LensReading which lens this session carries.
Read this before reading anything else on this page

Fairness readings computed on modeled data demonstrate the instrument, not a real audit result. Every ratio, every alarm and every drift reading below was computed by the same code that would run against a customer’s real population — but the population underneath is modeled. Nothing on this page is a finding about any real group of people, and it must not be quoted as one.

Govern · item 18

Adverse-impact monitoring

Any decision that touches a person’s job, pay or prospects is measured across cohorts, in every reporting period, against a stated disparity threshold. Where a cohort is too small to produce a defensible ratio, no ratio is produced — the cell is shown as unmeasured rather than passed. Where monitoring is not running at all, that is shown as its own alarm.

Risk class and appeals

Health

Whether agent decisions land evenly

People decision points under monitoring, the alarms raised against them and the appeals outstanding.

Decision points
8
25,100 people affected each month
Under monitoring
7 of 8
1 people-affecting decision runs unmonitored
Readings computed
188
6 periods · 5 cohort dimensions
Outside threshold
14
8 readings below the reporting floor
Alarms raised
23
23 still open · 6 high severity
Review overdue
1
13 appeals recorded against these decisions

Every decision that affects a person

Population, the agent that makes or shapes the decision, the named human accountable for it, and the state of its monitoring.

Decision pointKindPopulationAgentAccountable humanRisk classReadingsWorst ratioAlarmsReviewStatus
Candidate screening and shortlist
Talent Acquisition
Selection8,420/mo
Screening Agent
A3
Recruiting pod lead, per requisition
Recruiter
High risk300.76
3 breaches all periods
4 open
Monthly cohort review, quarterly threshold review
on schedule
Monitored
Interview slot allocation
Talent Acquisition
Allocation2,140/mo
Scheduling Agent
A3
Coordinator on the requisition
Recruiting coordinator
Limited risk180.832 open
Quarterly
on schedule
Monitored
Performance calibration preparation
Performance Management
Ranking4,180/mo
Calibration Prep Agent
A1
Function leader chairing the session
Calibration chair
High risk420.871 open
Per cycle, twice yearly
on schedule
Monitored, Germany excluded
Succession and promotion slate
Talent Mobility
Selection610/mo
Succession Slate Agent
A2
Morgan Idris, chairing the talent council
Talent council
High risk300.74
7 breaches all periods
4 open
Quarterly
on schedule
Alarm open
Merit and pay recommendation
Total Rewards
Scoring5,210/mo
Merit Cycle Agent
A2
Marcus Weill, VP Total Rewards, as final reviewer
Line manager and reward partner
High risk240.941 open
Per cycle, annual, with a mid-cycle equity read
on schedule
Monitored
Attrition risk flagging
Exit Management
Flagging
adverse-direction
3,840/mo
Attrition Signal Agent
A2
HRBP for the population
HR business partner
High risk361.31
4 breaches all periods
5 open
Quarterly
51 d overdue
Review overdue
Performance improvement plan drafting
Performance Management
Drafting
adverse-direction
180/mo
PIP Drafting Agent
A1
HRBP countersigns every plan
Line manager with HR business partner
High risk8
8 below floor
unmeasured5 open
Quarterly
on schedule
Below reporting floor
Access revocation on exit
Exit Management
Allocation520/mo
Deprovisioning Coordination Agent
A2
Exit coordinator in the Hire-to-Retire tower
Exit coordinator
Limited risk0unmeasured1 open
Quarterly
on schedule
Not monitored

Alarms

Derived from the readings, never authored. An alarm is raised when a ratio breaches, when it drifts across consecutive periods, when a cohort cannot be measured, or when monitoring is missing entirely.

