Patch 5.5.1
Tellius 5.5.1 is here! 🎉 From dynamic metric logic that eliminates endless CASE WHEN sprawl, to improving how Kaiya interprets and learns user language, to smarter axis scaling that avoids visual misrepresentation—the updates are designed to reduce manual effort and improve accuracy across Tellius. In addition, we’ve made numerous enhancements to chart behavior, filtering, calendar mapping, and user experience—alongside a long list of minor fixes that make day-to-day workflows smoother and more predictable.
🚀 New features
Dynamic metric match: No more CASE WHEN sprawl
The all new dynamic metric match turns multiple hand-coded CASE WHEN formulas into a single “smart” metric. Under Data → Business Views → Calculated Columns → Dynamic Match, define your dynamic metric, then list your rules in a CASE block. At runtime, Tellius checks which fields you’ve grouped by (… IN GROUP_BY_COLUMNS) and which filters you’ve applied ('Field = Value' IN FILTER), fires only the matching CASE branch (in parallel, not one after the other), and returns your result instantly. Add new hierarchy levels or filter combinations simply by appending another rule—faster queries, cleaner logic, and a future-proof way to maintain complex business rules.
Write rules from most-specific to most-general using supported operators (=, !=, <, >, <=, >=, IN, NOT IN). Include spaces around operators, parentheses for lists, and single quotes for filters.
In any chart, click the ⋮ menu and select Show Dynamic Match Details to view the measure, group-by, filter, and selected column.
Preview is disabled for dynamic match.
Dynamic match is supported only in Vizpads, Search, and Kaiya—not in Insights, Feeds, change-aggregation charts, or market-share target charts.
Dynamic match columns cannot be used inside other calculated columns—always reference their underlying measures.
Kaiya now understands you with SRL and context-aware adaptation
We’ve done major enhancements to Kaiya’s query interpretation system so that it understands more variations in how you ask about data. Kaiya can more accurately resolve column names, values, and time filters, ask clearer clarification questions, and avoid unnecessary or incorrect filters. The engine also improves top/bottom insights logic and handles date-based queries more intelligently. Over time, Kaiya adapts to your language and learns common phrases and terms used within your organization and responds better to them. Redundant or unsupported filters (e.g., unused preceding periods) and rules are automatically cleaned up, reducing the chance of confusion or misinterpretation.
As part of these updates, Kaiya now supports Phrase Learning within its Semantic Role Labeling (SRL) engine, allowing it to adapt to frequently used phrasings, synonyms, and user-specific language patterns. SRL has also been enhanced to better handle ambiguous filters, automatically default to relevant time dimensions when none are specified, and ask smarter clarifying questions when column names or values are unclear. Additionally, native support for Market Share analysis is now integrated directly into SRL-based interpretation.
What is Semantic Role Labeling (SRL) in Kaiya?
We've enhanced Kaiya’s ability to interpret your questions more precisely using Semantic Role Labeling, a method that identifies each part of your query by its analytical purpose (such as metrics, filters, dimensions, or time constraints). Kaiya can now accurately handle more complex, ambiguous, or casually phrased questions, ask clearer clarification questions only when necessary, and avoid applying unnecessary or incorrect filters.
This update also introduces Phrase Learning, allowing Kaiya to learn common expressions and synonyms specific to your organization. Additional improvements include better handling of time comparisons, automatic defaulting to relevant columns, and native support for market share analysis—resulting in faster, clearer, and more accurate insights.
SRL replaces Kaiya’s previous interpretation engine with a smarter, prompt-driven model that delivers faster and more accurate responses—particularly for domain-specific, complex, or conversational queries."
Smart axis scaling: single or dual scales that match your data
When you plot two measures in charts—like NRx in the thousands and TRx in the millions—each typically gets its own Y-axis. However, separate scales can distort perception: a small value from the lower-scale metric (e.g., 10K NRx) might appear taller than a larger value from the higher-scale metric (e.g., 1M TRx), misleading users into thinking NRx is greater than TRx. This forces business users to interpret axis labels manually and adjust for scale, increasing the chance of misreading charts.
