If you haven’t already, check out Arun Ulag’s hero blog “FabCon and SQLCon 2026: Unifying databases and Fabric on a single, complete platform” for a complete look at all of our FabCon and SQLCon announcements across both Fabric and our database offerings.
In December, we released two new optimizations for statistics maintenance in the Fabric Data Warehouse and the SQL Analytics Endpoint: Proactive Statistics Refresh and Incremental Statistics Refresh. These features help keep query optimizer statistics aligned with how your data changes over time, leading to quicker query plan generation, less regressions after ingestion, and further minimizing the need for manual statistic maintenance.
Modern analytics workloads are rarely static: tables grow continuously; distributions shift, and “hot” partitions change every day (or every few minutes). When the underlying data changes significantly, query performance can drift if statistics are not managed correctly or fast enough. The goal of these releases is simple: make statistics freshness more automatic and more efficient, so your warehouse stays fast as it evolves.
Why statistics matter
The query optimizer uses column statistics to estimate row counts and data distribution over a particular selection. Those estimates directly influence key planning choices like join order and resource estimations to fund your queries, which in turn affects the performance of the query. When a query is submitted and the optimizer finds that the needed statistics are stale (after heavy data change), it typically requests that the statistics be refreshed synchronously to the query executing to avoid creating a plan based on outdated information. However, sometimes this synchronous process can extend the overall duration of a query—especially as each statistic needs to re-sample the entire column.What’s new in Fabric
- Proactive statistics refresh automatically detect statistic staleness after data change (rather than waiting for a user-triggered query) so that statistic updates can be kicked off in the background. With smart policies baked-in, proactive statistics refresh helps reduce the likelihood of your SELECT query still needing to update statistics during query compilation, saving you time while maintaining accuracy.
- Incremental statistics refresh is designed for large and fast-growing tables where statistics refresh can be expensive. Instead of resampling entire columns (which is the case for typical statistic updates), it only needs to assess the recently added data and combine this information with the existing histogram statistic. As a result, the statistic update operation itself is much faster.
A_GIF_of_multiple_T-SQL_queries_in_a_Fabric_Data_Warehouse._First_inserting_into
Figure: When applicable, statistics refresh proactively and incrementally, minimizing user query duration.
Current observations and reports
Since release, we’ve observed fantastic improvements for workspaces across the globe. As of March 2026, we have seen 90% of workspaces cut statistic updates from query compilation by half thanks to these features. This reduction directly shortens the overall query duration users perceive day-to-day.We’re also seeing encouraging usage from customers who run high-change, low-SLA analytic workloads in Fabric. These are environments where data changes often; dashboards refresh frequently, and even small plan shifts can have tremendous impact. In dedicated testing with the IDEAS internal customer responsible for serving Copilot Analytics, the enablement of Proactive Statistics Refresh and Incremental Statistics Refresh together reduced their statistic maintenance compute by 5x and improved predictability.
As always, outcomes depend on workload shape (data volatility, query patterns, table sizes, concurrency), but we hope these findings further instill your confidence in these efficient and smart optimizations.
These two innovations are just some of the latest improvements we’re thrilled to add to the Fabric SQL Analytics Endpoint and Data Warehouse, as part of our ever-growing suite of built-in performance boosters. For more information about these features and statistics in general, explore the Data Warehouse Statistics documentation resources.