WorkflowEmployment & Labor Law

Reduction in Force Adverse Impact Analysis

Reduction in force adverse impact analysis that tests a layoff list unit by unit for age, sex and race, and builds the release's age disclosure from that tested list.

Watch it work Book a walkthrough

Walk the bench, the disclosure and the release ↓

Watch with transcript →

The problem

The layoff list keeps moving, and the release's age table moves out of step

Once managers have picked their people, the employer's lawyer has two questions before anyone is told. Is the list fair inside each decisional unit, by age, sex and race? And will the release hold for the workers over forty? The economist tests the census in a workbook, and an associate builds the release's age table again by hand from a different export.

Then the list changes: a name swapped, a resignation, a transfer, and each version is run and checked again. Meanwhile a company-wide test passes while one unit fails, a reason written as near retirement sits unread until discovery, and a site's own notice law gets checked after the date is set. One wrong age in the disclosure, or ages given in bands, and the age waiver can fail for everyone who signed.

Built forEmployment partnersLabor economistsEmployment associatesHuman resources leaders

The Problem Today

Layoff lists are tested in spreadsheets, and the age table in each release is retyped from a second export.

1.2M: job cuts employers announced in 2025, the most since 2020 published
45 days: to consider a group-layoff release, or the age waiver fails published

Spread across workbooks today

Census
Tests
Age table
Notices

Aggregates hide a unitA company-wide pass can mask one unit that fails
Reasons read too lateAge-coded reasons surface in discovery

The Solution

List matched to the censusEach selection tied to its person, unit and site, version by version
Each unit testedFour-fifths ratio and significance by age, sex and race, unit by unit
Counsel holds or clearsA failing unit is held for a documented re-review, with a written reason
Age disclosure, generatedJob titles and ages of those selected and not, matching the analysis

Built-in controls

Analysis kept privileged, at counsel's direction
No selection changed on the statistics alone
Managers never see age, sex or race

What You Get

Counsel sends notices knowing which unit failed, why, and what was done about it.

Day 1: tests, disclosure and memo the day after the list arrives our estimate

A release that holdsThe group consideration and revocation periods, with the age disclosure attached
Notice dates set by siteA site's longer state notice period caught before the date is fixed
Reasons fixed before noticeAge-coded words found and re-scored while the list can still change

From census to privileged memo

Reduction in force adverse impact analysis, one decisional unit at a time

Nine features that carry a layoff list from the census match through unit tests, selection reasons, the partner's decision, the disclosure, the release and the memo.

01Layoff list v3
A layoff list matched to the census, with changes between versions
How it works
- Each of 180 selections tied to its census row by employee id
- 0 missing demographics and 0 duplicates before a test runs
- Changes from v2 listed: 4 added, 2 removed, with the reason for each removal
- All 70 tests re-run on v3 after the change
Employment partnerKnows which version of the list was tested and what moved since the last one, without comparing two spreadsheets line by line.
02Decisional units
Decisional units proposed from the org chart, and doubtful ones questioned
How it works
- Nine units, each with who decided and on what criteria
- Sites, people, selections and layoff rate on each row
- Where one manager ranked staff across three sites, the mapper asks before testing
- Units under 10 selections go to the economist to read, not tested alone
Employment partnerSettles who is compared with whom before any test runs, and has that answer on the record with the org chart behind it.
03All 70 tests
A test matrix that shows why the company-wide pass misleads
How it works
- Seven groups in each of nine units, plus the aggregate row
- Each cell shows the four-fifths ratio and the standard deviations
- All 9 units, age 50+: 0.84 and 1.15 standard deviations (SD), a pass
- The engineering unit, age 50+: 0.28 and 3.6 SD, a fail
Labor economistSees in one grid that a firm-wide number hides a failing unit, and opens any cell on the bench with one click.
04Test bench, engineering
A test bench showing one unit's people, rates and significance together
How it works
- Each person in the unit as a dot on an age line, selected ones filled
- Group and comparison layoff rates side by side, against the 0.80 line
- 13 of 40 aged 50 and over selected, against 11 of 120 under 50
- 3.57 standard deviations and a Fisher exact p of 0.001, cited to census v3
Employment partnerDecides a unit with the people, the rates and the significance on one screen, not spread across an economist's workbook and an email thread.
05Selection rationales
Selection reasons read against the written criteria, age-coded words marked
How it works
- All 24 engineering reasons read against rating, skills and seats per team
- 3 reasons carry age-coded words and 2 lack a skills score
- 13 aged 50+ selected, of whom 7 are mainframe developers
- The reviewer asks the client for a re-score; it does not accuse a manager
Employment associateFinds a reason like not a digital native while a documented re-score can still answer it, long before it would surface in discovery.
06Hold for re-review
A written-reason hold that keeps a failing unit out of the disclosure
How it works
- Hold for re-review, or clear with a reason; both need words
- The reason goes on the record and into the privileged memo
- A held unit stays out of the disclosure and the release packets
- Clearing a failing unit needs a reason that is not statistical, and the economist is told
Employment partnerCan show later which unit failed, what was decided and why, in the partner's own words with the date beside them.
07Disclosure, generated
A decisional unit disclosure built from the tested list, in one-year ages
How it works
- Generated for the 8 cleared units: 1,260 people, 156 selected
- Job titles and ages of those selected and not selected, unit by unit
- Ages as of the notice date, in one-year steps with no bands
- Reconciled with tested list v3 at a difference of 0
Employment associateStops retyping the age table from another export, the step where one wrong age can void the age releases across the program.
08Release check
A release template checked element by element, wrong periods caught
How it works
- 8 required elements checked against template v4, each cited to its section
- Section 6 gives twenty-one days, the period for a single release; a group program needs 45
- Consideration flagged where one site's statutory severance is already owed
- 6 of 8 pass before the fix goes in by redline
Employment associateCatches the wrong consideration period in the client's own template before it reaches a release packet for each employee over forty.
09Privileged memo
A privileged memo tracing each figure to its source, sent on approval
How it works
- What was done, what was found, what to do, and privilege, drafted from the bench
- Figures link to the census, the criteria, section 6 and the headcounts
- A ready-to-send list: engineering decided, disclosure generated, section 6 fixed
- Read by the general counsel and the HR vice president only
Employment partnerGives the client one privileged account of what was tested and done, with each figure tied to the list version it came from.

