Skip to main content

Preparing your data

There are two ways to obtain data for the analysis:

  1. Export a Person Query as a CSV from Viva Insights — recommended.
  2. Export a CSV from an existing Super Users Report, if you already have one.

Which method? We generally recommend Method 1 (Person Query) — it guarantees comprehensive coverage of the covariates needed for a result you can be confident in. Method 2 is quicker if you already have a Super Users Report, but some metrics may be missing.

Whichever you choose, save the resulting CSV into the toolkit’s data/ folder.

Your data stays local The analysis runs entirely on your machine — no data is uploaded anywhere. The data/ folder is covered by a .gitignore rule (*.csv), so your exports won’t be committed if you’re working inside a Git clone. Keep filenames ending in .csv (lower-case) so that rule reliably applies on every operating system.

Method 1 — Export a Person Query from Viva Insights

  1. Open the Viva Insights analysis page.
  2. Select Person Query → ‘Set up analysis’.
  3. Configure:
    • Time period: Last 6 months (rolling)
    • Group by: Week
    • Metrics: Include the columns listed under Columns to include below for your scenario.
    • Filter: Is Active = True (if available) — you can validate the number of employees here.
    • Attributes: Include Organization and Function Type (others optional) — this is the last box on the page.
  4. Save and Run the query. Wait until Status = Completed, then export the CSV into copilot-causal-toolkit/data/.

Employee Engagement scenario The Person Query must include Glint survey data as the outcome. This needs a prior setup step to import survey data into Viva Insights — see Import survey data from Viva Glint. Once imported, metrics such as eSat appear as columns in the export.

Method 2 — Export from a Super Users Report

Show Super Users Report export steps

This assumes you already have a Super Users Report (.pbit/.pbix) populated with Viva Insights data. It contains pre-aggregated data that can be exported without setting up a new Person Query.

From Power BI Desktop:

  1. Open your Super Users Report. If you don’t have the file, ask your Viva Insights admin or check your organization’s shared workspace.
  2. Open the Table view — click the Table icon on the left sidebar.
  3. Find the data table — look for person-level data with columns like PersonId, Date, Total_Copilot_actions_taken, collaboration metrics, and organizational attributes. It’s usually called Table, but may have been renamed.
  4. Export to CSV:
    • Option A: Right-click the table name → Copy table → paste into Excel → save as CSV.
    • Option B: Select the table → Home → Transform data to open Power Query Editor → right-click the table → Export → CSV.
  5. Verify the export contains multiple weeks (ideally 12–26), all required columns, and no excessive blank rows.
  6. Save the CSV into copilot-causal-toolkit/data/ with a descriptive name (e.g. SuperUsersReport_Export_2025.csv).

Important notes for Super Users Report data:

  • The report uses Date instead of MetricDate as the date column.
  • The SUR notebooks (ending in _SUR.ipynb) are designed to handle this schema difference.
  • Some Person Query metrics may be unavailable in the report (e.g. Available_to_focus_hours, Weekend_collaboration_hours). If critical metrics are missing, use Method 1 instead.

Alternative: export from Power BI Service (online). Open the report at powerbi.com → navigate to the page with the data table → click … (More options) on a visual → Export data → Underlying data → .csv. Some organizations restrict export; contact your Power BI admin if needed.

Columns to include

The treatment variable is always Total_Copilot_actions_taken. Select the scenario you’re running for its outcome and recommended confounders. Organizational attributes (e.g. Organization, Function, Level, IsManager, Area) are used for heterogeneity analysis in every scenario — include as many as you can; exact names vary by organization, so update them in the notebook configuration.

Goal: understand how Copilot usage changes time spent collaborating with external stakeholders and customers.

  • Outcome: External_collaboration_hours — hours in meetings, emails, chats, and calls with people outside the organization.
  • Treatment: Total_Copilot_actions_taken.
  • Confounders (time-varying controls):
    • Meeting_hours, Email_hours, Chat_hours, Collaboration_hours
    • Internal_network_size, Networking_outside_organization
    • Total_focus_hours
    • Other relevant behavioral and network metrics from your Person Query.

    </div>

    Goal: understand how Copilot usage changes after-hours work patterns, which affect wellbeing and burnout risk.

  • Outcome: After_hours_collaboration_hours — work-related activity outside standard business hours.
  • Treatment: Total_Copilot_actions_taken.
  • Confounders (time-varying controls):
    • Meeting_hours, Email_hours, Chat_hours, Collaboration_hours
    • Internal_network_size, Networking_outside_organization
    • Total_focus_hours, Workweek_span
    • Other relevant behavioral and network metrics from your Person Query.

    </div>

    Goal: understand how Copilot usage influences employee engagement, measured by an ordinal survey outcome (e.g. a Glint metric). Unlike the other scenarios, the outcome is survey-based rather than a continuous Viva Insights metric.

Template notebook Glint metrics vary across organizations and may be custom-defined. Review and update the outcome variable, its scale, and the confounders to match your data before running.

  • Outcome: a Glint metric such as eSat (Employee Satisfaction) — set this to whichever ordinal survey outcome you intend to evaluate.
  • Treatment: Total_Copilot_actions_taken.
  • Outcome scale configuration: because the outcome is ordinal, set OUTCOME_SCALE_MIN / OUTCOME_SCALE_MAX to match the survey scale (e.g. 1–5, 1–7, 1–10). These drive ceiling/floor diagnostics and interpretation.
  • Confounders (starting point — revise per outcome):
    • Collaboration_hours, Available_to_focus_hours, Active_connected_hours, Uninterrupted_hours
    • After_hours_collaboration_hours, Collaboration_span
    • Meeting_and_call_hours_with_manager_1_1
    • Other relevant behavioral and network metrics from your Person Query.

Data requirement: the Person Query must include Glint survey data — see Import survey data from Viva Glint. Only the Person Query (PQ) notebook exists for this scenario; there is no SUR version.

Explore which HR / organizational attributes are in your dataset

Run this to list every HR attribute and its value counts:

hrvar_str = vi.extract_hr(data, return_type = 'vars').columns

for hr_var in hrvar_str:
    hrvar_table = vi.hrvar_count(data, hrvar = hr_var, return_type = 'table')
    print(f"\nValue counts for {hr_var}:")
    print(hrvar_table)

for hr_var in hrvar_str:
    vi.hrvar_count(data = data, hrvar = hr_var, return_type = 'plot')
Last updated: Jun 22, 2026 Edit this page on GitHub