# Codebook Synthetic educational data for a fictional UK business-services company; no real employees or staff-number linkage are included. One row is one sampled employee; team-level attributes repeat for sampled members of the same team. CSV blank cells denote missing values. Dates are ISO 8601. Numeric categories are codes, not continuous measurements unless stated. The approved data type, dataset level, measured entity, timing and collection window below describe the variables in the protocol; the physical CSV always has one employee per row. ## `employee_id` - Role: identifier - Additional analysis role: identifier - Data type: text - Dataset level: Employee - Measured entity: Employee - Timing: Before randomization - Collection window / frequency: Once, 18 January 2027 - Collection mode: direct - Source: Generated by the study team at sampling - Values: Text; E0001–E1000, unique - Definition: Study-generated ID before randomization; no real staff-number key exists in this synthetic bundle. - Missingness: Never missing. ## `department` - Role: baseline_covariate - Additional analysis role: baseline covariate / stratum - Data type: categorical - Dataset level: Employee - Measured entity: Employee - Timing: Before randomization - Collection window / frequency: Once, 4 January 2027 roster extract - Collection mode: direct - Source: HR roster extract - Values: Sales; Customer Service; Engineering; Operations; Finance; Marketing; HR & Administration - Definition: HR roster, 4 January 2027. - Missingness: Never missing. ## `team_id` - Role: cluster_group - Additional analysis role: team cluster - Data type: categorical - Dataset level: Team - Measured entity: Team - Timing: Before randomization - Collection window / frequency: Once, 4 January 2027 roster extract - Collection mode: direct - Source: HR roster extract - Values: Text; T001–T420 - Definition: Anonymous line-manager team from HR; members share team and site. - Missingness: Never missing. ## `site` - Role: baseline_covariate - Additional analysis role: baseline covariate - Data type: categorical - Dataset level: Employee - Measured entity: Employee - Timing: Before randomization - Collection window / frequency: Once, 4 January 2027 roster extract - Collection mode: direct - Source: HR roster extract - Values: London; Manchester; Birmingham; Glasgow - Definition: Attached office site from HR; remote workers retain a site. - Missingness: Never missing. ## `work_arrangement` - Role: moderator - Additional analysis role: baseline moderator / stratum - Data type: categorical - Dataset level: Employee - Measured entity: Employee - Timing: Before randomization - Collection window / frequency: Once, 4 January 2027 roster extract - Collection mode: direct - Source: HR roster extract - Values: Office-based; Hybrid; Remote - Definition: Contracted arrangement from HR. - Missingness: Never missing. ## `job_level` - Role: baseline_covariate - Additional analysis role: baseline covariate / moderator - Data type: ordinal - Dataset level: Employee - Measured entity: Employee - Timing: Before randomization - Collection window / frequency: Once, 4 January 2027 roster extract - Collection mode: direct - Source: HR roster extract - Values: 1 Associate; 2 Senior associate; 3 Team lead; 4 Manager - Definition: HR job grade. - Missingness: Never missing. ## `tenure_years` - Role: baseline_covariate - Additional analysis role: baseline covariate - Data type: continuous - Dataset level: Employee - Measured entity: Employee - Timing: Before randomization - Collection window / frequency: Once, 4 January 2027 HR extract - Collection mode: derived - Source: HR record of employment start date - Values: 0.25–40 years; one decimal - Definition: Service on 4 January 2027, employment-start-date interval divided by 365.25 and rounded. - Missingness: Never missing. ## `experience_years` - Role: baseline_covariate - Additional analysis role: baseline covariate - Data type: continuous - Dataset level: Employee - Measured entity: Employee - Timing: Before randomization - Collection window / frequency: Once, 4 January 2027 HR extract - Collection mode: direct - Source: HR record (self-reported at hiring and updated at annual reviews) - Values: 0–50 years; one decimal - Definition: HR-reported paid work experience; normally at least tenure. Outdated records can be below tenure. - Missingness: Blank when HR record incomplete. ## `age` - Role: baseline_covariate - Additional analysis