Mosaic Studio and Model Limitations

The current known limitations for Mosaic Studio and Mosaic models are summarized below. This includes Mosaic model linking, time transformations, and the derived Period-to_date metrics. Some areas are being actively improved on for upcoming releases.

• AI-Assisted Modeling

• Error Messages and Diagnostics

• Performance and Infrastructure

• Mosaic Linking and Multi‑Source Scenarios

• Time Transformations

• Period‑to‑Date (PTD: MTD, QTD, YTD) : Additional Details

AI-Assisted Modeling

• Inconsistent AI results: In some cases, AI‑generated suggestions (for example, relationships, metrics, names) may be inconsistent or not fully aligned with user expectations. Strategy is iteratively improving quality and stability.

• Non‑aggregate metrics are not identified automatically: AI does not reliably detect or classify “non‑aggregate” metrics. Users may need to explicitly configure these metrics rather than relying on the automatic process.

Error Messages and Diagnostics

• Error messages are often not intuitive: Many system and query errors are still reported with low‑level or generic messages. These may not clearly indicate root cause or required corrective action.

• Multi‑source diagnostics need improvement: For multi‑source/multi‑model workflows:

  • Error handling is not yet robust or user‑friendly.

  • Query details can be difficult to interpret.

  Enhancements are planned to make these errors more actionable and easier to understand.

• Mosaic Sentinel reporting for nested Mosaic‑linked models: Mosaic Sentinel does not yet fully support reporting on nested Mosaic‑linked models. Monitoring and observation for those scenarios are limited.

Performance and Infrastructure

• Slow performance on very large models (2000+ attributes):

  • Authoring and consumption on very large models can experience noticeable slowness.

  • Mosaic linking for multi-source scenarios is noticeably slower than for single-source models.

• Instance‑based deployments – Trino memory tuning: For customer‑managed/instance‑based deployments:

  • Trino memory configuration may require manual tuning for optimal performance.

  • Documentation is planned to guide recommended settings.

Mosaic Linking and Multi‑Source Scenarios

• Unsupported linking types: 

  • Linking on grouped or compound attribute forms

  • Linking on geo‑enabled attributes or attributes with compound forms (such as ID and non‑ID)

  • Importing the same Mosaic model twice to mimic the behavior of attribute roles

• Unsupported SQL functions in multi‑source or Mosaic‑linked scenarios: Some SQL functions are not yet supported in these scenarios, which can result in:

  • Errors or unexpected behavior when using advanced expressions

  • The need to simplify or redesign certain calculations

• Feature coverage limitations:

  • Auto Metrics work only for the current model. They are not applied to external or linked models.

  • Freeform SQL (FFSQL) live tables are not supported for multi‑source use cases.

  • Access control lists (ACL) on linked attributes is not fully supported. ACL for linked attributes always inherits the most restrictive governance settings from their source models. Their ACLs cannot be modified independently in Mosaic Studio.

  • Merged attribute forms for linked attributes do not support governance or display-form modifications in the Mosaic model.

  • Multi‑select and bulk rename with AI is not yet available on linked attributes.

  • Sentinel does not report on nested Mosaic‑linked models.

Time Transformations

Time transformations provide capabilities like Previous Day, Week, Month, or Year, as well as Period‑to‑Date (PTD) metrics such as Month‑to‑Date (MTD), Quarter‑to‑Date (QTD), and Year‑to‑Date (YTD). These limitations apply to the general time transformation framework and, where specified, to PTD.

Supported Functions and Aggregations

Currently available time transformation functions are listed below.

Time transformations are designed primarily for simple aggregate metrics. For Live PTD metrics, the avg is not supported.

Function	Supported Aggregations for Live Data	Supported Aggregations for In-Memory Data	First Release
PreviousDay	Sum, Min, Max, Count, Avg	Sum, Min, Max, Count, Avg	Strategy One (January 2026)
PreviousWeek	Sum, Min, Max, Count, Avg	Sum, Min, Max, Count, Avg	Strategy One (February 2026)
PreviousMonth	Sum, Min, Max, Count, Avg	Sum, Min, Max, Count, Avg	Strategy One (January 2026)
PreviousYear	Sum, Min, Max, Count, Avg	Sum, Min, Max, Count, Avg	Strategy One (January 2026)

General Time Transformation Limitations

• Simple aggregate metrics only: Time transformations are supported only for basic aggregates (sum, min, max, count, product; avg with some limits). For other functions, results may appear as null.

• Compound metrics are not supported: Metrics that combine other metrics (for example, Metric1/Metric2, Metric1 + Metric2, or more complex formulas) are not supported with time transformations or PTD.

• Unsupported analytic functions: The following functions are not supported with time transformations:

  • Lag

  • Rank

  • PercentRank

  • RunningSum

  • First

  • Last

• Metrics with non‑aggregating behavior: Metrics that rely on non‑aggregating behavior (for example, row‑level or complex analytic logic) are not supported. They may return incorrect or null results when combined with time transformations.

• Base datetime and derived attributes restrictions: Time transformations, including PTD, do not work when either of the following is placed directly on a dashboard visualization or used in filters:

  • The base datetime attribute used to define the transformation

  • Any derived attribute built on that base attribute

• No nested or cascaded transformations: Multi‑step or nested time transformations (such as “previous month last year” built on top of another time transformation) are not supported.

• Attribute groups over time‑enriched attributes: Attribute groups built on top of time‑enriched attributes may return incorrect data in some cases. These scenarios are considered unsupported.

• Guardrails and documentation still maturing: Not all unsupported scenarios are blocked with clear error messages. In some cases, results may appear as null or partially correct rather than failing explicitly.

Period‑to‑Date (PTD: MTD, QTD, YTD) : Additional Details

PTD metrics (MonthToDate, QuarterToDate, and YearToDate) are built on the time transformation framework, and inherit all limitations listed in Time Transformations above.

Current supported scope for Strategy One (March 2026)

• Functions covered

  • MonthToDate (MTD)

  • QuarterToDate (QTD)

  • YearToDate (YTD)

• Supported aggregations

  • Live: sum, min, max, count

  • In‑Memory: sum, min, max, count, avg

  • In Live, avg is not supported for PTD metrics

PTD‑specific limitations for Strategy One (March 2026)

• Subtotals and grand totals return null: Subtotals and grand totals are not supported for PTD metrics. Totals are not computed using PTD logic and therefore appear as null.

• Dataset‑level PTD metric Filters and metric set filters: Applying PTD metrics as either of the following is not supported in this release and may produce errors or unexpected results:

  • Dataset‑level metric filters

  • Metric set filters

• Filters on multiple time hierarchies: Filters that involve two or more time attributes (for example, mixing different time hierarchies) are not supported when used with PTD metrics if only one of those time attributes appears on the template. Results may be partially filtered, incorrectly filtered, or misaligned with the intended time context.

• Cases that explicitly return null: The following scenarios will return null:

  • Metrics that cannot be transformed, such as:

    • Count Distinct

    • Smart metrics or other unsupported advanced metric types

  • Time levels that do not align with the PTD scope (such as using MTD on a year‑only grain)

Observable Symptoms in Unsupported PTD Scenarios

Because of the limitations described above, unsupported or partially supported PTD use cases may manifest as:

• Blank or null results

• Blank subtotals or grand totals

• Partially applied or misapplied filters

• Differences between Live and In‑Memory answers for the same view

• Errors or incorrect results when using PTD metrics in dataset‑level metric filters