Outlier Detection helps you identify groups whose overall review ratings differ significantly from the rest of the review cycle. Instead of manually analyzing spreadsheets, you can quickly spot unusual rating patterns across departments, locations, tenure groups, teams, and other employee segments.
Outliers can help uncover potential fairness concerns, highlight exceptional performance trends, and guide deeper review discussions. An outlier is a signal to investigate further, not a conclusion.
Where to Find Outlier Detection
Open a Performance Review cycle.
Select the Outliers tab.
Review any groups that have been identified as statistically significant outliers.
Each outlier card includes:
The group that was flagged
The number of employees in the group
A summary of how the group's ratings differ from the company average
A visual comparison of the group's rating distribution against the overall review cycle
To explore the underlying data, select Show in Table.
Tip: Review Outlier Detection before calibration discussions to quickly identify groups that may benefit from additional review and context.
How Outliers Are Detected
Outlier Detection compares the distribution of overall ratings within a group with the distribution across the entire review cycle.
Rather than looking only at averages, the detector evaluates whether the overall pattern of ratings is unusual enough to be unlikely to have occurred by chance.
Groups are flagged when they meet a 95% confidence threshold (alpha = 0.05).
In practical terms, this means the system highlights groups only when there is strong evidence that their rating patterns differ from the overall population.
How Is This Calculated?
The detector compares each eligible group to the entire review cycle.
For example, if a department contains six employees with unusually high overall ratings, the system asks: if six employees were selected at random from the entire review cycle, how often would a result this unusual occur?
If a similar result happens less than 5% of the time, the group is flagged as an outlier.
The analysis considers the full distribution of ratings, not just the average score. This means a group can be identified as an outlier even when its average rating appears similar to the company average, if the overall rating pattern is unusually different.
What Data Is Analyzed?
The detector evaluates employee groups based on available employee attributes, including:
Department
Team
Location
Gender
Tenure
Additional employee attributes may also be included when available in your account.
Minimum Group Size Requirements
To improve reliability, only groups with at least five employees are included in the analysis.
Very small groups can produce misleading results because a single rating may have a large impact on the overall outcome.
Note: The minimum group size is fixed and cannot be changed.
Understanding the Results
An outlier does not automatically indicate bias or a problem. A flagged group simply means its rating pattern differs significantly from the overall review cycle.
For example:
A high-rating outlier may reflect exceptional team performance or indicate a manager who tends to rate more generously.
A low-rating outlier may reflect challenging business conditions or indicate unusually strict rating behavior.
Outlier Detection provides statistical signals. Human review and organizational context remain essential when interpreting the results.
Tip: Review outlier findings alongside qualitative feedback, manager comments, and business context before taking action.
No Outliers Detected
If you see a message indicating that no outliers were detected, this typically means one of two things.
Rating patterns are consistent. Reviewers may be applying ratings consistently across the organization, resulting in no statistically significant differences between groups.
Groups are too small. Some employee groups may not meet the minimum size requirement for analysis. Groups with fewer than five employees are excluded from Outlier Detection.
Best Practices
Use Outlier Detection as a starting point for investigation, not as a final conclusion.
Review flagged groups alongside qualitative feedback and business context.
Compare results across multiple review cycles to identify recurring patterns.
Follow up when unusual rating distributions appear repeatedly.
Consider both high and low outliers, as each can provide valuable insight.
Frequently asked questions
What exactly is an Outlier?
What exactly is an Outlier?
An outlier is a group whose overall rating pattern differs from what would normally be expected when compared to the rest of the review cycle. This does not mean the group's ratings are incorrect. It simply means the pattern is unusual enough to warrant a closer look.
Does an outlier mean there is bias?
Does an outlier mean there is bias?
No. An outlier indicates a statistically unusual pattern, not proof of bias. Additional review and organizational context are required before drawing conclusions.
Does the detector only look at average ratings?
Does the detector only look at average ratings?
No. The detector evaluates the full distribution of ratings, not just the average score. This allows it to identify unusual patterns even when a group's average appears similar to the company average.
Can I include smaller groups in the analysis?
Can I include smaller groups in the analysis?
No. Groups must contain at least five employees to be included. This threshold helps ensure results are reliable and meaningful, rather than being driven by the ratings of only a few individuals.
Is the statistical threshold configurable?
Is the statistical threshold configurable?
No. Outlier Detection uses a fixed confidence level of 95% (alpha = 0.05) for all analyses.