Release 6.0

Meet Tellius 6.0 ✨ A step-change release that turns one tough question into a 20× deeper, governed answer you can ship. Ask in your own words; Kaiya plans a multi-step workflow on the fly, joins the right data, runs advanced methods in SQL/Python, validates the outputs, and returns a single narrative with timing, magnitude, ranked drivers, confidence, and next steps. No pre-setup. Just analyst-grade depth every time.

Also included in the release: one-click KPI switching, selective exports/filters, multi-conditional alerts, faster login, and responsive Vizpads under load. This is your fastest way to move from question → insight → action without detours.

🚀 New features

20x deeper automated insights with multi-agent planning

Kaiya now turns your real-world questions into coordinated, multi-step analysis. Ask something truly complex—“How did Cancellation Rate change week-over-week in Q3, and which regions, products, or channels contributed most to the WoW delta?”—and Kaiya plans the workflow, runs safe SQL to pull the right slices, uses Python for methods SQL can’t cover (e.g., changepoint detection, contribution/variance with volume weighting and sample-size checks), and validates outputs before presenting the answer.

Our semantic layer and knowledge graph encode metric logic and business relationships (e.g., net_revenue = gross_sales − returns − discounts), so Kaiya chooses the right breakdowns automatically and returns the timing and magnitude of the swing, a ranked driver list across region/product/channel, confidence cues where supported, clear visuals, and suggested next steps.

• There’s no pre-configuration required: the plan is assembled on the fly from your question.

• When you like the result, save that exact analysis as a reusable workflow.

• From a single prompt you can drive multi-step variance analysis, cohort retention tracking, anomaly detection, association/correlation checks, causal probes, forecasting with intervals, segmentation/clustering, benchmarking/cohort comparisons, contribution analysis, and what-if explorations—then reuse the same analysis pattern whenever you need it.

• You don’t just get what changed, but when, why, who drove it, and what to do next—and the ability to repeat that depth of analysis at scale without building a workflow upfront.

This allows Kaiya to provide personalized, business-aware responses—it knows your fiscal calendar, preferred chart types, current initiatives, and can connect new questions to previous work. It builds on past conversations and applies domain-specific knowledge, understanding your business, role, and analytical patterns.

Kaiya is ready for your toughest, richest questions

Ask your question and Kaiya chooses the best execution path—SQL for warehouse-native queries, Python for advanced methods—running directly against your Business Views and live sources. A built-in classifier automatically picks the best route per question (no setup required from your end), so you can ask compound prompts with multi-metric and multi-dimension logic—including joins, subqueries, dynamic calculations, and true nested aggregations. When needed, Kaiya asks concise clarifications and proceeds with sensible, disclosed defaults.

Answers are produced via a governed query layer with full auditing and data-source queuing; row-level policies and calendar mappings (including fiscal) are honored end-to-end. Guardrails in the execution plan avoid fan-out and chasm traps, and familiar flows stay fast—straightforward questions follow the standard path, while complex ones seamlessly use the optimized warehouse route.

One click metric switcher: Change just metrics, not Vizpads

You can now add a global Measure Switcher to any Vizpad so a single dropdown swaps one KPI (e.g., Sales) to another (e.g., Profit or Units Sold) across the Vizpad—only charts that actually use that measure will re-query and re-render; everything else stays as-is to preserve visual context. In View mode you just pick a measure and affected visuals update immediately. Under the hood we,

• Identify all charts whose primary measure matches the base measure

• Replace that measure in the chart’s query and visual mappings

• Trigger targeted re-rendering only.

Date-alignment filters return the right dates for the chosen measure; existing filters continue to apply to the refreshed data; number/formatting is per-measure (you add formatting per metric), and when you switch, prior formatting is disabled—not deleted.

Smarter Feeds: Date-controlled, multi-condition alerts with instant summaries

Feed alerts come with the following enhancements to make your alerts actionable and insightful.

• A date field that controls how data is grouped (day/week/month), how periods are compared (WoW/MoM/YoY), and ensures alerts fire against the correct business timeline.

• Include the Kaiya Summary checkbox in alerts so emails come with a PDF that explains the why behind a spike or dip (top contributors, reasons for change, suggested actions).

• Define multiple conditions in one feed with AND/OR (a single or a group of conditions).

For example, say

• You track Monthly Revenue on the ECommerce_360 BV but switch the date field to sales_date so the alert lines up with when revenue is actually recognized.

• You add two conditions: % drop ≥ 10% AND absolute drop ≥ $200k. This cuts noise on small bases.

