It is intended for analysts or statisticians who need to understand where the pay gap arises, which positions or employees drive it and what must be corrected, explained or recalculated. The section also supports the legal reporting requirement for differences between worker categories by base and variable pay components.
Chart 1: Gender pay gap by job position
Controls above the chart
Control | Values | Function |
View | By position / By work group | Aggregates the chart by position or by grade group. |
Detail | Off / On | Switches from chart to employee-level table. |
Info | Tooltip | Shows methodology explanation. |
Scatter plot
The default scatter plot shows average pay by position. The Y-axis shows average salary and the X-axis shows job positions. For each position, two points may be shown: one for men and one for women. Positions represented by only one gender show only one point and cannot be evaluated in a gender comparison.
The chart should be read by looking for positions where the male and female points on the same vertical line differ substantially. Such positions may require deeper investigation.
Detail mode — identifying specific employees
When Detail mode is turned on, the chart is hidden and replaced by a search field and table. The table shows the position and employee ID together with the employee’s contribution to the total pay gap in percentage points.
This is the direct channel from aggregate result to action. If the scatter plot indicates a problem in a position such as Logistician or Commercial Clerk, the analyst can search for that position and identify the specific employees driving the gap.
Chart 2: Pay gap by work group — core debugging workflow
This chart is shown when the view is switched from By position to By work group. It is the most important chart for iterative debugging of the grading model.
Sub-controls above the chart
Control | Values | Meaning |
Method | Regression / Average | Uses adjusted regression result or simple average. |
Work group grading scenario | All scenarios or system 1, 2, 3 | Selects which grading system is shown. |
Pay component | All pay / Base pay / Variable pay | Filters the gap by salary component. |
The chart allows the user to identify which grade has the largest adjusted or unadjusted gap, and whether the issue comes from base pay or variable pay.
Iterative debugging workflow
Identify a problematic grade in the chart.
Switch pay component from all pay to base pay or variable pay to determine the source of the gap.
Return to Input data > Manage data.
Filter the dataset to the problematic grade using the relevant Level 1 / 2 / 3 column.
Sort by the relevant pay component and look for extremes.
Decide whether each anomaly is a classification error or an explainable exception.
If incorrect, correct the grade or pay component, click Update database and run the analysis again.
If explainable, document the explanation in Notes for audit purposes.
The same process is applied separately to variable pay. Large positive or negative variable-pay differences often come from bonuses, allowances, exceptional payments or missing employee-contribution information.
Explaining or correcting findings
For every problematic employee or grade, the client has two options:
Explain — add a justification in Notes, for example exceptional project bonus, scarcity skill or documented performance reason.
Correct — fix a data error such as wrong grade, missing contribution information or incorrect pay value, then recalculate.
Employee-level table
The expert table can show gaps at the level of specific employee IDs, both by position and by grade. It can include total gap, base-pay gap and variable-pay gap. Negative values mean the employee earns less than expected by the regression model given their qualification, seniority and position.
The same employee may look like an extreme anomaly at position level but normal at grade level, because a grade contains more people and dilutes individual deviations. This is why both views are needed for interpretation and reporting.
Position versus grade
The position view is narrow and sensitive to individual outliers. It is useful for identifying concrete issues in a specific job role. The grade view is broader and closer to regulatory reporting by worker category. A robust explanation should consider both.
Blinder-Oaxaca decomposition
The Expert analysis can also contain decomposition of the pay gap into explained and unexplained parts. The explained part is attributable to observable structural factors such as education, experience, position, seniority or grading. The unexplained part remains after controlling for those factors and is the key candidate for deeper investigation.
Raw data
The Raw data section shows the internal variables used by the regression model. These data are visible and downloadable and, together with the automatic e-mail sent after each run, form the audit trail.
Internal column | Example / meaning |
Id | Employee identifier. |
Age | Age calculated from source data. |
Year Of Service | Years of service, for example 16.00. |
Years In Education | Education years, for example 8. |
Years Of Experience | Experience years. |
Years Of Experience Squared | Squared term for non-linear experience effect. |
Employee Scarcity | 0 or 1 scarcity indicator. |
Employee Evaluation | Performance evaluation. |
Sex | Gender variable. |
Skill / Skill2 / Skill3 | Level 1 / 2 / 3 grading systems. |
Seniority | Seniority level. |
Employee Contribution | Pension or employee contribution indicator. |
Log Adjusted Net Income | Logarithm of adjusted total income. |
Log Adjusted Basic Income | Logarithm of adjusted base income. |
Log Adjusted Variable Income | Logarithm of adjusted variable income. |
Average Salary | Average salary value. |
Basic Income Average Adjusted | Adjusted average base income. |
Bonuses Average / Personal Bonuses Average | Bonus components used in the model. |
Logarithmic variables
The regression works with logarithms of pay values rather than nominal values because:
Logarithms linearise the relationship between pay and explanatory variables such as education and experience.
The coefficient of gender in a log-regression can be interpreted directly as a percentage difference after multiplication by 100.
Residuals of the log model are usually closer to a normal distribution.
Audit trail
The Expert analysis, raw data export and automatic e-mail together create a complete audit trail: what data were used, which parameters were selected, what results were produced and which employees or grades require explanation.