I recently attended The Escape Conference at Robert Morris University. The conference focuses on data science, with a theme of Sports Analytics this year. It was broken into two speech sessions with hands-on classroom workshops in between where brands like IBM and Microsoft walked attendees through different analytics products, or even taught them introductions to programming languages like R and Python.
I attended a morning session for IBM’s Watson Analytics. Watson is the name of IBM’s overarching Artificial Intelligence technology that first burst onto the national scene after using it to design an AI contestant that could compete successfully on the gameshow Jeopardy! IBM has introduced Watson into many high-profile consumer partnerships, uniting with H&R Block on tax software and advertising its insights in a partnership with Professional Tennis.
The intrigue of Watson is its focus on taking queries in natural language and returning deep and even predictive insights. This promise could make analytics much more accessible throughout an organization.
They have released Watson Analytics, a light-touch Business Intelligence Analytics platform to analyze data and create dashboards out of it. It is entering a space with established products such as Tableau, Microsoft’s PowerBI and Qlik. I was excited to get my hands on Watson and see how it integrated its natural-language query system into a BI reporting tool.
Here are my takeaways from IBM Watson Analytics.
- It’s fun. Every time I open a new Adobe program – its blank canvas and dozens of confusing icons - I get heartburn for a week. But, Watson Analytics takes a lot of the guesswork (and control, as I’ll get to) out of the user’s hands. After uploading a set of data to analyze, it doesn’t take you to a blank work page with sets of tools like Tableau. Instead, it prompts you to ask a question or provides you with sample starting points with premade insights and visualizations. Watson is constantly providing new suggestions to explore in the data, and hours can be spent cutting across and exploring variables.
- Grading the data. After a data set is uploaded, Watson provides a percentage score for how clean and complete not only the dataset as whole is, but each individual column. As you scan across the rows you can see if any data cells are null and why Watson is giving the column the score it has.
- Exploring is easy. Always looming at the top of a new workspace is the prompt, “Ask a question about your data.” With most BI tools, the start of a new report requires the work of a whole new visualization, but this use of Watson allows for quick queries that are answered with pre-set visualizations. For effective questions, you do need to be familiar with the column names in your data set, but Watson does well in understanding natural language queries with relevant responses.
After asking a question to Watson, it will respond with interesting pre-set reports that can be further explored.
- Lack of data cleaning tools. Remember when I complimented the cool feature where Watson grades how clean your dataset is? That is a nice feature, but when it comes to being able to clean that data within Watson Analytics, you’ll have to look elsewhere. Granted, it usually ends up being easier to clean a data set before entering it into the BI Analytics tool, but as I’ll get to, it’s unclear that the target user would have experience in data cleaning. In addition to cleaning, the only calculated columns you can add are pre-set functions such as multiply, average, etc.
- Lack of design customization. Watson Analytics are pretty much what-you-see-is-what-you-get. Even the colors on the graphs can only be changed from a pre-set color palette. There was some additional basic formatting functionality around labeling, but even putting trend lines was only available if it was a part of the template visualization. But, I should stress that the IBM team pushed this as a data exploration tool for finding insights, not necessarily the best tool for customizing and visualizing the data set. And with that…
- I’m not quite clear who this tool is for. This is a “lite” offering from IBM, who sell much more robust tools like SPSS for data scientists to use. An IBM representative said the product was for business users at lower levels to ask questions about their business and get inspired for ideas to ask their data scientists. But that’s an odd space to me. Typically those that would work with PowerBI or Tableau would be making reports to distribute to different internal and external groups, and that process often requires customizing the data for each audience. That’s not a strength for Watson Analytics.
As we walked through the tool I kept thinking back to the Nintendo Wii. Nintendo began falling behind in the gaming industry after decades of defining it. Rather than match their competitors in system specifications and graphics, they approached it from a different angle with the Wii. It offered a new motion-sensored controller that relied on user movement on a series of simple games. It was a hit, for a while, and brought gaming to some interesting new audiences. But the simplicity that made it accessible also made it a bit one-note for gamers, who had begun experiencing games with complex worlds and storylines. Ultimately, it failed to win over the gaming audience that made up much of the industry market.
Watson Analytics is refreshingly different, it’s a fun experience, but it’s also a little gimmicky and short on substance. There are also possible integrations with other Watson analysis tools that I did not have the fortune of being able to demo. But as it stands, it would be a really nice add-on to another BI Analytics tool, not necessarily as a standalone tool. On the plus side, Watson Analytics isn’t very expensive and includes a free tier.
If you are new to analytics and have a data set you’d like to explore, this is a great option. If you are responsible for creating a measuring and reporting program for your team, I’d make sure you feel comfortable in Watson’s limitations.