• You check “Include Kaiya Summary” and specify what the summary should highlight so the alert email explains the drivers (e.g., “Drop concentrated in West, BOGO promo ended; Electronics down 18%”). You act immediately without any detours.

Actionable, interactive tables: Drill, Keep, Remove

You can act directly on any cell in a table or pivot table with Drill / Keep / Remove / Insights / ML-Models.

• Keep / Remove adds or removes a filter for the cell’s dimensions so the table shows only that row (and in an advanced pivot aggregate, it keeps/removes the entire row and its dependent data).

• Drill lets you pick a new drill dimension that’s not already on the Vizpad table; the chosen dimension filter is applied to the entire Vizpad, and the table returns details for that dimension with all available measures; all row/column dimensions present in the pivot are added as filters.

You can save these pivot/table changes as a Custom View so they persist: pinned columns, autosizing, expand/collapse state (Expand All/Collapse All), expanded rows, multi-sort configurations, all filters (chart/tab/global), and all formatting (number/date/grid/conditional).

You trim noise, answer “show me this slice” fast, and reopen the exact state later without rebuilding. For example, click West → Drill by State → Drill by City → Keep “San Francisco” (or Remove “Oakland”) → save as “West–City Deep Dive”.

📈 Enhancements

Stakeholder ready Vizpads: Export and apply filter only to the selected charts

Selective export:

You can now export only the charts you need—pick individual charts or tables using checkboxes (with Select/Deselect All per tab), and the selection persists while you configure the export.

What this means for daily work: you stop cloning dashboards just to present, cut bloat, and keep one source of truth. For example, in a Clinical Trial Reporting: in a 5-tab, 40+ chart workspace, you click Selective Export, select 8 enrollment/safety/milestone charts, include filters/BV/last refresh, and produce an 8-page PDF tailored for regulators.

Selective filtering:

You can now select multiple charts from a tab and apply filters exclusive to those charts, instead of the whole tab. In the filter panel for control or tab filters, you pick the filter, choose the visuals it should affect (bulk add/remove supported) .

Global filters stay global; chart-native filters stay chart-scoped; selective filters apply only to the selected charts, not to other control filters. New visuals you add to the Vizpad are not auto-included. You can switch a selective filter to “entire tab,” delete it while keeping the same visual selection. While exporting, if you enable “include filter details,” the export documents which selective filters are applied to which visuals.

Zero-friction SAML: Simpler setup, safer sessions, faster login

In this release, clicking Login with SAML redirects to Okta and returns to the app in a streamlined flow that preserves SAML’s secure POST, reduces redirects, signs you in faster, and keeps sessions consistent across browsers. When enabled by your organization, first-time users are automatically provisioned after successful SSO, so access is granted immediately without manual setup.

For admins, setup is simplified into a single guided screen with built-in validation, and the backend flow centralizes session handling and removes extra hops. The result is smoother onboarding, fewer configuration touchpoints, and a sign-in experience that’s robust at scale.

Stay fast under load

You now get a real queue and an isolated execution lane for synchronous work (Vizpad and Kaiya queries). Sync jobs run in their own scalable setup, separate from Business View (BV) publishes and dataset refreshes, so interactive analysis isn’t slowed by background workloads. You control hard concurrency limits at the application, live data source, and user levels. When capacity is free, a job runs immediately; otherwise it queues with visible position and estimated wait. You can cancel running or queued sync jobs (creators can cancel their own, if allowed). Because the sync lane is independently scalable, you can add more capacity for Vizpad/Kaiya without touching background job pools—keeping Vizpads responsive even during traffic spikes and eliminating resource dogpiles.

The Notifications tab now surfaces sync activity with the creator name and time taken. In practice, when multiple teams open heavy Vizpads at once, your executive KPI board still renders because sync analysis runs on its own isolated, right-sized lane; background BV or dataset jobs proceed separately. The combination of isolation, limits, cancellation, and clear visibility delivers predictable performance and a single control point for reliable, real-time analytics.

Time resolution now persists with Custom Views

In Vizpads view mode, when you set a Time Resolution (Monthly, Quarterly etc.), the chosen resolution can now be saved as a Custom View—so everyone you share it with and export has the applied resolution. Once you make any Time Resolution changes, the message "There are unsaved changes to your selected view" will be displayed with Save View Changes button.

For example, if you have a quarterly QBR view, you can change the Time Resolution to "Monthly Ops" and save the Custom View for weekly standups.

Tellius, en deine língua

Tellius can now be run in Spanish, German, and Portuguese. Go to Settings → My Profile → Language Preferences, change the language to the desired one, and Tellius will start speaking your language.