RefKindDecision pointCohortObservedThresholdAgeOwnerStatus
FAIR-0004Threshold breach
High
Candidate screening and shortlistage-band · 50 and over0.760.8078 d
Priya Raman
Chief People Officer
Open
FAIR-0008Threshold breach
High
Succession and promotion slateGender · Women0.740.8078 d
Priya Raman
Chief People Officer
Open
FAIR-0011Threshold breach
High
Succession and promotion slateage-band · 50 and over0.750.80109 d
Priya Raman
Chief People Officer
Open
FAIR-0015Threshold breach
High
Attrition risk flaggingage-band · Under 301.261.2517 d
Priya Raman
Chief People Officer
Open
FAIR-0016Threshold breach
High
Attrition risk flaggingtenure-band · Under 2 years1.311.2578 d
Priya Raman
Chief People Officer
Open
FAIR-0023Monitoring absent
High
Access revocation on exitall · all cohorts170 d
Yusuf Adeyemi
Head of Identity and Access
Open
FAIR-0002Never reportable
Low
Candidate screening and shortlistGender · Non-binary and self-described170 d
Nadia Kovac
Data Protection Officer
Monitoring
FAIR-0005Never reportable
Low
Interview slot allocationGender · Non-binary and self-described170 d
Nadia Kovac
Data Protection Officer
Monitoring
FAIR-0007Never reportable
Low
Performance calibration preparationGender · Non-binary and self-described170 d
Nadia Kovac
Data Protection Officer
Monitoring
FAIR-0009Never reportable
Low
Succession and promotion slateGender · Non-binary and self-described170 d
Nadia Kovac
Data Protection Officer
Monitoring
FAIR-0012Never reportable
Low
Merit and pay recommendationGender · Non-binary and self-described170 d
Nadia Kovac
Data Protection Officer
Monitoring
FAIR-0014Never reportable
Low
Attrition risk flaggingGender · Non-binary and self-described170 d
Nadia Kovac
Data Protection Officer
Monitoring
FAIR-0018Never reportable
Low
Performance improvement plan draftingGender · Non-binary and self-described170 d
Nadia Kovac
Data Protection Officer
Monitoring
FAIR-0001Sustained drift
Medium
Candidate screening and shortlistGender · Women0.870.80170 d
Nadia Kovac
Data Protection Officer
Acknowledged
FAIR-0003Sustained drift
Medium
Candidate screening and shortlistage-band · 40 to 490.850.80170 d
Nadia Kovac
Data Protection Officer
Acknowledged
FAIR-0006Sustained drift
Medium
Interview slot allocationregion · Asia Pacific0.830.80170 d
Nadia Kovac
Data Protection Officer
Acknowledged
FAIR-0010Sustained drift
Medium
Succession and promotion slateEthnicity · Minority ethnic groups, aggregated0.810.80170 d
Nadia Kovac
Data Protection Officer
Acknowledged
FAIR-0013Sustained drift
Medium
Attrition risk flaggingGender · Women1.151.25170 d
Nadia Kovac
Data Protection Officer
Acknowledged
FAIR-0017Below reporting floor
Medium
Performance improvement plan draftingGender · Women0.911.25109 d
Nadia Kovac
Data Protection Officer
Acknowledged
FAIR-0019Below reporting floor
Medium
Performance improvement plan draftingage-band · Under 301.121.25109 d
Nadia Kovac
Data Protection Officer
Acknowledged
FAIR-0020Below reporting floor
Medium
Performance improvement plan draftingage-band · 40 to 490.931.25109 d
Nadia Kovac
Data Protection Officer
Acknowledged
FAIR-0021Below reporting floor
Medium
Performance improvement plan draftingage-band · 50 and over0.801.25109 d
Nadia Kovac
Data Protection Officer
Acknowledged
FAIR-0022Review overdue
Medium
Attrition risk flaggingall · all cohorts51 d
Priya Raman
Named reviewer for this decision
Acknowledged
None of the 23 alarms has been closed. That is the honest state: alarms are being raised and owned, and none has yet been worked to a documented resolution. A closure rate would be a more comfortable number to show and it would not be true.