role: baseline covariate - Data type: continuous - Dataset level: Employee - Measured entity: Employee - Timing: Before randomization - Collection window / frequency: Once, 4 January 2027 HR extract - Collection mode: derived - Source: HR record (date of birth converted to age; the date of birth itself is not kept) - Values: 18–67 completed years - Definition: Age at 4 January 2027 from HR date of birth; date of birth is not included. - Missingness: Never missing. ## `gender` - Role: baseline_covariate - Additional analysis role: baseline covariate - Data type: categorical - Dataset level: Employee - Measured entity: Employee - Timing: Before randomization - Collection window / frequency: Once, 4 January 2027 HR extract - Collection mode: direct - Source: HR equality-monitoring record - Values: Woman; Man; Non-binary or self-described; Prefer not to say - Definition: HR equality-monitoring response; Prefer not to say is a category, not missing. - Missingness: Never missing. ## `education` - Role: baseline_covariate - Additional analysis role: baseline covariate - Data type: ordinal - Dataset level: Employee - Measured entity: Employee - Timing: Before randomization - Collection window / frequency: Once, 4 January 2027 HR extract - Collection mode: direct - Source: HR record (self-reported) - Values: 1 Secondary school; 2 Vocational or apprenticeship; 3 Undergraduate degree; 4 Postgraduate degree - Definition: Highest reported completed qualification in HR. - Missingness: Blank when unreported. ## `caring_responsibilities` - Role: baseline_covariate - Additional analysis role: baseline covariate - Data type: binary - Dataset level: Employee - Measured entity: Employee - Timing: Before randomization - Collection window / frequency: Once, baseline survey (4 to 15 January 2027) - Collection mode: direct - Source: Baseline staff survey - Values: 0 No; 1 Yes - Definition: Regular caring responsibilities for a child or adult; baseline survey 4–15 January. - Missingness: Blank if baseline survey unanswered. ## `commute_minutes` - Role: baseline_covariate - Additional analysis role: baseline covariate - Data type: continuous - Dataset level: Employee - Measured entity: Employee - Timing: Before randomization - Collection window / frequency: Once, baseline survey (4 to 15 January 2027) - Collection mode: direct - Source: Baseline staff survey - Values: 0–150 minutes, multiples of 5 - Definition: Self-reported usual one-way commute on office days; 0 for Remote respondents. - Missingness: Blank if baseline survey unanswered. ## `full_time` - Role: baseline_covariate - Additional analysis role: baseline covariate - Data type: binary - Dataset level: Employee - Measured entity: Employee - Timing: Before randomization - Collection window / frequency: Once, 4 January 2027 roster extract - Collection mode: direct - Source: HR roster extract - Values: 0 Part time; 1 Full time - Definition: Contracted FTE status from HR. - Missingness: Never missing. ## `annual_salary_gbp` - Role: baseline_covariate - Additional analysis role: baseline covariate - Data type: continuous - Dataset level: Employee - Measured entity: Employee - Timing: Before randomization - Collection window / frequency: Once, January 2027 payroll - Collection mode: direct - Source: Payroll record - Values: Whole pounds, ordinarily 18000–150000; possible erroneous values below 1000 - Definition: Full-time-equivalent base annual payroll salary; some values are mistakenly entered in thousands and require cleaning. - Missingness: Never missing. ## `team_size` - Role: baseline_covariate - Additional analysis role: team covariate - Data type: count - Dataset level: Team - Measured entity: Team - Timing: Before randomization - Collection window / frequency: Once, 4 January 2027 roster extract - Collection mode: direct - Source: HR roster extract - Values: 2–25 people - Definition: Roster number reporting to the manager, including the employee; not just study-sampled teammates. - Missingness: Never missing. ## `prior_performance_rating` - Role: baseline_covariate - Additional analysis role: baseline covariate - Data type: ordinal - Dataset level: Employee - Measured entity: Employee - Timing: Before randomization - Collection window / frequency: Once, from the 2026 review cycle - Collection mode: direct - Source: Performance management system - Values: 1 Below expectations; 2 Partly meets; 3 Meets; 4 Exceeds; 5 Outstanding - Definition: 2026 annual review