 26 deleted) |
| GPTs |
Omada_A_S_Enterprise_gpts_export__YYYY-MM-DD_-_YYYY-MM-DD_.csv |
Jul 2025 – May 2026 (one per month) |
36 rows (all live) |
| Projects |
Omada_A_S_Enterprise_projects_export__YYYY-MM-DD_-_YYYY-MM-DD_.csv |
Jul 2025 – May 2026 (one per month; Jul 2025 empty) |
72 rows (all live/active) |
| Impact Survey |
Omada_A_S_Enterprise_impact_survey_export_20260512.csv |
Through April 2026 (single export) |
18 rows (3 respondents × 6 questions) |
| Division Mapping |
userdepdiv.csv |
Static lookup — Department → Division |
N/A |
2. Metric Definitions
Enabled Users
Users with user_status == "enabled" in the monthly export. Deleted accounts (user_status == "deleted") are excluded from all user-level metrics. In May 2026: 150 enabled, 26 deleted.
Active Users
Users with is_active == 1 in the monthly export. This field is set by OpenAI based on whether the user sent at least one message during the export period. Includes both enabled and deleted accounts in the raw export; active rate is reported against enabled users only.
Total Messages
Sum of the messages field across all enabled users. Null values are treated as 0. This count includes all message types (direct, GPT, project, tool).
Median Messages
The 50th percentile of the messages distribution across all enabled users (including those with 0 messages). Half of enabled users are above and half below this value.
P90 Messages
The 90th percentile of the messages distribution. 10% of enabled users sent this many messages or more. Used as the "power user proximity" threshold in the Workspace report.
User Segments
Applied to enabled users only, based on the messages field for the relevant month:
— Not Using: messages == 0
— Regular: 1 ≤ messages ≤ 499
— Power: messages ≥ 500
These thresholds are fixed and applied consistently across all months.
Activation Rate
Active users ÷ Enabled users × 100. Represents the percentage of licenced users who engaged with the platform at least once in the reporting month.
GPT Adopted Signal
A GPT is flagged as "adopted" if unique_messagers_workspace ≥ 2. This indicates the GPT was used by more than one unique workspace user during the month. It is a usage-based proxy for shared adoption; the API does not return an explicit visibility/sharing field, so this remains an inference.
GPT Live / Draft
Derived from the config_type field in the GPT export. Values: live or draft. In May 2026, all 36 GPTs are live.
Project Active
Projects with is_active == 1. In May 2026, all 72 projects are active. Projects with messages_workspace == 0 would be flagged as dormant but none exist in this period.
Division
Derived by joining the user's department field (a JSON array) to the internal userdepdiv.csv lookup table. The first listed department is used for users with multiple departments. Users with no department or an unmapped department are classified as "Unassigned." The olku@omada.net division is manually overridden to "Customer Experience" (source data shows Customer Success, which is incorrect).
3. Exclusion Rules
Deleted accounts
Users with user_status == "deleted" are excluded from all user-level metrics (counts, averages, segments). They remain in the raw export and appear in the total row count.
x_ prefix and onmicrosoft.com accounts
Test or system accounts with an x_ prefix or @*.onmicrosoft.com email domains are excluded from all analytics and cost/licence reporting.
4. Known Limitations
Division coverage: 98.7% resolved (148 of 150 enabled users in May 2026). Division data is derived via name-based lookup against the Omada Identity directory (userdepdiv.csv), not from the raw department field in the ChatGPT export (which contains lowercase internal codes that do not map to Omada's actual division structure). Only 2 users remain unassigned: Sara Muryjas (blank division in Omada Identity, excluded by policy) and Sofus Andersen (not found in Omada Identity directory).
July 2025 active count exceeds enabled count. The July 2025 export shows 15 active users against 13 enabled. This is a known artefact of the data — likely because some users who were subsequently deleted or deprovisioned had activity recorded in the first pilot month. It does not reflect actual overage and has not been corrected in the trend data.
No GPT visibility/sharing field available. The GPT export does not include a field indicating whether a GPT is shared workspace-wide or kept private to its creator. Adopted vs single-user classification is inferred from unique user counts only.
Survey response count is too low for statistical analysis. With only 3 respondents, the impact survey data provides directional signals only. No averages, percentages, or comparisons from this data should be treated as representative of the user population.
Monthly exports are independent snapshots. Each monthly export reflects activity within that calendar month only. A user who was active in April but not May will appear in the "not using" segment for May regardless of historical activity. Trend data represents the state of each month as reported, not a continuous tracking of the same user cohort.
5. Reporting Cadence & Refresh Process
Cadence: Manual (ad hoc). These reports are refreshed manually when new monthly export files are available from the OpenAI Enterprise admin console. There is no automated data pipeline or scheduled refresh.
Monthly refresh inputs required
To refresh this report package for a new month, the following exports are needed from the OpenAI admin console:
— Monthly Users export (one file, covering the target month)
— Monthly GPTs export (one file, covering the target month)
— Monthly Projects export (one file, covering the target month)
— Impact Survey export (if updated)
The userdepdiv.csv division mapping only needs updating if department names have changed.
6. Report Package Contents
| File | Description |
|---|
| Exec_Summary.html | Executive summary with KPIs, assessment, recommendations, and charts. Primary audience: EMT / CFO. |
| Analytics_Index.html | Navigation index for the full report package. |
| Workspace_Analytics_Report.html | Detailed user metrics, division breakdown, model usage, and top user table. |
| User_Segments_Report.html | Power / Regular / Not Using segment breakdown with full user tables. |
| Impact_Survey_Report.html | Survey response analysis and productivity signal interpretation. |
| Projects_Report.html | Project activity, top projects, creator analysis, governance guidance. |
| GPTs_Report.html | GPT usage, adoption classification, full GPT listing, creator analysis. |
| Trend_Report.html | 11-month trend data with charts and full monthly data table. |
| Data_Appendix.html | This file — source files, definitions, exclusions, limitations, assumptions. |
7. Assumptions
Segment thresholds are organisationally defined. The Power User threshold of 500+ messages is set at the organisational level, not derived from statistical analysis of the distribution. It is applied consistently across all months and all reports.
May 2026 is treated as the primary reporting period. All static KPIs, tables, and segment counts in the Workspace, Segments, GPTs, and Projects reports reflect May 2026 data only. Trend data uses all 11 months (Jul 2025–May 2026).
Messages field represents total workspace messages. The messages field in user exports counts all messages sent by the user in the period, including GPT messages, project messages, and direct messages. It is not decomposed further in the user-level reports.