Introduction

QuickSight is a BI service by AWS, that allows you to easily connect to many types of data sources, analyse your data, create visualisations and share them with your users and co-workers. It’s a server-less service running in Amazon’s cloud, and everything, including authoring reports, is done in the browser.

In other words, this is Amazon’s response to Power BI and other cloud-powered BI services. It has launched at the end of 2016, and is, as far as I know, the cheapest BI product out there. Pricing, for readers is 0.30$ per session of 30 minutes, and capped at 5$ per month per user. For report authors, it’s 24$ per month/user, or 18$ with annual subscription.

I’ve recently spent some time having a play with QuickSight and trying to build a few reports that, for comparison, I’ve also built with Power BI and Tableau. In this article I’ll share my experience and my findings. I’ve also been in touch with AWS and got some insights into their plans, so I’ll include that as well. Let’s start.

Data connectors

The first thing you need to do, as is the case with most other BI products, is connect to your data source(s). QuickSight allows you to connect to most of the popular data sources, cloud hosted or on premise. Being a part of the AWS suite, it easily connects to Redshift, all sorts of AWS relational database services (RDS): PostgreSQL, MySQL, Oracle, SQL Server, Aura etc., as well as your own database servers. Other types of data sources are popular file types such as CSV, XLSX and JSON. The list of online services is pretty limited, and includes Salesforce, GitHub and Twitter. The full list of sources can be found here.

If your other solutions are running in AWS, you can use S3 as a data lake type of source. Athena is another AWS service that allows you to build SQL queries on top of S3 and then use them as sources to QuickSight. Read about Athena to find out how to do so.

Data Modelling

After connecting to a data source, the next step is to prepare your data – define relationships between tables or files, rename fields and change data types. You can also create calculated fields that you can then use like any other field. One positive feature that I immediately noticed is that you create data sets independently of reports. This makes it really easy to reuse them. Another positive is that the experience of importing data and preparing it looks and feels very much like Tableau’s. This is, unfortunately where the positive ends. Two major disadvantages are: no automatic relationship detection, and no support for including multiple sources in one data set.

Automatic relationship detection between your tables is done in other tools by comparing field names, as well as values. QuickSight has none of that, and you’ll need to manually map each relationship – you can choose from the list of fields and decide if you want to use left join/right join/inner join or outer join.

Data can be loaded to your data set from one source only. This is not a big problem if you have your data pre-prepared in a data warehouse or if you want to analyse one source. Data modelling is one of Power BI’s strong points, allowing you to merge data from multiple sources. QuickSight simply doesn’t have that at all. My first thought was – this isn’t very bad, because anyway I usually prefer to merge data sources in the data warehouse layer and then just read from one source. However, I’ve then realized it cannot even create reasonable relationships if you use two fact tables sharing the same dimensions.

If you want your report to show data coming from two fact tables, even on the same schema, you’ll need to create two separate data sets, and create reports that connect to both data sets and display visuals from both. You cannot merge the data and show them together in one visualization. Power BI allows that in a very simple way, Tableau requires two separate data sets but allows data blending into one visualization, QuickSight does not allow that at the moment. The good news is that it’s in AWS’ plans for Q4 of 2018.

Report Authoring

A report is called Analysis. When you start creating an analysis you need to choose a data set to be used by it. You can add additional data sets and build visualizations from any of the datasets, but as mentioned above, it is impossible to blend them together. The analysis creation UI looks like Tableau’s. It’s simple and very easy to create basic visualizations by simply dragging and dropping from the list of fields to the canvas.The visual types at the moment are limited to KPIs, tables, pivot tables, and the usual bar/line charts and pie charts, as well as heat maps, treemaps,scatter charts and a map. In checking with AWS, they are working on adding additional visual types. The basics are there though, and unless you need something very fancy and creative, this should usually be enough.

One of the features that I love in Power BI reports and Tableau dashboards is interactions between visualizations – the ability to use one visualization to filter another. If you have a table and a pie chart, clicking on a piece of the pie acts as a filter to the table, and changes the table to show only data that corresponds with that piece of the pie. QuickSight does not allow that at the moment, but the team has confirmed it’s in their backlog.

Other features which I enjoy in other products and are not yet included in QuickSight are:

• Conditional formatting – paint or highlight cells in a table according to values
• Showing data labels on a chart
• Controlling data colors on charts – it is available in some of the visual types but not all
• Sorting a field by the values of another field

Some of these features are planned for future releases.

Drilling through from one visualization to another is done by using a feature called “URL Actions”. I haven’t tried it yet but according to the documentation it is a bit complicated and not as simple to create as in Power BI, where all you need to do is to create a drillthrough filter in one of your pages.

Calculated Fields and Measures

QuickSight supports basic calculated fields, either on a row level (arithmetics, concatenation etc.) or an aggregate level (sum,avg,min,max etc.). They are basic, which should cover most of what you usually need, but there’s no support for rich data modelling like you can do with DAX in Power BI. Again, for most reports this should be fine, especially if you rely on a data warehouse where as much as possible is pre-calculated.

Table calculations (like % of total) are supported on pivot tables only. I’d like to see them become available in other visual types, and that’s indeed the roadmap.

It will be interesting to see, in a future release when QuickSight allows blending several data sources together, if calculated fields will be supported on top of more than one source, for example dividing a value from one source by a value from another.

Security

I’m not much of a security expert but from what I could understand, if your data is stored in AWS, this is perfect for you. QuickSight supports the existing security structures like VPC and allows a secure connection to your sources. Your users do not need AWS accounts, and logins can be created for them.

Online Resources & Community

One of the things that are most important to me when choosing software is the ability to get help when using it, and be able to interact with others who are facing the same challenges. In that sense Tableau and Power BI have very strong communities, with the later currently having around 300,000 posts in their online forums. As or QuickSight, well – there’s absolutely nothing. The QuickSight forum on AWS has zero threads.

Bottom Line

Being server-less is something I appreciate allot, after years of maintaining Tableau, Cognos and Oracle servers.I like the look and feel, and the fact that while being young, QuickSight’s features are already sufficient for basic reports authoring. I also like the pricing which, as far as I know, cannot be matched. I was also very impressed with the quick answers to my questions and the fact that most of the missing features are planned already.

There’s no one clear bottom line, it really depends on your requirements. For the work I usually do for my clients, it seems immature at the moment so I cannot recommend it, as of now (July 2018). I intend to check it again at the end of the year.

Thanks for reading, and as always, I’m happy to hear your opinions and questions in the comments below.

If you enjoyed the content of this article, please consider sharing it with your network on LinkedIn. As an independent consultant, this helps my business. Thank you!