rating. - Missingness: Blank for employees hired after 1 April 2026 without an annual review. ## `sick_days_prior_12m` - Role: baseline_covariate - Additional analysis role: baseline covariate - Data type: count - Dataset level: Employee - Measured entity: Employee - Timing: Before randomization - Collection window / frequency: Once, covering January to December 2026 - Collection mode: derived - Source: HR absence system - Values: 0–60 whole days - Definition: Recorded sickness in 2026, over service period if less than 12 months. - Missingness: Never missing. ## `baseline_cross_team_contacts` - Role: baseline_covariate - Additional analysis role: baseline covariate - Data type: count - Dataset level: Employee - Measured entity: Employee - Timing: Before randomization - Collection window / frequency: Counted once over the 8-week baseline window (26 October to 18 December 2026) - Collection mode: derived - Source: Aggregated collaboration-platform metadata (counts only; no message content) - Values: 0–200 distinct colleagues - Definition: Distinct colleagues in other departments in emails, chats or meetings across 26 October–18 December 2026; metadata counts only. - Missingness: Never missing. ## `baseline_output_index` - Role: baseline_covariate - Additional analysis role: baseline covariate - Data type: continuous - Dataset level: Employee - Measured entity: Employee - Timing: Before randomization - Collection window / frequency: Weekly for 8 weeks (26 October to 18 December 2026); averaged - Collection mode: derived - Source: Department operational systems: Sales CRM (value of closed orders), Customer Service ticketing (tickets resolved to quality standard), Engineering tracker (accepted story points), Operations (orders processed), Finance (reconciliations and reports completed), Marketing (approved deliverables), HR & Administration (cases closed) - Values: 0–400 index points; one decimal - Definition: Mean weekly department output, FTE-adjusted and divided by the fixed department median of baseline employee-weeks, times 100; averaged over baseline weeks worked. The synthetic reference uses sampled worked employee-weeks as a proxy for the company-wide department median. - Missingness: Blank if fewer than two baseline weeks worked. ## `job_satisfaction_baseline` - Role: baseline_covariate - Additional analysis role: baseline covariate - Data type: continuous - Dataset level: Employee - Measured entity: Employee - Timing: Before randomization - Collection window / frequency: Once, baseline survey (4 to 15 January 2027) - Collection mode: derived - Source: Baseline staff survey - Values: 1–7 in increments of 0.2; higher is more satisfied - Definition: Baseline survey: mean of five 1–7 items, only when all are answered. - Missingness: Blank if baseline survey unanswered. ## `belonging_baseline` - Role: moderator - Additional analysis role: baseline moderator - Data type: continuous - Dataset level: Employee - Measured entity: Employee - Timing: Before randomization - Collection window / frequency: Once, baseline survey (4 to 15 January 2027) - Collection mode: derived - Source: Baseline staff survey - Values: 1–7 in increments of 0.25; higher is stronger belonging - Definition: Baseline survey: mean of four 1–7 items, only when all are answered. - Missingness: Blank if baseline survey unanswered. ## `retreat_arm` - Role: treatment_assignment - Additional analysis role: randomized assignment - Data type: categorical - Dataset level: Employee - Measured entity: Employee - Timing: At randomization - Collection window / frequency: Once, 18 January 2027 - Collection mode: direct - Source: Randomization list - Values: Retreat; Control - Definition: Employee-level assignment within department × work arrangement strata, 18 January 2027. - Missingness: Never missing. ## `session_date` - Role: time - Additional analysis role: reference date - Data type: date_time - Dataset level: Employee - Measured entity: Employee - Timing: At randomization - Collection window / frequency: Once, 18 January 2027 - Collection mode: direct - Source: Randomization list - Values: YYYY-MM-DD; 2027-02-02, 2027-02-04, 2027-02-09, 2027-02-11, 2027-02-16, 2027-02-18, 2027-02-23, 2027-02-25, 2027-03-02, or 2027-03-04 - Definition: Assigned week-0 date for either arm, regardless of attendance; follow-up weeks 1–16 run after this date. - Missingness: Never missing. ## `attended_retreat` - Role: compliance_measure - Additional analysis role: compliance