I just received an ad for IBM’s new cloud based BI Tool, IBM Watson Analytics. Watson became famous initially for beating two champion Jeopardy contestants through its natural language processing.
However, despite the brand, this IBM Watson Analytics service isn’t really related – it just seems to be a basic data exploration and visualization tool. Watson Analytics combines aspects of Microsoft’s Azure ML (predictions and data science) and Power BI (visualizations and dashboards) to provide a business user targeted data exploration tool.
Here is an initial comparison between Power BI and Watson Analytics. Look for a comparison of Azure ML and Watson Analytics in a future post!
Price Comparison
Both Power BI and Watson Analytics have free versions and paid versions. IBM has two paid levels, “Personal” at $30.00 US per month and “Professional” at $80.00 US per month. The differences are based on the amount of data the service will support:
Power BI is $9.99 US per user month and includes a 10 GB per user of data capacity. In addition, there are no limits on the number of rows other than the amount of rows per hourly refresh (which is 1 million).
It should be noted that Power BI has no predictive analytics capabilities in itself, where IBM Watson Analytics has this built in to the tool. Microsoft has an entirely different product for Machine Learning called Azure ML. Azure ML has its own pricing structure.
Uploading Your Data
One of Azure ML’s tutorials involves using Auto pricing data. I created the tutorial experiment using Azure ML and extracted the dataset to a CSV file. I could then use the same sample data to upload to Watson Analytics and to Power BI. I created a “raw” auto data file that has some null values and question marks where there was missing data in the original file and a cleaned up version that has these rows stripped out.
Dropping data into Power BI is easy – you just go to “Get Data” and upload a file. However, unlike Watson Analytics, you can also pull data from a myriad of other data sources including databases, cloud based services, web pages, REST APIs, etc. At least with the free version, Watson Analytics limits you to files and with the paid versions the data sources are still pretty limited (e.g. Box, DropBox, IBM DB2, IBM DashDB, IBM SQL, and Twitter).
With Watson Analytics, you can upload Excel files or CSV files and it does some analysis on the quality of your data.
Ensuring Clean Data
One of the challenges with the raw data file is that it contains blank rows and rows with “?” as the value for price.
If you upload the raw file to either Power BI or Watson in its raw form, it no longer sees price as a numeric column because of the “?” in the data. The column is no longer useful for running averages, totals, etc. and the only option you can use is to count the values (either distinct or not).
In both cases, Watson and Power BI are at least smart enough to not count rows that are blank – for example, if you take an average of 6 rows and 1 row is blank your average is the 5 rows that have data.
Exploring the Data through Natural Language
Once you have uploaded your data, you can ask questions and receive answers through natural language queries. Both Power BI and Watson provide similar facilities. However, Watson seems limited to drilling down to a table of values while Power BI will drill down to a single value.
For example, asking the question “What is the average price for audi” shows the following results in Power BI:
In Watson, the same question sends me to a list of average prices by make instead:
Creating Dashboards
Both Power BI and Watson have the ability to combine charts into dashboards.
In Watson, dashboards are called “Assembled Views”. You create them by picking a layout and then filling it with charts. There a quite a few layouts including traditional single page layouts, tabbed layouts, a vertical infographic style layout, a slide show, and a time journey layout.
Here is a dashboard I created as an Assembled View.
Strangely, the layouts don’t enforce a grid so you can still dashboards around, you can create some awkward designs etc. The layouts seem to be more guides than grids.
In Power BI, you have two levels of dashboards – reports which are dataset specific and dashboards which can aggregate widgets from reports. Reports have more sophistication and support drill down – dashboards provide an easy view while linking to the underlying reports when you drill into an individual widget.
Using Power BI Studio, you can create DAX driven formulas, add additional columns, etc. that go beyond just basic dashboarding. There does not seem to be an equivalent power user tool with IBM Watson Analytics.
One feature that Watson Analytics has that Power BI lacks is the ability to add non-BI information into the dashboard such as external web pages, images, text, etc. Power BI only allows you to insert text or images.
Watson Analytics also provides some interesting layouts specifically for telling a story and creating timelines. These allow you to assemble dashboards in a series and page through them.
Sharing Your Dashboard / Report
Power BI only allows you to share dashboards and not reports. In addition, you can only share within your organizational network – you cannot share with external users at all.
Watson Analytics does not currently support sharing – it says it is “coming soon”.
