Known Limitations

Limitation: On the Analytical Engine the ALL function overrides filters only for measures, not for slicer/dimension subqueries, so it does not fully remove the filtered table from the query path the way the original translator did.

Workaround: Use Measured Values or the Filtered Measure add-on instead of the ALL function.

Limitation: After switching to the new translator/AE, null measure values display as blank rather than 0, because in the AE null + 0 evaluates to null.

Workaround: Wrap the measure with ISNULL, e.g., If(ISNULL(sum(deals)),0,sum(deals)).

Limitation: Certain formulas using the Round function with filters fail on the Analytical Engine with an ElastiCube 'too many digits' error; the AE falls back to the original translator for these cases.

Workaround: Rely on the automatic fallback to the original translator (OT) for the affected Round-with-filter formulas.

Limitation: A syntactically valid formula can fail on the AE ('operator is null') and fall back to the original translator OT depending on parenthesis placement, while a logically equivalent parenthesization runs.

Workaround: Rewrite the formula with the parenthesis arrangement the AE accepts.

Limitation: Dashboards sending Main, Headers, and Grand-Total JAQL concurrently can fail in the Analytical Engine with a MonetDB MALException due to unsupported generated SQL.

Workaround: Fall back to the original translator (OT) / compatibility mode for the affected widget.

Limitation: A pivot reusing a very complex formula across multiple metrics can be substantially slower on the new Analytical Engine than on the legacy engine.

Workaround: Manually fall back to the original translator (OT) for the affected widget.

Limitation: On Live models, the RANK function combined with ALL always returns 1; the Analytical Engine does not support Measured Value over Window functions (RANK/RSUM).

Workaround: Use an ElastiCube in compatibility mode, where RANK with ALL returns the correct result.

Limitation: When a cube's Analytical Engine setting is AE-only, YTD formulas (YTDAVG/YTDSUM with ALL) return errors because XTD / Measured Value over window functions are unsupported.

Workaround: Set the cube to Compatibility mode, where the YTD formulas return data.

Limitation: With the Analytical Engine, conditional statements such as ISNULL must be wrapped in an aggregation, otherwise the widget fails with “Unsupported operator.” (The reported formula also used an invalid measured-value structure.)

Workaround: Wrap the conditional in an aggregation per the conditional-statements documentation, or switch the cube to Compatibility Mode.

Limitation: After upgrading to 2026.1.1, a widget filter can return different results when the Analytical Engine is used versus the original translator. The Analytical Engine output is the correct, by-design behavior.

Workaround: If legacy results are required, set the translation strategy to OLD_ONLY (Compatibility Mode); otherwise rely on the Analytical Engine results.

Limitation: New Data Groups default to an 8 GB memory cap even on larger machines (e.g., 64 GB RAM), which can cause ElastiCube build failures on fresh instances until adjusted.

Workaround: Increase the Data Group memory in its settings to match the available machine resources.

Limitation: On a Live datasource, including filterRelations in a JAQL query causes count/offset pagination to be ignored, returning the full dataset instead of the requested page.

Workaround: Avoid filterRelations where pagination is required, or apply row limiting another way.

Limitation: For a chart without a category, CSDK mode honors the widget’s “stacking” configuration (stacking multiple values) while classic Fusion ignores it, so the same widget can look different on the dashboard in CSDK mode.

Workaround: Adjust the widget’s stacking configuration to get the desired rendering; CSDK follows the stacking setting by design.

Limitation: After changing the Default Database in Connection Management, the data model UI may continue to display the old database name even after updating the connection.

Workaround: Manually update the database in each model via Change Database; the default database only defines the default and does not block access to other databases.

Limitation: Custom-code notebooks that call internal Sisense APIs using a session cookie fail to build once that cookie expires, because the cookie is not auto-regenerated.

Workaround: Author the notebook to obtain a fresh token rather than relying on a session cookie.

Limitation: After upgrading (L2025.3 → L2025.4) on systems where widget deltas are enabled, dashboards can display with no widgets even though the underlying widget records still exist, when the widget deltas were not fully populated during the migration.

