What salary anomalies are and why they matter
A salary anomaly is an individual compensation data point that deviates significantly from the expected range for a given combination of role, seniority level, location and tenure. In a well-managed compensation structure, employees in equivalent positions with similar experience and performance records should earn within a predictable range of each other. When an individual's salary falls materially outside that range — whether above or below — an anomaly exists.
Salary anomalies matter for several distinct reasons. On the underpayment side, they represent a retention risk: employees who are significantly below the market rate for their role are the most likely to respond positively to recruiter outreach, to actively search for alternatives, and to accept offers from competitors. Research consistently shows that compensation dissatisfaction is the leading stated reason for voluntary departure in competitive labour markets. Employees who know — or suspect — they are underpaid are already partway to the exit.
On the overpayment side, anomalies represent a budget management problem. An employee paid significantly above market for their role is not necessarily performing better than peers at the market rate — in many cases, overpayment results from historical decisions, retention bonuses that are no longer warranted or legacy structures from acquisitions. Uncorrected overpayment compounds over time through merit increases applied to an already inflated base.
There is also an equity dimension. Salary anomalies that fall along demographic lines — where employees in a protected category are systematically paid below the median for their peer group — represent both legal and moral risk. Proactive anomaly detection is an essential component of any pay equity programme, since systemic underpayment by gender, ethnicity or other characteristics rarely announces itself through obvious data patterns; it has to be specifically looked for.
The Underpayment Flight Risk
Employees who know they are underpaid — and they often do know, through recruiter conversations, published salary surveys and peer discussions — are substantially more likely to be actively job-searching than those who feel equitably compensated. The risk is highest for employees in roles where external market data is transparent: technical roles with visible salary benchmarks, roles with active recruiter networks, and any position where competitors routinely share offer details to attract candidates. HR teams that detect underpayment anomalies and act on them proactively retain employees who might otherwise have been lost to a competitor with a straightforward competitive offer.
Types of salary anomalies: underpayment, overpayment and inconsistency
Not all salary anomalies have the same cause or the same remedy. Understanding the type of anomaly being dealt with is the first step to addressing it appropriately.
Underpayment anomalies occur when an employee is paid materially below the market median or the internal salary band for their role and level. The most common causes are: the employee was hired during a period when the organisation's salary offers were below current market rates; the employee has been in role for several years with only modest merit increases while the market moved faster; the employee negotiated weakly at hire and the initial anchor has persisted; or the employee was promoted into a role without a full market adjustment to the new level's compensation range.
Overpayment anomalies occur when an employee is paid materially above the market median or above the top of the salary band for their role and level. Common causes include: a retention bonus or counter-offer that elevated the salary above a sustainable long-term level; the employee's role was reclassified downward while their salary was protected; a legacy arrangement from a company acquisition where compensation harmonisation has not been completed; or an exceptional performance-based increase that has not been followed by corresponding career progression.
Inconsistency anomalies are perhaps the most revealing type: employees in identical or near-identical roles with similar tenure and performance records earning materially different salaries. These inconsistencies often reflect the accumulated effect of different hiring managers using different negotiation practices, different recruiters applying different discount or premium judgements to offers, or different managers applying merit increases with different philosophies. Inconsistency anomalies are the most directly actionable from an equity standpoint, since they reveal internal inequity that is difficult to justify on any defensible basis.
How to detect salary anomalies in workforce data
The starting point for anomaly detection is a clean, consistent dataset: every employee's current base salary, their role title standardised to a common taxonomy, their seniority level, their location, and their tenure in role and in the organisation. Without role standardisation — where "Senior Engineer" might appear as ten different variants across hiring managers and departments — the analysis will be fragmented and unreliable.
Once the dataset is clean, the most direct detection method is peer group comparison: group employees by the same role and level, calculate the median and interquartile range for each group, and flag anyone who falls more than a defined percentage outside the expected range. The percentage threshold should be calibrated to the role type: tight bands with low variation might flag deviations of more than 15%, while leadership or highly variable specialist roles might use a 25-30% threshold.
Location adjustment is critical for organisations with employees in multiple geographies. An employee in a high cost-of-living city should be compared against the adjusted median for that location, not against a national average. Failing to apply location adjustments will systematically flag all employees in expensive markets as overpaid and all employees in lower-cost markets as underpaid — obscuring the genuine anomalies.
For organisations with access to market benchmark data — from published salary surveys, compensation data providers or industry associations — the analysis can be extended to compare internal salaries against external benchmarks. This adds a second dimension: an employee may be at the median internally but 30% below the external market for their role. Both pieces of information are relevant, and only the combination reveals the complete picture.
Automated Salary Outlier Detection in Treegarden
Treegarden's compensation analytics module automatically flags employees whose salaries deviate more than 20% above or below the median for their role and level combination. The detection runs continuously against live HR data — so when a new hire is added, when a salary change is processed or when the benchmark data is refreshed, any resulting anomalies are surfaced immediately. HR leaders see a prioritised list of flagged employees with the deviation amount and severity, enabling rapid triage rather than a periodic manual audit.
