Compensation Outliers

Overview

The Compensation Outlier Flag helps you identify employees whose pay differs significantly from what would be expected based on your organization's broader compensation patterns. This flag is particularly valuable during merit cycles for:

• Proactive equity review: Surface potential pay disparities before finalizing compensation decisions
• Targeted manager conversations: Equip people leaders with data to justify or correct compensation anomalies
• Informed budget allocation: Ensure merit increases address systematic pay gaps rather than perpetuating them
• Documentation and compliance: Demonstrate thoughtful, data-driven compensation review processes

How It Works

Understanding the Analysis Method

The Outlier Flag uses regression modeling to analyze compensation patterns across your entire organization. This is fundamentally different from simple peer-to-peer comparisons, and understanding this distinction is critical to using the flag effectively.

What regression modeling does:

• Analyzes compensation data across your entire employee population
• Builds a predictive model based on the analysis attributes you select (such as Department, Job Family, and Level)
• Predicts what each employee's compensation should be given their unique combination of attributes
• Flags individuals whose actual compensation deviates significantly from this prediction

What regression modeling does NOT do:

• It does not create narrow peer groups of employees with identical job titles, levels, or departments
• It does not directly compare Employee A only to other employees who share all the same attributes
• It does not require a minimum number of "exact match" peers to generate insights

Why Regression Modeling Matters

This approach offers significant advantages over direct peer-to-peer comparison:

1. Captures nuanced pay patterns Traditional peer grouping might compare a Senior Software Engineer in the Product department only to other Senior Software Engineers in Product. Regression modeling instead asks: "Based on how we compensate people across departments, job families, and levels, what should this person earn?" This reveals more sophisticated patterns in your compensation philosophy.
2. Works with small or unique populations If you only have two Principal Engineers in the Sales department, a peer comparison approach would be limited or impossible. Regression modeling can still generate meaningful predictions by understanding how seniority and department influence pay across your entire organization.
3. Identifies intersectional disparities The model can detect when specific combinations of attributes lead to unexpected pay patterns—for example, if employees in Department X at Level Y are consistently paid differently than the model would predict based on how those attributes typically influence pay.
4. Handles complexity at scale As your organization grows and compensation becomes more complex, regression modeling adapts automatically. It can weigh multiple factors simultaneously without requiring you to manually define every possible peer group.

Example in Practice

Consider an employee with these attributes:

• Department: Engineering
• Job Family: Software Development
• Level: Senior

A traditional approach might compare them only to other Senior Software Developers in Engineering. But what if your organization has different pay scales across different engineering teams, or if job family influences pay more than department?

The regression model learns from your entire compensation dataset: how much does Level matter? How much does Department matter? How do these factors interact? It then predicts this employee's expected salary based on those learned patterns. If their actual salary is 15% below prediction, the flag appears—signaling a potential outlier worth investigating.

Configuration Guide

1. Name Your Flag

Give the flag a clear, descriptive name like "Peer salary differential" or "Compensation outlier."

2. Add a Description

Write a description that will help other planners understand the flag's purpose. For example:

• "Employees whose salary is lower than their peers"
• "Flags compensation that deviates significantly from predicted ranges"

3. Choose a Color

Select a color to make the flag easily identifiable in your planning interface. Light blue is the default, but you may want to use red or yellow for higher-visibility issues.

4. Set Visibility

By default, the flag appears in both the Smart Flags column and worksheet filters. Check "Hide from Smart Flags column and only show in worksheet filters" if you want the flag visible only when filtering, not as a persistent column indicator.

Setting Up Conditions

Step 1: Select a Compensation Component

Choose which compensation element you want to analyze for outliers. Common choices include:

• New Base Salary (most common for merit cycle planning)
• Current Base Salary
• Total Cash Compensation

Step 2: Choose Analysis Attributes

This is the most important configuration step. Select 2-4 attributes that you believe most significantly influence compensation at your organization. These might include:

• Department - if pay scales vary across different parts of the organization
• Job Family - if different types of roles have different compensation structures
• Level - if seniority is a primary driver of pay
• Location - if you have geographic pay differentials
• Performance rating - if pay-for-performance is central to your philosophy

:::info Important: The attributes you select train the regression model. The model will learn how these factors influence pay across your organization and use that learning to predict expected compensation. Choose attributes that genuinely matter to your compensation philosophy.