By cohort dimension

Which dimensions produce readings, and which produce alarms.

age-band3 cohorts
78 readings · 8 outside threshold · 6 below floor
Ethnicity1 cohorts
24 readings · 0 outside threshold · 0 below floor
Gender1 cohorts
38 readings · 3 outside threshold · 2 below floor
region2 cohorts
12 readings · 0 outside threshold · 0 below floor
tenure-band2 cohorts
36 readings · 3 outside threshold · 0 below floor

Alarms by kind

Not every alarm is a breach. Two of these kinds mean we cannot see.

Threshold breach5
Impact ratio outside the disparity threshold in the latest period.
Monitoring absent1
A people-affecting decision point runs with no adverse-impact monitoring attached.
Never reportable7
Cohort declared for monitoring but never once large enough to produce a reading.
Sustained drift5
Ratio moving against a cohort across consecutive periods without yet breaching.
Below reporting floor4
Cohort too small to compute a defensible ratio.
Review overdue1
The documented review cadence has lapsed.

Review cadence

A monitoring instrument that nobody reviews is a log file. Each decision point carries a stated cadence, a named reviewer and a date the review is next due.

Candidate screening and shortlist
Monthly cohort review, quarterly threshold review
Reviewer: Nadia Kovac, Data Protection Officer, with Priya Raman
Next review scheduled
Interview slot allocation
Quarterly
Reviewer: Priya Raman
Next review scheduled
Performance calibration preparation
Per cycle, twice yearly
Reviewer: Priya Raman
Next review scheduled
Succession and promotion slate
Quarterly
Reviewer: Priya Raman
Next review scheduled
Merit and pay recommendation
Per cycle, annual, with a mid-cycle equity read
Reviewer: Marcus Weill with Nadia Kovac
Next review scheduled
Attrition risk flagging
Quarterly
Reviewer: Priya Raman
51 days overdue
Performance improvement plan drafting
Quarterly
Reviewer: Priya Raman
Next review scheduled
Access revocation on exit
Quarterly
Reviewer: Yusuf Adeyemi
Next review scheduled

Actions

What is waiting on a person

These are decisions that land on a person. Every item below has somebody on the other side of it waiting.

Operations

What this desk is allowed to start

A surface that only reports is not operable. This is the work this page can set in motion, and the bound it runs into.

Trigger and bound

This desk can take a fairness reading against a declared metric and raise an alarm. It cannot change a decision, overturn an appeal, suppress an alarm, or alter a threshold — and it deliberately cannot infer a protected characteristic that was not lawfully collected, which is why several readings on this page carry an inadequate sample.

Live observability

What the record shows right now

Alarm state. Monitoring means measured and inside the threshold.

Current distribution

23 alarms

Monitoring730%
Acknowledged1043%
Open626%

Is policy and strategy coming to fruition

Whether the written intent is holding here

7 of 8 people decision points are monitored. 16 fairness alarms are live.

Not holding on the record

The written commitment is that no person is affected by an agent decision without a route to challenge it and a measurement behind it. Partly held: 7 of 8 decision points are monitored, 1 are past their review date, and 8 readings have samples too small to conclude anything from. 2 appeals are overdue. Where a reading is marked inadequate it is because the population is too small to measure without identifying individuals, and it is left unmeasured rather than estimated.

What this page is, and what it is not

Fairness readings computed on modeled data demonstrate the instrument, not a real audit result. The arithmetic is the arithmetic a regulator would expect — impact ratio against a reference cohort, a stated threshold, a minimum sample before a ratio is computed at all — but the population is modeled. Read this page as evidence that the measurement exists and produces uncomfortable answers, not as evidence about any real workforce.

8 of 188 readings sit below the reporting floor and are shown as unmeasured. One decision point produced readings in only two of 6 periods and not one of its cells cleared the floor — so it has no fairness position at all, which is stated rather than smoothed. One cohort is declared for monitoring across every decision point and has never once been large enough to report.

One people-affecting decision runs with no adverse-impact monitoring attached at all, and an appeal has already been raised against it. That combination is the reason this page exists.