An issue was corrected where the task runner component that computes estimated adjusted pay gaps did not respect the residual method feature flag configuration. This caused inconsistent calculation results when the log_residual approach was configured for the instance. The task runner now uses the method configured for the instance, ensuring consistent estimated adjusted pay gap calculations across all platform components.
PA-7912
The Pay Transparency report and Pay Explainability report now automatically identify and apply the correct sub-group context when generated for an employee belonging to a sub-grouped analysis. When HR analysts or Compensation managers generate a report from a parent analysis that includes sub-groups, the system now scopes the average values and comparisons to the employee's specific sub-group rather than using dataset-wide aggregates, ensuring consistency with the sub-group-level results shown in the analysis view. For employees whose sub-group does not have a corresponding analysis (due to small segment size), calculations are restricted to that sub-group and the Analytics Deep-Dive section displays a message indicating that no statistically meaningful analysis is available for that employee.
PA-7941
Web-rendered reports now generate in the logged-in user's selected language rather than the tenant's system-wide language setting. All built-in report types — including Overview, By Category, Outlier, Universal Fair Pay, EU Directive, UK, Ireland, Japan, and Empty reports — are affected. When a saved report is reopened, it displays in the language it was generated in, regardless of the current viewer's language or the tenant's system language. Users whose language preference differs from the tenant's system language will now see report labels and text in their own language.
PA-7942
The Gender Equality Index report is now available to users on all seven supported platform languages. Previously, users with platform languages other than English or French were shown an 'unsupported language' message and could not generate the report. This restriction has been removed, and the report can now be generated and exported regardless of the user's selected platform language.
PA-7944
Country names embedded in report titles (e.g. 'Remuneration register (Spain)', 'Gender Equality Index (France)') were extracted into independent translation keys so that both the report name and the country can be translated separately. Layout issues caused by longer translated strings were corrected across the report creation picker, export panels, and printed/PDF output—containers that previously caused labels to clip, truncate, overflow, or overlap adjacent elements now accommodate translated text properly.
PA-7971
The EEO-1 and California regulatory report exports now respect the user's selected language setting. Previously, these Excel exports always displayed worksheet names, column headers, job category labels, and race/ethnicity labels in English regardless of the application language. Exported files now display labels and headers in the appropriate locale, ensuring consistency with the rest of the interface. English translations are currently available; users on other locales will see English labels via fallback until additional translations are added.
PA-8011
An issue was corrected where the 'target reached' indicator in the Pay Gaps by Groups table displayed incorrect results when users manually selected a reference group (e.g., 'vs. Female') that differed from the system's automatically chosen reference, specifically affecting demographic columns with three or more values (e.g., Male/Female/Non-binary). The issue occurred because the indicator was evaluated against the user's selected viewing direction rather than the canonical direction used by the underlying pay gap calculation. The indicator now correctly evaluates against the canonical direction in all cases, ensuring consistent and accurate target-reached status regardless of which reference group is selected for display.
PA-8028
An issue in the raise suggestion algorithm was corrected where employees who did not require raises were incorrectly allocated maximum raises, and the within-group gender pay gap overshot the configured target. Two related defects were identified: when an employee's salary fell far enough below predicted value that the outlier floor mandated a raise exceeding the configured cap, the algorithm's internal measurement incorrectly used the raw cap as the ceiling while the actual pipeline applied the higher outlier floor value. This mismatch caused the system to misclassify demographic groups as unclosable and allocate unnecessary cap-level raises to multiple employees in the same group, inflating remediation costs and causing the pay gap to land on the wrong side of the target. In the reported case, a within-group gap switched from 9.5% women lower to 5.5% women higher, past a 4.9% target. The raise suggestion algorithm now correctly applies the maximum of the configured cap and the outlier floor requirement when evaluating closeability and adjusting raises, ensuring that outlier corrections are properly factored into pay gap calculations.
PA-8032
The fields-metadata API endpoint (which returns aggregate statistics such as distinct values, counts, and summary statistics for dataset fields) now accepts filter parameters to scope results to a subset of records, mirroring the filtering mechanism already available on the dataset-metadata endpoint. When a filter is applied, all returned aggregates—distinct values and counts for text fields, summary statistics for numerical fields, and intersection breakdowns—reflect only the records matching that filter. For users without per-record-details permission, the existing minimum-group-size privacy guard now extends to this endpoint: if a filter narrows results below the configured threshold (currently 6 records), aggregates are suppressed to prevent inference of individual employee details.
PA-8040
The frontend fields-metadata layer (the internal service that fetches employee field statistics and aggregates for use in reports, wizards, and data configuration screens) was extended to accept an optional filter parameter and to expose a suppressed privacy-guard flag. This enables future features to request field statistics for a specific subset of employees in a single round-trip and to display appropriate messaging when results are suppressed due to small group sizes. No user-facing functionality changes in this release; existing screens that display field metadata continue to behave as before.
PA-8050
Documentation was added to the repository to codify the Linear pilot rules during the trial period. A LINEAR_PILOT.md file was created at the repository root specifying that Jira PA-XXXX identifiers should be used in commit messages, PR titles, and branch names, and that the Linear MCP server (a tool for programmatic ticket operations) should be used for ticket operations during the pilot. A cross-reference to this documentation was added to the existing CLAUDE.md file under the Commits & Pull Requests section.
