RIF Compliance

RIF selection criteria: how to build a defensible process

RIF selection decisions must rest on legitimate, non-discriminatory, documented business reasons. The employer bears the burden of showing the criteria were set in advance, applied consistently, and were not a pretext for discrimination. This guide covers how to choose criteria, build a scoring matrix, define the decisional unit, and clear the protected-class traps before notice goes out.

SelectionAdverse ImpactADEADocumentation

Before

Set criteria first

4/5ths

Run adverse impact

ADEA + FMLA

Two flags to check

What RIF selection criteria are, and why they matter

Selection criteria are the rules that decide which employees are included in a reduction in force. They are the single most scrutinized part of any RIF, because they are where discrimination, intentional or not, most often enters the process. The legal principle is straightforward: every selection must be based on a legitimate, non-discriminatory, documented business reason.

When a terminated employee challenges a RIF, courts apply a burden-shifting framework. In plain English: the employee shows they are in a protected class and were let go. The employer must then articulate a legitimate, non-discriminatory reason for the selection. The employee then gets to argue that the stated reason is a pretext, a cover story for discrimination. Documented, consistently applied criteria are what let the employer carry its burden and defeat the pretext argument. Vague or after-the-fact criteria are what let the employee win.

The burden of proof on consistency sits with the employer. It is not enough to have a reason. You must be able to show the criteria existed in writing before names were attached, that the same criteria were applied the same way to similarly situated employees, and that the outcome was tested for adverse impact before notice went out.

Criteria first, names second

The order of operations is the whole game. Define and weight the criteria, then score employees against them. Setting criteria after you have looked at the list, so the people you already wanted to cut happen to score lowest, is called reverse-engineering, and it is the most common source of RIF litigation. If the timeline shows criteria were finalized after names were reviewed, the rest of the process rarely survives.

Objective vs. subjective criteria

Objective criteria are measurable and applied the same way for everyone, which makes them far easier to defend. Subjective criteria depend on judgment and are harder to apply consistently, so they invite a pretext argument. Use objective criteria wherever possible. Where subjective judgment is unavoidable, anchor it to a defined rubric and document the rationale.

Objective criteria (preferred)

  • Position elimination

    The role no longer exists after a restructuring. The cleanest criterion when an entire function is removed.

  • Skills matrix / skills inventory

    Documented competency gaps measured against future-state requirements, scored on a defined scale.

  • Performance ratings

    Must be documented, consistent, and free of recency bias. Use a completed review cycle within the last 12 months.

  • Seniority / last-in-first-out (LIFO)

    Easy to defend and hard to attack, but not always aligned with the business need for specific skills.

  • Attendance / disciplinary record

    Documented and consistently applied across the decisional unit, not pulled selectively.

  • Required certifications or licenses

    Credentials the remaining roles require. Employees without them can be selected on a defensible basis.

Subjective criteria (higher risk)

  • "Critical to the business" designationsHigh risk

    Hard to define and easy to apply inconsistently. Often masks a predetermined outcome.

  • Manager judgment without documented rationaleHigh risk

    A bare opinion with no written basis is the easiest selection to attack as pretext.

  • "Cultural fit" assessmentsHigh risk

    Frequently a proxy for age, race, or other protected traits. Rarely survives scrutiny.

  • Potential-based ratings without behavioral anchorsMedium risk

    Speculative judgments about future value, with no objective measure, skew toward bias.

  • Vague "skills" assessments without a rubricMedium risk

    A skills judgment with no defined scoring scale is subjective in everything but name.

The selection matrix

Best practice is a documented selection matrix: criteria across the columns, every employee in the decisional unit down the rows, and a score in each cell. Each criterion carries a weight assigned in advance. The total weighted score determines selection. The matrix forces consistency, creates a paper trail, and makes the order of operations auditable, criteria and weights first, names second.

1

Define the criteria and weights

Choose the criteria, mostly objective, and assign a weight to each based on the business need. A skills-driven reduction might weight the skills matrix at 50%, performance at 30%, and seniority at 20%. Lock the weights before any names are scored.

2

Define the decisional unit

List every employee in the unit, the department, function, location, or job family within which selection is made. Everyone scored against the same criteria must be genuinely comparable.

3

Score each employee against each criterion

Use a defined scale (for example 1 to 5) with written anchors so two managers would score the same employee the same way. Record the basis for each score, especially for any criterion involving judgment.

4

Run adverse impact analysis before finalizing

Calculate weighted totals, rank the unit, and identify the selections. Before finalizing, run adverse impact analysis on the selected group. Only then is the list ready for legal review.

Use the adverse impact calculator

The one rule that holds the matrix together

Criteria and weights are set before names are attached. If weights are adjusted after scoring begins, or criteria are added once the ranking is visible, the matrix stops being evidence of a fair process and becomes evidence of reverse-engineering. Date and version the matrix so the timeline is unambiguous.