Statistical approaches: percentile analysis and regression
Beyond simple peer group comparison, two statistical methods are particularly useful for salary anomaly detection in larger workforces: percentile analysis and regression-based expected salary modelling.
Percentile analysis extends the peer group comparison approach by expressing each employee's salary as a percentile within their peer group rather than just a deviation from the median. An employee at the 5th percentile of their peer group is almost certainly underpaid, regardless of the specific dollar or euro amount involved. An employee at the 95th percentile warrants investigation. This approach is intuitive, communicable to non-technical stakeholders and directly actionable: the remediation target is to bring employees below the 25th percentile to at least the 25th percentile of their peer group, restoring them to an equitable position within the distribution.
Regression-based modelling is more sophisticated and appropriate for larger datasets. The approach builds a statistical model that predicts expected salary based on observable factors: role, level, location, tenure, performance rating and any other legitimate salary determinants. Each employee's actual salary is then compared against their model-predicted salary. Large positive residuals (actual significantly above predicted) flag overpayment; large negative residuals flag underpayment. The advantage of regression is that it controls for all the legitimate factors simultaneously — it will not flag a high-performing senior employee in an expensive city as overpaid just because their absolute salary is high. It flags cases where the salary cannot be explained by the legitimate factors, which is a much more precise signal.
Regression analysis also enables pay equity analysis at scale: by adding a demographic variable to the model and examining whether it is a statistically significant predictor of salary after controlling for legitimate factors, organisations can determine whether systematic gender or ethnicity-based pay gaps exist in their data. This is the standard methodology used in external pay equity audits, and running it internally provides early warning before an external audit reveals it.
Overpayment Is Harder to Fix Than Underpayment
When an underpayment anomaly is confirmed, the remedy is straightforward: increase the salary to bring the employee within the expected range. When an overpayment anomaly is confirmed, the options are far more constrained. In most jurisdictions, reducing an employee's salary unilaterally constitutes a breach of contract and is legally impermissible without the employee's agreement. Practically, an employee earning above market is unlikely to agree to a reduction. The realistic options for managing overpayment anomalies are: freeze merit increases for the employee until the market catches up to their salary; restructure their role to match the responsibilities to the compensation level; or manage progression upward so that their responsibilities eventually justify their pay. Overpayment management is therefore a multi-year process, not a quick correction.
Investigating flagged anomalies: legitimate reasons versus problems
Automated anomaly detection produces a flag, not a verdict. Every flagged employee requires a manual investigation to determine whether the anomaly reflects a genuine problem or has a legitimate explanation that the detection algorithm cannot see.
The investigation process should start with the employee's compensation history: when was the last salary change, what was the reason, and who approved it? A salary significantly above the band midpoint may reflect a documented retention bonus, an approved market adjustment for a hard-to-fill specialisation, or a legacy arrangement explicitly agreed with senior leadership. Any of these is a legitimate explanation — but it should be documented, not just informally understood by whoever approved it at the time.
For underpayment anomalies, the investigation should also check whether the employee's role title reflects their actual responsibilities. Misclassification — where an employee is performing at a higher level than their title indicates — is a common cause of apparent underpayment. If someone is doing senior-level work at a mid-level title and salary, the anomaly is real, but the remedy is a reclassification and promotion adjustment rather than simply a salary increase within the current title.
The investigation output should be one of three conclusions: the anomaly is a confirmed problem requiring remediation; the anomaly has a legitimate documented explanation and requires no action other than noting the reason; or the anomaly requires further investigation — for example, if there is conflicting information about whether a documented arrangement was formally approved. This classification ensures that flagged anomalies are resolved rather than accumulating as an unreviewed backlog.
Compensation Data Visualisation in Treegarden
Treegarden's compensation visualisation tools present salary distribution by department, role and tenure through interactive charts, making it straightforward to identify clustering patterns and distribution gaps that indicate systemic anomalies rather than individual outliers. A team where all employees cluster at the top of the salary band suggests a structural problem with the band definition. A team with a bimodal distribution — clustered at the low and high ends with few employees in the middle — suggests a historical discontinuity in hiring or merit practices. Visual distribution analysis surfaces these patterns in seconds rather than through spreadsheet inspection.
Remediation: what to do once you have found an anomaly
Once an anomaly is confirmed as a genuine problem, the remediation plan needs to address three questions: what is the target salary, when will the adjustment be made, and how will it be communicated to the employee?
For underpayment remediation, the target salary should be set at the appropriate point within the salary band for the employee's role and level — typically the band midpoint or the 25th percentile for an employee performing at standard level, with higher positioning for high performers. The adjustment should not simply bring the employee to the minimum of the expected range, since this produces an adjustment that is technically compliant but still leaves the employee at the bottom of the distribution where further drift could quickly return them to anomaly status.
