When data distribution is heavily skewed, cardinality estimation (how many rows the query optimizer expects each operator to process) can be wildly incorrect, resulting in poor quality query plans and degraded performance. You’ve probably seen the advice to update all statistics if a query plan looks wrong – but is that the right advice? In many cases, no! These are “sledgehammer” approaches, and while they might solve some problems (usually parameter sniffing problems), they don’t solve the actual problem. In this session, you’ll learn a generalized yet tailored-to-the-table way to solve query plan quality problems for very large tables (VLTs). Topics will include creating, using, and updating filtered statistics; using forced parameterization and templatized plan guides; and understanding stored procedures and how they can leverage filtered statistics.