Workaround: Ensure widget deltas are fully populated during/after migration; contact Support to restore widget rendering.

Limitation: Columns defined as TIMESTAMP_TZ (timestamp with timezone) are not supported. On a chart, values with different timezone offsets that fall on the same calendar day appear as duplicate dates.

Workaround: Create a custom column that casts the TIMESTAMP_TZ column to a timestamp without timezone.

Limitation: With a Live model and Continuous date display, a Days-type date column in a Table widget can show a date one day earlier than the dashboard-filter selection. The root cause is the same TIMESTAMP_TZ limitation; it does not occur with Date & Time type or in a Pivot.

Workaround: Use Date & Time display, or cast the TIMESTAMP_TZ column to a timestamp without timezone.

Limitation: Smart Matching requires build-generated statistics, which Live models do not have, so it cannot be enabled on a Live model.

Workaround: Use an ElastiCube model where a build generates the required statistics.

Limitation: With Model Statistics enabled, the build finalization phase can hang for hours with database connection timeouts when the data contains invalid numeric (NaN) values.

Workaround: Disable Model Statistics for the cube, or remove the invalid (NaN) values from the data.

Limitation: POST /elasticubes/live/{title}/datasecurity creates a new data-security rule on every call, so repeated calls produce what look like duplicate rules and the UI may show 'access to nothing'.

Workaround: Use GET and PUT instead of POST.

Limitation: When both an 'Everything' data-security rule and a more specific rule exist on the same model, a user on the Everything rule can see fewer values in a widget than a user on the specific rule.

Workaround: None.

Limitation: Data security filtering fails (MALException / 'failed to apply parallel default logic') when the field used for data security is defined with a numeric type too small to hold its values - for example a small integer (Int16/smallint, max 32,767) field whose actual values exceed that range or contain blanks. Affected widgets return a generic query error and formulas cannot be edited.

Workaround: Change the data-security field to Big Int (or Text) consistently across the data security table and all related tables, then run a full build. Data-security relationships can then remain in place.

Limitation: An ElastiCube custom-table build can fail with 'Invalid cube schema (3000) / no such table' for a dictionary table while Parse and Preview succeed.

Workaround: Apply the workaround shared by Support, which resolved the build on the customer side.

Limitation: The Data Group “Query timeout” is enforced per translation path, so when a query falls back (Original Translator vs Analytical Engine) the effective timeout can be roughly double the configured value, and long queries can occupy the simultaneous-query queue. This is by design.

Workaround: Account for the per-path timeout when sizing concurrency and timeout settings.

Limitation: If an ElastiCube is in the Stopping state when a dashboard is refreshed, its widgets show “This dashboard requires ElastiCube ‘<name>’ to run. It wasn’t found on LocalHost.” This message is by design for the stopping state.

Workaround: Wait for the cube to finish stopping (and start again), then reload the dashboard.

Limitation: A custom table defined with SELECT * does not automatically include new columns added to its source table. The * is only re-expanded when the schema is re-resolved, so until then a manual build or preview can fail with 'Invalid cube schema' / missing SchemaID because the built table lacks the expected columns (scheduled builds may still succeed).

Workaround: After adding columns to a source that feeds a SELECT * custom table, refresh the custom table's schema (Connection Settings > Change tables > Preview) or run a full build so the column list is re-expanded. Adjusting affected column data types and rebuilding also resolves it.

Limitation: With White Labeling enabled, setting the Primary Button text color to white makes text in several Explanations modal windows render in white and become unreadable.

Workaround: Avoid white for the Primary Button text color; choose a color that contrasts with the modal background.

Limitation: The sort order applied to a widget in the web UI is not preserved in the exported CSV file, so the CSV row order can differ from what is shown on screen.

Workaround: Re-sort the data in the exported file as needed.

Limitation: Exporting a dashboard to PDF from the /adminAccess page uses the dashboard's default filter values instead of the filter selections made before export.

Workaround: None.

Limitation: In the new filters, the From and To date fields do not accept keyboard input; the whole date value is highlighted and cannot be edited by typing.

Workaround: None.