measure - Data type: binary - Dataset level: Employee - Measured entity: Employee - Timing: Week 0 (reference date) - Collection window / frequency: Once - Collection mode: direct - Source: Venue sign-in sheet, entered by HR - Values: 0 Did not attend; 1 Attended - Definition: Venue sign-in on assigned date. Control employees cannot attend during the study. - Missingness: Never missing. ## `cross_team_contacts_w1_8` - Role: mediator - Additional analysis role: post-treatment mediator - Data type: count - Dataset level: Employee - Measured entity: Employee - Timing: After treatment, weeks 1 to 8 - Collection window / frequency: Counted once over the 8-week window - Collection mode: derived - Source: Aggregated collaboration-platform metadata (counts only; no message content) - Values: 0–200 distinct colleagues - Definition: Distinct colleagues in other departments contacted in weeks 1–8 after the reference date. - Missingness: Blank if employee left before the end of week 8. ## `output_index_w1_8` - Role: outcome - Additional analysis role: primary outcome - Data type: continuous - Dataset level: Employee - Measured entity: Employee - Timing: After treatment, weeks 1 to 8 - Collection window / frequency: Weekly for 8 weeks; averaged - Collection mode: derived - Source: Department operational systems (the same metrics as baseline_output_index) - Values: 0–400 index points; one decimal - Definition: Mean of weekly output indices for weeks worked in weeks 1–8, using the same fixed department baseline reference. - Missingness: Blank if fewer than two weeks worked in this window. ## `output_index_w9_16` - Role: outcome - Additional analysis role: secondary outcome - Data type: continuous - Dataset level: Employee - Measured entity: Employee - Timing: After treatment, weeks 9 to 16 - Collection window / frequency: Weekly for 8 weeks; averaged - Collection mode: derived - Source: Department operational systems (the same metrics as baseline_output_index) - Values: 0–400 index points; one decimal - Definition: Mean of weekly output indices for weeks worked in weeks 9–16, using the same fixed department baseline reference. - Missingness: Blank if fewer than two weeks worked in this window. ## `job_satisfaction_w4` - Role: outcome - Additional analysis role: secondary outcome - Data type: continuous - Dataset level: Employee - Measured entity: Employee - Timing: After treatment, week 4 - Collection window / frequency: Once; survey open for 7 days - Collection mode: derived - Source: Follow-up staff survey 1 - Values: 1–7 in increments of 0.2 - Definition: Mean of five 1–7 satisfaction items in the week-4 survey; survey open seven days. - Missingness: Blank for nonresponse or if already left. ## `job_satisfaction_w12` - Role: outcome - Additional analysis role: secondary outcome - Data type: continuous - Dataset level: Employee - Measured entity: Employee - Timing: After treatment, week 12 - Collection window / frequency: Once; survey open for 7 days - Collection mode: derived - Source: Follow-up staff survey 2 - Values: 1–7 in increments of 0.2 - Definition: Mean of the same five satisfaction items in the week-12 survey; survey open seven days. - Missingness: Blank for nonresponse or if already left. ## `left_firm_w16` - Role: outcome - Additional analysis role: secondary outcome - Data type: binary - Dataset level: Employee - Measured entity: Employee - Timing: After treatment, by week 16 - Collection window / frequency: Once, at the end of week 16 - Collection mode: direct - Source: HR leaver records - Values: 0 Still employed; 1 Left - Definition: Any exit between the assigned reference date and end of week 16, from HR leaver records. - Missingness: Never missing. ## Analysis and limitations Use assignment for intention-to-treat analyses, including non-attenders. Account for team clusters and randomization strata. Session dates also define calendar timing for controls. Follow-up windows are relative to each employee’s date, not a fixed calendar date. Observed scores are aggregate measures: survey items, individual weekly outputs, number of weeks worked, and exact exit dates are not in this file. Consequently, the proposed weeks-worked and zero-for-nonworked-weeks checks, and exact pre-week-8 exit timing, cannot be reconstructed from these columns alone. Contact counts concern other departments, despite the variable name. Statistical associations, including mediator relationships, do not establish causation.