Defining the decisional unit

The decisional unit is the group within which the selection is made. It is the pool of employees who are compared against each other on the same criteria. Defining it correctly matters because the adverse impact analysis is run within the decisional unit, and the unit also drives the OWBPA disclosure required for group terminations of employees 40 and older.

The unit must be defined before selection begins, following the actual organizational structure and the business rationale for the reduction. It is typically one of the following:

Department

All employees in a single department, for example all of Customer Support.

Function

A function that spans the organization, for example all individual-contributor recruiters.

Location

A single site or office where the reduction is taking place.

Job family

A group of comparable roles, for example all software engineers at a given level.

Why the boundaries matter

The size and shape of the decisional unit changes the adverse impact result. Drawing the unit narrowly or broadly to manufacture a favorable outcome, or to dilute a disparity, is a recognized red flag. Define the unit from the business rationale first, then run the analysis. Do not redraw the unit after seeing the numbers.

Protected class traps

Even a clean, criteria-driven matrix can produce a selection that triggers independent legal exposure. These four traps surface after the list is drafted and must be checked before it is finalized.

Age (ADEA)

40 and older

Employees 40 and older are protected under the ADEA. A statistical age skew in the selected group raises ADEA risk even when each individual selection looks defensible. The OWBPA additionally requires a decisional unit disclosure for group terminations, listing the job titles and ages of those selected and not selected. Run an age-band analysis, not just race and sex, before finalizing.

Action: Before finalizing the list, run an age distribution comparison of selected vs. non-selected employees. Flag for legal review if the selected group skews significantly older.

FMLA / ADA leave

Leave status

Selecting an employee because they are on leave, or disproportionately selecting employees who recently took FMLA or ADA leave, is independently unlawful regardless of the RIF rationale. Flag every employee on leave or recently returned, and have counsel confirm the selection rests on criteria applied to everyone, before the list is finalized.

Action: Identify all employees currently on or recently returned from FMLA or ADA leave before the selection list is finalized. Have counsel review each before including them.

Pregnancy

PDA / parental leave

Selecting a pregnant employee, or one who recently returned from parental leave, requires documented justification that has nothing to do with the pregnancy or the leave. The same criteria and scores must be shown to apply, and the timing relative to the leave should be reviewed by counsel.

Action: Flag any employee who is pregnant or returned from parental leave in the last 6 months. Document the specific objective business reason for inclusion that is unrelated to the leave.

Prior complaints

Retaliation

Selecting an employee who recently filed an HR complaint or an EEOC charge raises retaliation risk regardless of how sound the RIF rationale is. Flag any recent protected activity and confirm the selection would have occurred on the merits independent of the complaint.

Action: Search HR records for any EEOC charges, internal complaints, or protected activity in the last 24 months. Have legal review each flagged employee before the list goes out.

The adverse impact connection

Good criteria do not guarantee a fair outcome. Even neutral, objective criteria can fall more heavily on one protected group. Once the selection list is drafted, run a 4/5ths rule analysis on the selected group against the decisional unit, and a Fisher's exact test where the sample is small. This step must happen before notice is given, while the list can still change.

If the analysis shows adverse impact, you have two options: adjust the list so the disparity is eliminated or meaningfully reduced, or document a compelling, criteria-based business justification for the selections as they stand. What you cannot do is ignore it. An unaddressed disparity sitting on the record is far worse than one you found and resolved.

Run the numbers before you finalize

Use the adverse impact calculator to check the 4/5ths rule across race, sex, and age bands within each decisional unit, with small-sample flagging built in.

Open the adverse impact calculator

Documentation requirements

If a RIF is challenged, the file is the defense. These five documents, created in this order and dated, are what let the employer carry its burden.

  • Written criteria and weights

    Documented and dated before any employee names are reviewed.

  • Completed selection matrix for each decisional unit

    Every employee scored against every criterion, with the basis for judgment-based scores recorded.

  • Business rationale memo

    Why these criteria, why these weights, and why this headcount, tied to the business reason for the reduction.

  • Adverse impact analysis results

    The 4/5ths rule output (and small-sample tests) for each decisional unit, run before notice.

  • Legal review sign-off before the list is finalized

    Counsel reviews the matrix, the analysis, and all flagged employees before anyone is notified.

Free download

Get the full RIF compliance checklist

66 steps across 9 phases, including selection criteria, the scoring matrix, adverse impact analysis, and the documentation your legal team will ask for. Formatted as an Excel workbook your team can track in real time.