Limitation: When a filtered column is renamed or deleted in the cube and the cube is rebuilt, the dashboard filter silently reverts to Include All with no error, making it hard to tell the filter is broken.

Workaround: None (renaming is being addressed by the Display Name feature).

Limitation: With 'Allow rendering Pivot Table content as HTML' enabled, the filter member list shows raw HTML tags instead of the rendered/plain values.

Workaround: None.

Limitation: Filtering with text conditions (e.g., Country Starts With 'A') always includes null and blank strings in the result set.

Workaround: None.

Limitation: After changing a default filter and selecting new values, Clear Selection reverts to the original default value rather than the new selection.

Workaround: None.

Limitation: When a dashboard uses a text-based Highlight filter on widgets that rely on the Analytical Engine (Compatibility Mode with data security), the widgets can fail with a “failed to get and process data-security rules” error and fall back to the original translator. Highlight filters on these widgets may also return “No Results.”

Workaround: Change the affected filter from Highlight to Slice/Filter so it runs on the Analytical Engine.

Limitation: When a widget combines a top/bottom ranking filter on a measure with other measure filters, results can be incorrect. Measure filters have no defined order, so a ranking filter is not consistently applied after the others.

Workaround: Turn off the widget filters and add a single ranking filter on the dimension — e.g., Bottom N ranked by a CASE WHEN formula that encodes the value threshold.

Limitation: Dependent filters cannot be starred individually, and starring the parent filter saves dependents as “Include All,” so a specific dependent selection cannot be preserved. This is by design, consistent with legacy filters.

Workaround: Only the top (parent) filter can be starred; save the parent filter to retain the overall configuration.

Limitation: Toggling "Dashboard Filters" on for a widget (e.g., via the FilterWidget design panel) can cause the date dropdown selection to be applied incorrectly. The same behavior also occurs with standard dashboard filters, so it is not specific to FilterWidget.

Workaround: Re-select the intended date value if an incorrect date range is shown.

Limitation: Creating a Calculated Dimension filter based on the same dimension as an existing FilterWidget can appear to conflict with the "Dashboard Filters" / FilterWidget features, although the filter value itself updates correctly. The same interaction occurs with regular (non-FilterWidget) filters.

Workaround: None.

Limitation: When PastYear is combined with a measured-value formula, results across months / grand totals can differ from a plain measure in ways users find unexpected.

Workaround: None.

Limitation: With the default NEW_THEN_OLD strategy, widgets using PREV/PASTYEAR plus an extra row dimension return partial data (missing future months) because the translator inner-joins and excludes dates lacking explicit current-period records.

Workaround: Enable the singleTableDatePastPeriodStrategy flag (or use OLD_THEN_NEW), or build a normalized-date custom field.

Limitation: MAX/MIN aggregations operate on numeric fields only, so wrapping DDIFF with MAX/MIN over a date field returns a 'must be wrapped in an Aggregation function' error.

Workaround: Use aggregations appropriate to the data type per the dashboard functions reference.

Limitation: A widget using GROWTH shows a widget error when a date list filter selects the very first available date (no prior date) or when no date filter is applied.

Workaround: Enable Analytical Engine compatibility mode on the cube.

Limitation: In the Advanced Formula plugin, a CASE/WHEN returning a concatenated string with an empty-string else branch raises an Analytical Engine parse error.

Workaround: Return a non-empty string (e.g., '0') in the else branch.

Limitation: RSUM totals can differ between a chart with Break By and one without when the data is sparse, because running totals are only calculated for values that appear continuously across the periods.

Workaround: Ensure continuous data per member if matching totals are required.

Limitation: In a calculated dimension, referencing a model field inside quotes keeps the field reference but parses it as a string constant.

Workaround: Reference model fields without surrounding quotes.

Limitation: Certain aggregation formulas that use a CASE-WHEN expression returning only constants — for example sum(case when [Brand ID]=1 then 1 else 0 end) — can fail in the Analytical Engine with “BE#208517 Internal Analytical Engine Error: …Column.getIdentifier() because column is null” instead of returning results.