In engineering, 13 of 40 aged 50 and over were picked, against 11 of 120

One list read two ways. Only the unit view shows the problem.

Comparison	Layoff rate, group and rest	Four-fifths ratio	Standard deviations
Engineering, age 40+	18.8% and 10.7%	0.57	1.44
Engineering, age 50+	32.5% and 9.2%	0.28	3.57
Engineering, age 60+	41.7% and 12.8%	0.31	2.68
All 9 units, age 50+	14.3% and 12.0%	0.84	1.15
All 9 units, age 60+	13.8% and 12.6%	0.91	0.35

Around the bench

Six more screens a client manager, an associate and a partner open

From the manager's view of the list to the practice's own thresholds.

Layoff list in the client manager's view: name, job title, unit, site and years for each selection, with the age, sex and race columns removed — **What a client manager sees**The same list with age, sex and race taken off: names, titles, sites, years and whether a reason is on file.

Notice check by site: one site triggered with 90 days' notice and 700 weeks of severance across 104 people, two sites with no duty, and a later date proposed — **Notice duties, site by site**Losses against headcount at each site, with a 90-day look-back. One site's state law is triggered: 90 days' notice and 700 weeks of statutory severance across 104 people.

Test bench with all nine units selected: 1,420 people, 180 selected, each group passing, age 50 and over at a four-fifths ratio of 0.84 — **The aggregate, on the bench**All 9 units: 58 of 406 aged 50 and over against 122 of 1,014 younger, 0.84 at 1.15 SD. The agent reads it with the units, never in their place.

Release check after the fix: all eight elements meet, section 6 reads forty-five days in template v5, and the state severance rider is added — **Section 6, fixed in redline**Template v5 reads forty-five (45) days, fixed by the associate and approved by the partner. The severance rider puts release pay on top of the statutory amount.

Audit trail for the matter: the hold, the disclosure for 8 units, the release fix and memo approval, above the 70 tests run and the 24 reasons read — **The matter's record of each act**Upload, census match, tests, reasons read, hold, disclosure, fix and memo, each with the person or agent behind it.

Settings for tests, groups, disclosure, roles, privilege and retention: a 0.80 four-fifths line, 2.0 standard deviations and a ten-selection minimum — **Thresholds the practice sets**A 0.80 four-fifths line, significance at 2.0 standard deviations, Fisher's exact on each table, and 10 selections as the minimum to test.

Age-coded reasons, a split unit, a short release period and an early date

Four findings on list v3, each put in front of a person with its source beside it.

Age-coded words in three reasons

Near retirement, set in his ways, not a digital native. The partner approves a re-score request to the client.

A unit that crosses three sites

Fourteen adjusters report to a team lead at another site, and one vice president ranked all 310 together. The mapper asks whether the unit stays whole.

The wrong consideration period

Section 6 of the client's template gives twenty-one days, the period for a single release. A group program needs 45, so a release signed on it could fail for each worker over forty.

A separation date set too early

One site's state law needs 90 days' notice and severance for 104 people. The planned separation date falls inside that window, so the checker proposes a later one.

The partner holds or clears, and no selection moves on the statistics alone

Holding or clearing a unitOnly the employment partner, with a written reason. The economist signs the method and cannot clear a unit.

Changing who is selectedA failing unit goes back for a documented re-score against the written criteria. The numbers alone never change a selection.

Writing to an employeeAgents send no notice, release or message to an employee. The memo and client requests go out on the partner's approval.

Seeing the statisticsClient HR sees results by unit and never per person. Managers never see age, sex or race.

Reduction in force adverse impact analysis, before the first notice goes out

What is reduction in force adverse impact analysis?

Reduction in force adverse impact analysis tests a proposed layoff list for whether older workers, women or a racial group are selected at a markedly higher rate than the rest. Here it runs in each decisional unit and in aggregate, at counsel's direction so it stays privileged.

Is a company-wide pass enough for an adverse impact analysis for a layoff?

No. Selections are made unit by unit, so a firm-wide figure can pass while one unit fails. On the list shown here, the company-wide age result passes and the engineering unit fails. Take that unit out and the other units select workers fifty and over at no higher rate than younger ones.

How does it keep the analysis privileged?

The matter opens at the direction of counsel, and each screen, export and memo carries the privilege label. The selection reasons are kept in a separate file that is not privileged. Client HR sees results by unit only, and managers see the list without age, sex or race.

Does the age disclosure include a unit still under review?

No. A held unit stays out until the partner clears it, and its section is then generated from the list as it stands. The disclosure states the decisional unit, the eligibility factors and the time limits, with the job titles and ages of those selected and not selected.

Does it check notice laws for each site?

Yes. The checker counts losses at each site against its headcount, looks back for earlier losses, and reads each state notice law beside the national one. When a site needs a longer notice period, it proposes a later separation date for the partner to send to the client.

Employment & Labor Law