Smart axis scaling eliminates this confusion by intelligently determining when a dual axis is necessary—and if so, how it should be applied. Tellius now evaluates units and value ranges in real time:
If two series share compatible units (like percentages or both in thousands) and differ by less than one order of magnitude, Kaiya anchors them to a single Y-axis. This ensures relative bar or line height accurately reflects the underlying data. For measures Sales and Profit (both in dollars, similar scale), Kaiya uses one Y-axis so the taller bar reflects the larger value (80,000) accurately.
If the measures differ significantly in scale or units, Tellius uses a dual-axis—both axes always start at zero, and distinct legends/colors are applied to maintain clarity. This prevents visual distortion from floating or truncated axes. For measures Revenue (dollars) and Conversion Rate (percent). Kaiya uses dual Y-axes—one for dollars, one for percentages—so each metric is plotted on its appropriate scale, starting at 0.
The result is a more consistent, and intuitive visual experience—ensuring that what users see always aligns with what the data actually means.
📈 Enhancements
Increased the column selection limit in the Trend Insights from 20 to 50 to support deeper analysis, with the added flexibility to configure this limit via environment variables.
Improved compatibility for time ranges used in Search and Vizpad. Formats like “
2012-M2” (monthly/quarterly) are supported and time filter pills appear even when no active BV filter is set.Users can now select and copy text directly from tables in Kaiya responses and other Vizpad tables where annotations (comments) are disabled.
When applying filters from Insight drill-downs (like “Understand Why” or “Contributors”), both categorical values and their associated date filters will now be correctly applied to the Vizpad. Additionally, the “Apply filter set to Vizpad” option is now available even at the top category level in contributors, making it easier to explore filtered insights directly.
Users can now teach Kaiya new phrases using unstructured text inputs. The UI has been updated to support this capability, making it easier to add, review, and manage phrase learning directly.
Under Settings → Users & Groups → Groups → +, an error message will now be displayed if the group name already exists during user group creation.
In Kaiya, clarification questions have been updated for improved tone and clarity—phrases like “Just to clarify” have been removed. Additionally, clarification messages and multi-fact error messages now have enhanced wording and adjusted font sizing for better readability.
Under Data, when setting up data source configurations, custom calendar mapping status is now clearly indicated, helping users understand whether a calendar is usable or requires configuration.
Calculated columns of dimension type within Business Views are now automatically indexed during Business View publishing, ensuring that both existing and newly created Business Views will include these calculated dimension fields in the search index.
In Vizpads, for bar charts with multiple measures and time comparisons, the chart now correctly places only one measure on the Y-axis, while the remaining measures are grouped using the secondary group-by.
Under Data → Connect → Custom SQL, enhanced the table list under “View Tables” to improve clarity and the “Copy” icon provides a confirmation tooltip for better feedback.
Under Data → Connect → Custom Calendar, validation errors during calendar table mapping (e.g., when a selected table's schema doesn't match the required format) will now appear directly within the pop-up.
Added support for displaying applied filters in the Insights pop-up within Kaiya. In Comparison Insights, the label "Compare Times" has been updated to "Time Period" for better clarity, and any filters used during Insight generation are now shown for full transparency.
🛠️ Minor fixes
Temporarily disabled the Stop Workflow/Conversation option in Kaiya to prevent interruptions during ongoing conversations or workflows.
In Kaiya → Manage Learnings section, fixed the issue where label and value text displayed the unintended bold formatting for a cleaner, more consistent look.
A spinning circle will be displayed during login for both Standard and LDAP authentication flows.
Under Data → Prepare → Data, fixed a UI alignment issue in the dataset options menu to ensure proper spacing.