Workaround: The error occurs because the CASE returns only literals (1/0), so the Analytical Engine has no real column to aggregate. Rewrite the formula so the aggregation operates on an actual field, or use a filtered aggregation. For example, replace sum(case when [Brand ID] = 1 then 1 else 0 end) with a filtered count of a real column — (count([Brand ID]), [Brand ID] = 1) — to count matching rows, or aggregate a real measure inside the CASE — sum(case when [Brand ID] = 1 then [Amount] else 0 end) — when summing a value.

Limitation: A measured-value flag built with CASE WHEN over a SUM can return extra records that don’t meet the criteria (not counted in totals and not found via the filter), depending on which table’s key the formula counts on. This is by design given how measured values and multi-pass are evaluated.

Workaround: Use the key/field from the table that is directly joined for the criteria (the customer’s working “en hh id” variant).

Limitation: For a Table widget, layout.json initially stores pxlWidth as 0, but opening the widget editor and applying changes it to a non-zero integer with no real layout change, producing spurious Git diffs and merge conflicts.

Workaround: None.

Limitation: Because connections contain sensitive credentials, they are not pushed to Git — only the connection OID is tracked. When pulling a data model into another environment, if no connection with a matching OID exists, a new empty connection is created (and the import can fail with a generic 500), requiring manual fix-up.

Workaround: Recreate the connection in the target environment with the same OID before pulling, then point the model at it.

Limitation: Designers granted only 'Can Use' permission on an ElastiCube get an error opening hierarchies, because the datamodel bytitle response returns no share data.

Workaround: Grant the user 'Can Edit' permission on the cube so hierarchies open properly.

Limitation: The installed Windows HA wizard folder contains 0-byte Cygwin junction-link files that cannot be archived or copied between Windows servers, so the copied wizard fails with 'command not found' for offline installs.

Workaround: Use the provided workaround for offline HA-wizard deployment built by Sisense.

Limitation: On the AE, switching a Live-model date field to Days/Weeks can fail (Unknown column / parse error) when a table contains two columns with the same name.

Workaround: Rename the duplicate custom column to a unique name.

Limitation: Notebooks can be created with up to 100 characters, but when editing the name you cannot add characters once the length is at/over 50.

Workaround: Keep the edited notebook name under 50 characters.

Limitation: With the NEW_THEN_OLD strategy, a widget joining unrelated tables (no connection between dimensions) returns No Results, whereas OLD_THEN_NEW returns results.

Workaround: Use the OLD_THEN_NEW translation strategy to obtain results.

Limitation: Hierarchies created on a perspective are not included when exporting the data model, so they are absent after import/duplication.

Workaround: Recreate the hierarchies on the perspective after import.

Limitation: In a pivot with Group By (columns), the Rank function skips values because the pivot query is limited to 10,001 rows, so ranking is computed over a truncated subset.

Workaround: Increase the pivot query row limit (e.g., to 100K) so all relevant rows are included.

Limitation: When a pivot's filter behavior is set to Highlight, highlighted values are always rendered in yellow with no Look and Feel option to change the color.

Workaround: None available; the highlight color cannot be customized.

Limitation: With a TOP-N filter on a pivot, sorting is applied to the filtered result set rather than the entire dataset, so the highest overall values may be omitted.

Workaround: None.

Limitation: In the Report Manager report table, some columns (e.g., Dashboard Name, Last Run) do not show sort and/or filter buttons, contrary to documentation implying every field is sortable/filterable.

Workaround: None.

Limitation: GET /api/v1/folders returns folders shared with you when called without parameters, but adding the name parameter filters out folders that are only shared with you (owned-only matching).

Workaround: Retrieve folders without the name parameter and filter results client-side.

Limitation: Sending an empty object {} to DELETE /api/v1/custom_data/query deletes all objects in the Custom Data collection.

Workaround: Always pass an explicit query body; never send an empty object to this endpoint.

Limitation: When the Display Name feature is enabled with the metadata plugin, field/column search using the == prefix fails with a 500 error on the searchByDisplayName endpoint.

Workaround: Search without the == prefix when Display Name is enabled.