PA-8054
The AskBee chatbot backend address was made environment-configurable in the frontend. Previously, all environments (development, staging, and production) communicated with the same chatbot instance, preventing proper isolation between environments. The backend address can now be set independently for each deployment environment through configuration rather than requiring source code modifications and a full release cycle.
PA-8058
A type alias ResidualMethod was introduced in the API codebase to replace repeated use of Literal type declarations for residual calculation methods (a Python typing construct that defines allowed string values). This refactoring improves code maintainability without changing any platform behavior.
PA-8059
Two new dropdown components (PaSelect and PaCombobox) were added to the design system to support rich content inside dropdown options, such as icons, multi-line labels, badges, and prefix text. The native HTML select component supports only plain text, limiting UI flexibility for features such as country pickers with ISO code prefixes or employee selectors with additional metadata. The new components include an automatic search/filter input for lists with six or more items, keyboard navigation, and full accessibility support including screen reader announcements for validation states.
PA-8062
An issue was corrected where the remediation process would hang indefinitely and fail to complete when running analyses with a large number of business object interaction terms. A performance improvement was implemented using a greedy optimization method to handle these regression scenarios. Analyses without remediation continued to run successfully throughout.
PA-8063
The analysis gender demographic reference group field was added to the analysis-results-simple API endpoint. Clients consuming raise impact data from this endpoint can now correctly interpret results without requiring a separate call to the analysis-results endpoint.
Release v2026-05-04_01
Type
Jira
Description
PA-7670
An issue was corrected where the Cancel button on the Run Analysis screen did not respond when clicked after scrolling down the page, preventing users from returning to their originating view. This occurred when users entered the Run Analysis workflow from a Reports view tied to a specific report. The Cancel button now correctly returns users to the page from which they initiated the Run Analysis workflow, preserving the previously selected report or dataset.
PA-7679
An issue was corrected where attempting to rename a dataset or analysis with an excessively long string displayed a generic "Operation failed" error message. The rename dialog and analysis rename controls now show an inline validation message directly below the input field when the name exceeds the character limit, making the constraint clear to users at the point of entry.
PA-7727
An issue was corrected in which the task runner component that computes the Estimated Adjusted Pay Gap did not respect the residual method feature flag configuration. This caused the task runner to use a different calculation method than other components in the platform when the log_residual approach was configured. The task runner now correctly applies the method configured for the instance, ensuring consistent adjusted pay gap calculations across all components.
PA-7791
An issue was corrected where analysis results were missing in the Subgroup analysis view when the grouping column contained only one distinct value and the dataset was configured without gender data. The Results section (showing compensation totals, number of raises, and related metrics) now correctly displays for single-group subgroup analyses instead of rendering as a blank area.
PA-7882
An issue was corrected where running a pay equity analysis on datasets with 4 or fewer employees resulted in a server error: "There was a problem running the analysis. An error occurred on the server." This issue also affected individual subgroups within a subgroup analysis when a subgroup fell below the minimum headcount threshold. The analysis now completes gracefully without regression results instead of failing with a server error.
PA-7911
An issue was corrected where compensation components in the Employee's group compensation table on the Right-to-Request report did not display in the same order as the columns in the uploaded Excel dataset. Aggregate totals (such as Total Target Compensation) were appearing before individual compensation components instead of after them. The table now displays compensation components in the order they appear in the source data, with aggregate fields positioned after individual components and salary-after-raise columns at the end.
PA-7915
Confirmation popup is visually cut off and triggers auto - confirmation without user intervention in Pay Bands.
PA-7979
An issue was corrected in the Pay Gaps by Groups summary where the "Reason target not reached" column displayed incorrect or misleading messages. In cases where the reference group was the higher-paid group but fell below the minimum size threshold to receive raises, the system incorrectly showed "Max limits reached" instead of accurately indicating that the higher-paid group was too small. Additionally, the tooltip displaying the group name and the reason was sometimes suppressed for groups with a swapped reference. The "Reason target not reached" column and its tooltip now display the correct reason and group information for all scenarios.
PA-7998
An issue in the Pay Gaps by Groups table where the pay gap target achievement status was calculated using the wrong reference group was corrected. The system was incorrectly using the system reference group instead of the user-selected custom reference group when determining whether the target was reached, resulting in misleading "Target reached" indicators and tooltip content. The target achievement logic now correctly evaluates against the appropriate reference group, and the informational tooltip on the "Target reached" indicator now displays accurate gap-versus-target summaries. When a custom reference group is selected that differs from the system default, the tooltip shows a two-row comparison with correct higher/lower labels.
PA-8007
MinIO binary download fails in CI due to redirect not being followed
PA-8008
Close-by-group e2e test times out waiting for grouping column dropdown
PA-8010
An issue was corrected where applying a checkbox filter with a blank value on a numeric column in the Compensation Assistant caused a 500 error and prevented the comparison group record count from updating. The issue occurred when a saved preset was deselected, which reset filter values to empty strings instead of null; those empty strings were then sent as filter parameters when attempting to filter numeric columns. The Compensation Assistant now correctly resets filter state to null, and the filter is handled gracefully.
