Amazon QuickSight: Reviewing the Tool & Building Out Analysis
Now more than ever organizations rely on large sums of data to make informed business decisions. In order to make the most informed decisions, Amazon QuickSight ensures your data is easy and inexpensive to digest.
Before I continue, you can read this related article I wrote, which documents cost optimization while setting up AWS accounts leveraging S3, QuickSight, and Cost Explorer.
DEMO QUICKSIGHT FOR FREE
Much of what I’ve learned about QuickSight was learned while playing with this free demo. There are six data examples on the lefthand column, and later in the article when we explore analyses from the “Petro Global” data (Image 1).
WHAT MAKES QUICKSIGHT POWERFUL?
QuickSight prides itself as the first BI service built natively for the Cloud with pay-per-session pricing & Machine Learning insights for all users. More specifically, there are five key selling points:
1) Pay As You Go: A common theme across most AWS services, you only pay for what you use. Also for users of QuickSight with only Read permission, you will pay 30 cents per viewing, and it is capped at $5 per month.
2) Auto Scaling & Serverless: QuickSight is highly available, and you can deploy to hundreds of thousands globally without provisioning servers.
3) Deeply Integrated with AWS Services: Data is securely & privately accessed across other services. Everything is easily integrated with a few clicks.
4) Developer Support: You can programatically onboard users and manage content. Also you can easily embed QuickSight into your own apps, and moreover you can pay to exclude all AWS & QuickSight branding if for example you are embedding BI dashboards into a customer-facing application.
5) Machine Learning: You can leverage built-in anomaly and forecasting functions. You can bring over your own model from Amazon SageMaker. Also this is new: You can ask questions of QuickSight using natural language.
The below outlines key functions performed by an administrator of QuickSight.
- Users: You can manage users, invite new users, delete users, reset passworks, etcetera (Image 2).
- SPICE Capacity: SPICE stands for “Super-fast, Parallel, In-memory, Calculation, Engine.” SPICE, sometimes referred to as the brain of QuickSight, rapidly serves and calculates data. In the “SPICE Capacity” tab you can manage and purchase capacity which is bundled and pooled across admins in the account (Image 3).
- Security & Permissions: If for example QuickSight’s access to Amazon Redshift was restricted, an administrator could go here to allow the access so that QuickSight can actually connect and pull the data (Image 4).
- VPC Connections: Again in the example of connecting QuickSight to a Redshift cluster or data source, you can go here to manage security groups and ensure that sensitive data isn’t public (Image 5).
- Domains and Embedding: If you want to embed QuickSight somewhere, you will need to whitelist the domain here (Image 6).
- Account Customization: Go here to make available to users some introductory information in the form of videos, datasets or analyses.
- Single Sign-On (SSO): Working at for example a large enterprise you don’t want to over-manage huge numbers of different User IDs & Passwords, so you can go here to integrate with SSO or Active Directory (Image 8).
MAKING A DATASET
Throughout this article you may have noticed the terms Datasource, Dataset, and Analysis. I want to take a moment to clarify the verbiage here, because it will be important for when we build out analyses.
Data sources could be Redshift, SQL, Cloudwatch, and really anything else to which QuickSight can connect. In Image 9, we are on the Datasets page, and we can click the New Dataset tab in the upper right.
In Image 10, we see the data sources from which we can create a dataset.
We then click Redshift, and when we populate the fields in Image 11, we create the data source.
Now in Image 12, we have created the data source, and we choose a schema to make a data set which is built on top of the data source.
Now that the data set is built, the interface in Image 13 is prompted. Much like CloudFormation, this becomes an easy whiteboard to interconnect other datasets, as we do in Image 14.
So, to be clear, Image 15 is a diagram I made to show that a dataset is derived from a data source, and analysis can then be built from that dataset or network of datasets.
BUILDING OUT ANALYSIS
Now that the dataset is built, we can click publish and visualize to build analysis (Image 16).
Then we have the ability to manipulate visuals, choose different graphs, customize a graph once selected, and more. In Image 17, we chose to play with forecasting data.
So far in this exercise we have viewed QuickSight from an Admin’s perspective. Now we see what users can do with the built analysis from a user’s interface in QuickSight. Specifically we will view the Petro Global sample data referenced in the beginning of this article.
While scrolling through built analysis for Petro Global, I click on on of the tiles, “3rd Party Driven Revenue” (Image 16), and then expand the tile (Image 17). Note that the orange buffered line is using Machine Learning to illustrate a forecast of data based of prior data, with upper and lower bounds visible when hovering the mouse over the line.
To summarize, we have reviewed QuickSight holistically, showed what admins can do behind the scenes, built a dataset from a data source, built an analysis from that dataset, and lastly we saw what analysis leveraging machine learning would look like from a user’s perspective.
For more information please see AWS’ overview of Amazon QuickSight.