Common mistakes

Setting criteria after reviewing names ("reverse engineering")

The most litigated RIF mistake. If criteria or weights were finalized after the list of names was visible, so the intended cuts happen to score lowest, the process reads as a pretext. Lock and date criteria before any names are attached.

Using performance ratings more than 18 months old

Stale ratings may not reflect current performance and suggest the employer reached back for a convenient justification. Use a completed review cycle within the last 12 months, and never create ratings after the list is drafted.

Failing to flag employees on FMLA or ADA leave before finalization

Selecting an employee on leave, or disproportionately selecting recent leave-takers, is independently unlawful. Leave status must be flagged and reviewed by counsel before the list is final, not discovered afterward.

Inconsistently applying criteria across similar roles

If two comparable employees are scored on different bases, or a criterion is applied to one decisional unit but not another, the consistency the employer must prove falls apart. Apply the same criteria the same way to everyone in the unit.

Running adverse impact analysis after notice is given

Once notice goes out, the list cannot be quietly adjusted. Running the 4/5ths analysis too late leaves any disparity on the record with no chance to address it. Run it before notice, every time.

No written business rationale memo

Without a memo explaining why these criteria, why these weights, and why this headcount, the matrix floats free of any business justification. The rationale is what connects the selections to a legitimate, non-discriminatory reason.

Frequently asked questions

What criteria can we use to select employees for a RIF?

RIF selection must be based on legitimate, non-discriminatory, documented business reasons. The strongest criteria are objective: position elimination, a documented skills matrix measured against future-state requirements, current performance ratings, seniority (last-in-first-out), documented attendance or disciplinary records, and certifications or licenses required for remaining roles. Subjective criteria such as "critical to the business" designations, manager judgment without documented rationale, and "cultural fit" assessments carry higher legal risk because they are harder to apply consistently and easier to attack as a pretext for discrimination. Whatever criteria you use, they must be documented in writing and set before any employee names are reviewed.

Can we use performance ratings as a RIF selection criterion?

Yes, but only if the ratings are documented, consistent, and current. Courts scrutinize performance-based selections for recency bias and for ratings that appear to have been adjusted to justify a predetermined outcome. Use ratings from a completed review cycle within the last 12 months. Ratings more than 18 months old are difficult to defend because they may not reflect current performance and suggest the employer reached back for a convenient justification. If ratings were never formally documented, do not create them after the layoff list is drafted, which is one of the clearest signs of reverse-engineering.

What is a decisional unit in a RIF?

A decisional unit is the group of employees within which the RIF selection is made. It is usually a department, function, location, or job family. The decisional unit must be defined before selection begins, because adverse impact analysis is run within it. The decisional unit also drives the OWBPA disclosure required for group terminations of employees 40 and older. Defining the unit narrowly or broadly to manufacture a favorable outcome is a red flag, so the boundaries should follow the actual organizational structure and the business rationale for the reduction.

What happens if our selection criteria have adverse impact?

Once the selection list is drafted, run a 4/5ths rule analysis (and Fisher’s exact test for small samples) on the selected group against the decisional unit. If a protected group is selected at a disproportionate rate, you have two options: adjust the list so the disparity is eliminated or reduced, or document a compelling, criteria-based business justification for the selections. This analysis must happen before notice is given, while you can still change the list. Running it after notice has gone out is too late to adjust and leaves the disparity on the record.

Can we select an employee who is on FMLA leave?

You can include an employee on FMLA or ADA leave in a RIF, but only if the selection has nothing to do with the leave and is supported by the same criteria applied to everyone else. Selecting an employee because they are on leave, or disproportionately selecting employees who recently took leave, is independently unlawful regardless of the RIF rationale. Leave status should be flagged before the list is finalized and reviewed by counsel. The employer must be able to show the employee would have been selected on the merits even if no leave had been taken.

Do we need legal review of the selection list before notifying employees?

Yes. Legal review of the finalized selection list before any employee is notified is a documentation requirement and a practical safeguard. Counsel reviews the written criteria and weights, the completed selection matrix, the business rationale memo, the adverse impact analysis results, and any flagged employees on FMLA or ADA leave, recently returned from parental leave, or who recently filed an HR complaint or EEOC charge. Once notice is given, the list cannot be quietly adjusted, so the legal sign-off must come first.

People Plan

Selection matrix and adverse impact, built in

People Plan builds the selection matrix for each decisional unit, scores employees against your weighted criteria, runs the 4/5ths rule and age-band analysis before you finalize, and flags employees on FMLA or ADA leave for legal review. Your team reviews rather than reconstructs.

Legal disclaimer

This guide is provided for general informational purposes and does not constitute legal advice. RIF selection analysis is fact-specific and depends on the decisional unit, the criteria used, and the workforce composition. Always have employment counsel review selection criteria and the finalized list before notifying any employee. People Plan is not a law firm.