:::

:::info Note: Changes to analysis attributes apply to all outlier flags in your organization to maintain consistency across analyses.

:::

Step 3: Add Population Filters (Optional)

By default, the model compares each employee to patterns across your entire workforce using regression analysis. Use population filters only if you want to narrow the analysis to a specific subset—for example:

• Analyzing only Software Engineers (comparing each to patterns within that function)
• Reviewing only employees in a specific location
• Focusing on a particular department undergoing compensation review

To add a filter:

1. Click "Add column"
2. Select the attribute to filter on
3. Choose specific values to include

When to use filters: Use filters when you believe compensation patterns are fundamentally different in a specific population, not when you simply want to review that population. For instance, if sales roles have entirely different compensation structures than other roles, filter to sales. But if you just want to review sales employees for outliers within your broader compensation framework, analyze the full population and filter your view later.

Step 4: Set the Outlier Threshold

Define how much deviation from the predicted compensation triggers the flag:

• Threshold amount: Enter a numeric value (e.g., 10)
• Unit: Choose "% percent" or "$ dollar"
• Direction: Select "Varies by" to flag deviations in either direction (both over and under)

Example: A 10% threshold means employees whose actual compensation differs from the model's prediction by more than 10% will be flagged.

Choosing the right threshold:

• Start with 10-15% for an initial analysis to surface clear outliers
• Tighten to 5-8% for more comprehensive equity reviews
• Consider your organization's pay ranges and compression tolerance

Permissions (Optional)

By default, any planner can see flags for the employees in their worksheet. Use permission settings to restrict flag visibility to specific roles, such as Comp Analysts, HR Business Partners or Directors if the flag surfaces sensitive information.

Using the Flag in Your Merit Cycle

Once configured, the Outlier Flag will automatically appear when employees meet your criteria. Use these insights to:

1. Review flagged employees with managers to understand the root cause of the deviation
2. Prioritize merit increases for underpaid outliers to address equity issues
3. Investigate overpaid outliers to understand if there's a justifiable reason (e.g., market adjustment, retention case)
4. Document decisions about whether to address each outlier or accept the deviation

The flag provides the signal; your judgment provides the context and decision-making.

Using the Peers Tool for Additional Context

When a planner clicks on an employee flagged as an outlier, they'll see an explanation of why the regression model detected the deviation. Directly below this explanation, planners can access the Peers tool to perform a complementary analysis.

The Peers tool allows planners to compare the flagged employee to colleagues who share exact matching attributes—such as identical Department, Job Family, Level, or Performance Rating. This provides a secondary perspective that can help planners feel more confident about their compensation decisions.

Why use both views together:

• Regression analysis (the Outlier Flag) shows how the employee compares to organization-wide compensation patterns, revealing systematic disparities
• Peer comparison (the Peers tool) shows how the employee compares to direct colleagues with identical attributes, providing intuitive context

For example, an employee might be flagged as an outlier because their compensation is lower than the regression model predicts. Using the Peers tool, a planner might discover that this employee is actually paid similarly to the three other Senior Engineers in their department with the same performance rating. This could indicate that the entire peer group is underpaid relative to broader patterns—a valuable insight for addressing team-wide equity issues rather than viewing the outlier in isolation.

Learn how to configure the Peer Comparison tool here: Comparison documentation

Frequently Asked Questions

Q: Why isn't the flag comparing my employee to others with the same exact title?

A: The flag uses regression modeling across your entire population rather than direct peer matching. This approach captures more nuanced patterns in how your organization compensates employees and works even when there aren't many employees with identical attributes.

Q: I have an employee with a unique role. Will the flag work for them?

A: Yes. Unlike peer-to-peer comparison, regression modeling doesn't require multiple employees with identical attributes. It predicts compensation based on patterns across your entire organization.

Q: An employee is flagged as an outlier, but I know their pay is correct. What should I do?

A: Outliers aren't errors—they're signals to investigate. Document the business justification (new hire market adjustment, retention counteroffer, specialized skills, etc.) and proceed with your planned compensation.

Q: Should I address every flagged outlier?

A: Not necessarily. The flag surfaces deviations from predicted patterns, but legitimate business reasons may explain those deviations. Use the flag as a starting point for conversations and investigation, not as an automatic mandate for adjustment.