In Search, updated the trend Insight pop-up on line charts to display clearer, user-friendly text when hovering over data points.
In Kaiya, fixed an issue where the "Validating Query" message continued to display even after results were returned.
Under Data → Prepare → Data → Python, fixed an issue where Python (Pandas) code validation was failing during data preparation.
In Kaiya generated summaries, fixed markdown rendering issues where numbered lists were not displaying correctly in Insight summaries, including comparison and trend Insights.
Fixed an issue with triggering comparison Insights directly from Kaiya.
In Vizpads, Excel exports from interactive and detail tables now preserve applied text formatting—such as font color—for both headers and data rows, ensuring consistency with in-app tables.
Under Settings → Security → Authentication, fixed an issue where users were unable to update and save LDAP authentication details.
Under Data → Prepare → Data, an error message will now be displayed when attempting to rename a dataset to an existing name from the Dataset List View.
Under Data → Refresh Pipelines, corrected the "Last Updated" timestamp to display the user's local timezone instead of server time.
Under Notifications → Job Details, restricted column reordering in the Logs table so that users can no longer move or drag table columns.
Under Data → Refresh Pipeline, updated the page design to align with the rest of the platform—added appropriate icons and standardized the layout.
In View and Edit modes of Vizpad, resolved an issue where the Revert and Apply buttons disappeared when bulk configuration was enabled during Business View swapping and formatting changes in Vizpad.
In Kaiya, resolved an issue where the order of chart responses generated during Agentic Flows was not preserved when viewing the conversation from chat history, ensuring a consistent and logical flow of results.
Resolved an issue where linked datasets were not getting updated during a parent dataset refresh. This update also prevents unexpected errors occurring during file refresh actions.
Under Data → Datasets, fixed an issue where the loading spinner in the datasets page did not disappear after the dataset finished loading.
Under Settings → Embed → Kaiya, Kaiya menu option will now be hidden if Kaiya is disabled on the instance.
In Vizpads, we've improved how error messages and icons are displayed and aligned in line charts with many values under "Color By".
When creating Insights from Vizpads, fixed an issue where filters applied during insight creation—such as control filters, Market Share filters, and color-by filters—were not carried over to the final insight.
Under Settings → Users & Groups → Synchronization, fixed an issue where user synchronization with user groups was not functioning as expected, ensuring that users are now correctly synced according to group configurations.
In the scatter plot charts of Vizpads, fixed an issue where the background color was not updating correctly when toggling formatting options in the formatting panel—especially in scenarios like resizing, cloning, or reapplying toggles.
Fixed an issue in the deduplication flow where calculated columns failed due to incorrect formatting in the expression.
In Vizpads, fixed an issue where applying a measure-based filter to Location Map charts with a Color By bucket caused errors due to improper handling of empty rows.
In Vizpads, resolved an issue where aggregation values for measures placed in the Color By bucket were not displayed in chart labels when dual axis was enabled on bar charts.
In Vizpads and Insights, fixed an issue where the "…Summary" loading text was not visible or when generating Kaiya summaries in 3x3 chart layouts.
In Search, fixed an issue where making a few formatting changes (like legend or background updates) re-rendered the charts when Kaiya summaries were turned on.
Fixed an issue where the Snowflake OAuth authentication flow did not function correctly during column mapping, ensuring smooth schema retrieval and mapping in connected datasets.
In Vizpads, fixed an issue where switching between "View Data" and "Chart" modes displayed inconsistent results.
In Kaiya, fixed an issue where updated Kaiya Learnings—especially when changed from specific Business View-level to system-wide—did not reflect properly even after a refresh.
In Vizpads, resolved an issue where aggregation values were still included in CSV exports for pivot and standard tables, even when aggregation display was disabled.
Fixed an issue in Kaiya where data labels were not displayed on bar charts when the x-axis values were dynamically generated from filters, even if the data label setting was enabled.
Last updated
Was this helpful?