Case 1 · Governed People Data

Can I trust Engineering Headcount?

Yes — this is the current published snapshot, not the APAC incident replay.

Engineering Headcount

15,954

Scope Engineering · Window month (as-of) · as_of 2026-08-31

  • Certified
  • Fresh
  • Healthy
As-of
2026-08-31
Owner
People Analytics
Certified run
Company 49,823

Scope Company · Window month (as-of) · as_of 2026-08-31

Company Headcount matches the parity table when serving_run.certified is true. SHA b28c00833782.

View historical incidents

Lineage

  1. Workforce file from HRISfrappe_hr.Employee
  2. Month-end worker historypeople_hist_worker_attr
  3. Workforce change eventspeople_evt_worker_change
  4. Certified Headcountpeople_get_metric_for(headcount)

Quality healthy · Publish published · health healthy

Definition

Certified workers as of month-end. Employment type and status follow BR-WF certified rules.

count of certified worker-month rows as-of

SQL: select count(*) from people_snap_worker_month where month_end = as_of and is_certified and (job_fam is null or job_family = job_fam)

Quality

74 tests · bronze 3 / silver 35 / gold 36 · all passed · last run against data-v1 (frozen snapshot)

bronze · 3

parse · 1

  • Pass · bronze_parse

schema_drift · 1

  • Pass · bronze_schema_drift

volume · 1

  • Pass · bronze_volume

silver · 35

business · 5

  • Pass · active_has_org
  • Pass · active_workers_unterminated
  • Pass · fte_valid
  • Pass · one_assignment_per_worker_min
  • Pass · salary_positive

not_null · 4

  • Pass · candidate_requisition_not_null
  • Pass · location_id_not_null
  • Pass · org_id_not_null
  • Pass · worker_id_not_null

referential · 13

  • Pass · dim_recruiter_covers_opening_recruiters
  • Pass · snap_recruiter_id_subseteq_dim_recruiter
  • Pass · candidate_requisition_fk
  • Pass · compensation_worker_fk
  • Pass · learning_worker_fk
  • Pass · location_fk
  • Pass · offer_application_ri
  • Pass · offer_requisition_ri
  • Pass · performance_worker_fk
  • Pass · term_has_separation_ri
  • Pass · worker_job_fk
  • Pass · worker_manager_fk
  • Pass · worker_org_fk

temporal · 7

  • Pass · application_before_hire
  • Pass · assignment_effective_window
  • Pass · hire_before_termination
  • Pass · hist_first_valid_from_eq_hire
  • Pass · hist_no_gap
  • Pass · hist_no_overlap
  • Pass · promotion_after_hire

transaction · 3

  • Pass · hires_eq_accepted_offers
  • Pass · rejected_without_later_accept_keeps_open_or_cancelled
  • Pass · window_hires_via_t1

unique · 3

  • Pass · unique_assignment_id
  • Pass · unique_requisition_id
  • Pass · unique_worker_id

gold · 36

business · 3

  • Pass · certified_ceo_count
  • Pass · headcount_non_negative
  • Pass · span_max_le_15

freshness · 1

  • Pass · overview_rows_present

metric_range · 26

  • Pass · attrition_rate_bounds
  • Pass · compa_ratio_range
  • Pass · metric_range_applications_per_opening
  • Pass · metric_range_average_headcount
  • Pass · metric_range_compa_ratio_median
  • Pass · metric_range_engagement_score
  • Pass · metric_range_headcount
  • Pass · metric_range_hires
  • Pass · metric_range_internal_mobility_rate
  • Pass · metric_range_involuntary_attrition_rate
  • Pass · metric_range_manager_turnover_rate
  • Pass · metric_range_offer_acceptance_rate
  • Pass · metric_range_promotion_rate
  • Pass · metric_range_quality_of_hire
  • Pass · metric_range_recruiter_load
  • Pass · metric_range_regrettable_attrition_rate
  • Pass · metric_range_rehires
  • Pass · metric_range_skill_coverage
  • Pass · metric_range_span_of_control
  • Pass · metric_range_time_in_stage_hours
  • Pass · metric_range_time_to_fill_days
  • Pass · metric_range_training_hours_per_worker
  • Pass · metric_range_voluntary_attrition_rate
  • Pass · mobility_rate_bounds
  • Pass · offer_acceptance_bounds
  • Pass · skill_coverage_bounds

referential · 1

  • Pass · orphan_manager_at_month_end

roll_forward · 3

  • Pass · evt_worker_change_has_hist_attr_switch
  • Pass · hist_attr_switch_has_evt_worker_change
  • Pass · snapshot_roll_forward

row_count · 2

  • Pass · active_headcount_scale
  • Pass · hire_events_present

All tests in this trusted snapshot passed.

What site visitor can read

Six cells: own org (Engineering), other org (Sales), company-level × internal Headcount vs confidential median compa-ratio. Engineering leader and HRBP are bound to Engineering: Sales confidential is org_scope. Company-level confidential is denied for the leader (company_scope_confidential) and allowed for HRBP. Visitor confidential is sensitivity.

ScopeMetricDecisionas_of
own orgheadcountallow2026-08-31
own orgcompa_ratio_mediandeny · sensitivity2026-08-31
other orgheadcountallow2026-08-31
other orgcompa_ratio_mediandeny · sensitivity2026-08-31
company levelheadcountallow2026-08-31
company levelcompa_ratio_mediandeny · sensitivity2026-08-31

Golden-set scorecard

Fail · tag phase4-p1-r6 · 101 cases

llm usage rate
0
metric accuracy
1
refuse accuracy
1
session consistency
1
chip differentiation
1
causal violation rate
0
tool selection accuracy
0.9888
suppression violation rate
0

LLM usage rate 0 — this run validates the deterministic skeleton path; the LLM-assisted path is covered by E1-LLM-* and the stub test, pending a successful upstream call.

Metric Explorer

Known limitations

Governance maturity for this demo — not a legal disclaimer.

  • Graded aggregate access has an inference path: an HRBP who can read company-level and own-org confidential aggregates can, for mean-type metrics, infer other-org values. Medians are weaker but not zero. Cross-query differential protection is not implemented (architecture §9.4 v1.1).
  • Suppression uses as-of month n as the denominator. Rates on this page are trailing-12m.
  • The dataset is frozen at data-v1. Quality tests run against that snapshot, not against a live daily rebuild.