Percentile

AggregationDialect-gated

Percentile(column, p)

The value below which a fraction p of the column falls (Percentile(x, 0.5) ≈ Median). p must be a literal strictly between 0 and 1 (SQX013).

Each engine gets its own native spelling — exact where the database has an exact grouped form, the engine’s own quantile function elsewhere:

Engine	Compiles to	Exactness
PostgreSQL · Redshift · Oracle · Snowflake	`PERCENTILE_CONT(p) WITHIN GROUP (ORDER BY col)`	exact
Databricks	`PERCENTILE(col, p)`	exact
Databend	`QUANTILE_CONT(p)(col)`	exact
ClickHouse	`quantile(p)(col)`	approximate
Trino / Presto	`approx_percentile(col, p)`	approximate
Druid	`APPROX_QUANTILE_DS(col, p)`	approximate
StarRocks	`PERCENTILE_APPROX(col, p)`	approximate
Vertica	`APPROXIMATE_PERCENTILE(col USING PARAMETERS percentile=p)`	approximate
Pinot	`PERCENTILE(col, p × 100)`	scale handled for you
CrateDB	`percentile(col, p)`	native aggregate
MySQL · SQL Server · BigQuery	—	fails loudly (SQX010)

On the last three, PERCENTILE_CONT is window-only (or absent) — SquareX refuses to fake a grouped percentile rather than silently compute something different.

When to use it

SLA / SLO reporting — P90 delivery days, P95 response time, P99 latency.
Distribution cutoffs — what does a top-decile order look like? (Percentile(Sales.Amount, 0.9)).

Example

P90OrderValue := Percentile(Sales.Amount, 0.9)

Under the hood

SquareX
Compiled SQL (simplified)

// On a bar chart grouped by order type
P90OrderValue := Percentile(Sales.Amount, 0.9)

SELECT "Sales"."order_type",
       PERCENTILE_CONT(0.9) WITHIN GROUP (ORDER BY "Sales"."amount")
         AS "P90OrderValue"
FROM "sales" AS "Sales"
GROUP BY "Sales"."order_type"

Good to know

Non-additive — rejected under YTD/RollingSum (SQX015).