Timing matters significantly. For underpaid employees who represent flight risks — those who have been flagged by performance as high value and whose roles are difficult to replace — the adjustment should be made immediately, outside the normal merit review cycle. Waiting for the next annual review cycle to address a known underpayment anomaly is a retention risk that the organisation is consciously accepting. For less urgent cases, incorporation into the next merit review cycle may be appropriate, provided the review cycle is within a few months.
Communication should be transparent about the process that led to the adjustment without disclosing peer salaries. The employee should understand that a review was conducted, that their salary was found to be below the appropriate range for their role and experience, and that the adjustment corrects this. This explanation reinforces the organisation's commitment to fair compensation and is more trust-building than an adjustment that arrives without explanation.
Pay Decision Audit Trail in Treegarden
Every salary decision in Treegarden is logged with a timestamp, the approving manager, the reason code and any supporting notes — creating an immutable audit trail that enables retrospective analysis of how anomalies developed. When an anomaly is detected and investigated, the HR team can immediately see the history of decisions that produced the current state: which merit increase rounds were missed, which adjustments were applied and which were deferred, and whether the anomaly was identified and noted in a previous review without action being taken. This historical context is essential for designing remediation that addresses the root cause rather than just the current symptom.
Building ongoing anomaly monitoring into HR processes
A one-time salary anomaly analysis is a snapshot, not a programme. The conditions that create anomalies — market movement, hiring decisions, merit increase cycles — are continuous processes. An effective anomaly monitoring programme must also be continuous rather than episodic.
The most efficient approach is to embed anomaly checks into existing HR process trigger points. When a new hire is added to the HR system, the system should automatically compare the new hire's salary against existing employees in the same role and level. When a salary change is processed, the system should check the resulting position against the band and against peer salaries. When market benchmark data is updated — which external data providers typically do annually — the system should re-run the comparison and surface any employees who have fallen below the new benchmark thresholds.
Regular reporting cadence is also important. A monthly or quarterly compensation anomaly report, reviewed by the HR Director or CHRO, ensures that flagged cases are being investigated and resolved rather than accumulating. The report should show not just new anomalies but the status of previously flagged cases: how many have been resolved, how many are under investigation, and how many have been noted with a legitimate explanation. This tracking converts anomaly detection from a one-time exercise into an ongoing governance process.
Finally, anomaly monitoring should be integrated with retention risk management. Employees who are both underpaid and identified as high performers by the performance management system are the highest-priority risk combination. These individuals represent disproportionate value to the organisation and are simultaneously the most likely to leave if the underpayment is not corrected. Cross-referencing anomaly data with performance data and flight risk indicators produces a prioritised action list that HR can act on proactively.
Frequently asked questions about salary anomaly detection
What counts as a salary anomaly in HR analytics?
A salary anomaly is any individual compensation data point that deviates significantly from the expected range for a given role, seniority level, location and tenure combination. In practice, most HR analytics tools define anomalies using a threshold — typically more than 20% above or below the median salary for a peer group with equivalent characteristics. However, the appropriate threshold varies by role type and salary range: in roles with tight pay bands, a 10% deviation may be significant, while in senior leadership positions with more variable compensation, a 25% threshold may be more appropriate.
How often should HR teams run salary anomaly analysis?
For organisations without automated anomaly detection, a structured salary anomaly analysis should be run at minimum quarterly — more frequently during periods of rapid hiring or significant merit increase cycles. For organisations using HR software with continuous monitoring, anomalies should be flagged as they emerge: when a new hire is added, when a salary change is processed, or when market benchmark data is updated. Annual-only reviews are insufficient for organisations experiencing meaningful headcount growth, since anomalies introduced through new hires accumulate throughout the year before being detected.
What is the difference between a salary anomaly and pay compression?
Pay compression is a systemic pattern affecting a cohort of employees — typically those with longer tenure whose salaries have been outpaced by new hire rates. A salary anomaly is an individual data point that falls outside the expected range for its peer group, which may or may not be part of a compression pattern. An underpaid senior employee is both a compression symptom and a salary anomaly. An overpaid junior employee may be a salary anomaly without contributing to compression. Both require attention, but compression is addressed through structural policy changes while individual anomalies may be addressed case by case.
What should HR do when an anomaly investigation reveals a legitimate explanation?
Not every salary anomaly represents an error or inequity. Legitimate explanations include: a specialised certification or skill set that commands a market premium; a retention arrangement agreed to prevent the loss of a critical employee; a geographic cost-of-living adjustment for a remote employee in a high-cost location; or a legacy arrangement from a company acquisition where salary harmonisation has not yet been completed. When an anomaly has a legitimate explanation, it should be documented in the employee record with the reason and the review date. This documentation prevents the same anomaly from being repeatedly flagged without context, and creates a clear record of what was considered and why no action was taken.