Field basics
Calculated fields use a function library covering aggregation (sum, avg, distinct_count), string (concat, split, toUpper), date (dateDiff, truncDate, addDateTime), and logic (ifelse, switch, isNull, coalesce).
As everywhere, a ratio must aggregate before dividing: sum(revenue) / sum(sessions), never a row-level division that gets averaged afterwards.
Dataset-level fields are shared; analysis-level fields are local. Put shared business logic in the dataset.
Level-aware aggregations
LAA functions compute at a specified level of detail rather than the visual's level — QuickSight's equivalent of Tableau's LOD expressions.
Pre-filter functions (sumOver with PRE_FILTER) evaluate before filters are applied. Pre-aggregate (PRE_AGG) evaluate after filters but before the visual's aggregation. Post-aggregate evaluate at visual level.
sumOver(sum({revenue}), [{customer_id}], PRE_AGG)That gives revenue per customer regardless of what the visual is grouped by — the standard pattern for customer-level metrics and percent-of-total denominators.
Table calculations
Post-aggregate window functions operate on the result set: runningSum, percentOfTotal, difference, percentDifference, rank, denseRank, lag, lead, windowAvg, windowSum.
runningSum(sum({sales}), [{order_date} ASC], [{region}])The second argument is the sort order; the third is the partition, which is where the calculation resets. Getting the partition wrong produces a plausible but incorrect chart — the same failure mode as Tableau's addressing and partitioning.
Parameters and controls
Parameters hold a value — string, integer, decimal or datetime, single or multi-value — set by a control or passed in via a URL or embedding.
They drive dynamic filters, metric switching via ifelse, threshold lines, and cascading controls where one parameter narrows the options of another.
Parameters passed through embedding URLs are how a host application pre-filters a dashboard for its user.