Limitation: In Search (Beta), results render as a single continuous ranked list. There is no divider and no separate 'similar widgets' section separating strong matches from looser ones. Match strength is instead conveyed only by the per-row match-score dots (5 to 3 dots).

Workaround: None. Use the per-row match-score dots to gauge relevance. The two-tier strong-vs-similar display (with divider) is planned for GA.

Limitation: An NLQ query with a WHERE clause adds the filtered field as a widget filter and also as a pivot row, which changes the aggregation so all items are returned instead of the intended top-N ranking.

Workaround: None.

Limitation: Smart field/table suggestions are not refreshed after publishing a Live model with a changed display name (works fine for ElastiCubes after a build).

Workaround: Manually trigger POST suggestion/refresh.

Limitation: In Sisense Mobile 3.2.4, a user can log in successfully but no dashboards appear in the app (no error shown), even though the same dashboards display correctly in the web application. Observed with SAML SSO.

Workaround: Access the dashboards through the Sisense web application.

Limitation: On some on-premises installations, opening the Sisense Intelligence / AI options in the UI fails with a console error and the page does not load.

Workaround: A configuration workaround was provided; contact Sisense Support for the on-premises configuration steps.

Limitation: The assistant tab and dashboard option do not appear in the UI when the Sisense-managed LLM (native AI) entitlement is disabled in the license. Even when the tenant has AI credits, the assistant toggle remains gated off until native AI is enabled in the license and the feature is turned on manually.

Workaround: Enable the Sisense-managed LLM (native AI) entitlement in the license, then in Admin turn on the AI features and, in the LLM configuration, select the native Sisense (managed) LLM. The assistant is not enabled by default and must be toggled on.

Limitation: Adding a Data Designer user within a newly created tenant can fail with "Your license is limited to 0 more designers", because tenant creation does not automatically allocate a Designer license count to the new tenant.

Workaround: Manually configure/allocate the Designer license count for the tenant before adding Designer users.

Limitation: When a user parameter with default/group/user-shared values is used in the 'Default Database' field of a MySQL live connection, only the default value is fetched; group- and user-shared values are ignored.

Workaround: None.

Limitation: When PDF and PNG export are removed for a role via the widgets manifest, the indicator-widget Download menu still shows the Image option alongside CSV.

Workaround: None documented; only CSV should be used when image export is intended to be restricted.

Limitation: Groups in the User Management page cannot be sorted alphabetically; the sort control has no effect.

Workaround: None.

Limitation: Jump to Dashboard in dashboards accessed via Web Access Token returns No Results; Sisense issues a /api/v1/groups/ call that returns 403 in the WAT context.

Workaround: Use Jump to Dashboard under a normal authenticated user session rather than through WAT.

Limitation: On a column chart with a break-by, the hover tooltip shows totals inconsistently (e.g., a total appears for some categories but not others), which users find confusing.

Workaround: Disable totals in the tooltip if they are not needed.

Limitation: Scatter/area map widgets cap results at 5000 rows (FETCH NEXT 5000 ROWS ONLY) regardless of the requested count, so not all points render on Live and B2D sources.

Workaround: Aggregate or reduce the number of plotted points so the dataset fits within the 5000-row limit.

Limitation: In a table widget the '>>' control advances a fixed number of pages per click rather than jumping to the last page.

Workaround: None.

Limitation: In Edit Widget mode the Add button disappears in the Categories/Break By/X-Axis section once two X-Axis categories are added, as the maximum is reached.

Workaround: Remove a category to add a different one.

Limitation: An Indicator and a Pivot built on the same measure and filters across joined tables can display different values, because of how each widget type generates its query. This is by design.

Workaround: Review the join/query semantics per widget type (details in the Jira); align the widget construction accordingly.

Limitation: On Column, Line, Area, and Polar charts, long category value names on the X-axis are truncated rather than wrapped to show the full text.

Workaround: None; this is documented behavior.

Limitation: When 'Widget Style' customization is enabled for the Filter Widget, the widget uses a fixed 68px height that can clip its content (for example the 'Select from list' dropdown). The same fixed-height behavior applies to other narrow widgets such as the Indicator.

Workaround: Manually increase the widget's height after enabling Widget Style customization.