Amazon QuickSight

What QuickSight is

Amazon QuickSight is the business intelligence service in AWS. It connects to data, lets an author build visuals from it and publishes the result as a dashboard other people read.

Being an AWS service shapes what it is good at. Connections to AWS storage, query and warehouse services are documented and direct, so a dashboard over data already held in AWS does not need a separate extraction path built for it.

Four objects, in order

QuickSight is much easier to follow once the objects are separated, because the words are used loosely elsewhere.

Data source. The connection to a system. Configured once, with credentials.

Dataset. What visuals are built from. It selects fields from the source and can carry calculated fields, filters and joins. Preparation happens here so it is not repeated in every visual.

Analysis. The working area. An author adds visuals, arranges them into sheets and configures how they interact.

Dashboard. The published, read only version of an analysis. This is what readers open.

The distinction between an analysis and a dashboard is the one that matters day to day: editing happens in the analysis, and publishing pushes those changes out to the dashboard.

Two ways to reach the data

This is the decision that most affects how a QuickSight deployment behaves.

Direct query. A visual sends a query to the source each time. Readers see what the source currently holds, and every interaction places load on it.

Import. The dataset is loaded into the in memory engine the service provides, and visuals are answered from there. Interactions are faster and the source is left alone, at the cost of the data being as current as the last refresh.

Neither is universally right. A dashboard read by hundreds of people all day is a different proposition from one checked twice a week.

Restricting rows per reader

Row level security associates users with the values they are permitted to see and applies that when a dashboard is opened.

This matters more than it first appears. Without it, serving ten regions means ten dashboards to build and keep in step. With it, there is one dashboard, and each reader sees their own rows.

It is also what makes sharing outside the organisation practical, since a recipient can be limited to their own data.

Who QuickSight suits

QuickSight suits organisations already running on AWS whose data sits in AWS services, and teams that want dashboards without operating a BI platform. The session based pricing option published by AWS also makes a wide reader audience a different calculation from per seat licensing.

Points to consider

It is an AWS service and is designed around AWS. That is efficient inside that environment and a constraint outside it.

The import engine also has capacity to manage. Datasets held in memory consume it, and refresh schedules need setting deliberately rather than left to chance.

Authoring is done in the service's own interface, so it is worth building a representative dashboard during evaluation rather than judging from a feature list.

Getting started

The user guide covers signing up, connecting a first data source, creating a dataset, building an analysis and publishing it. Separate sections deal with the in memory engine, refresh schedules, row level security and embedding.