IBM Watson vs. Power BI
What are we doing here? Comparing IBM Watson Analytics and Power BI using the same dataset, Dunnhumby: This dataset contains household level transactions over two years from a group of 2,500 households who are frequent shoppers at a retailer. It contains all of each household’s purchases, not just those from a limited number of categories.
Why are we doing this? Big Data is imperative in business to help understand many facets of company data. When Watson was first introduced outside of beta in 2014 it’s aim was to provide a natural language-based cognitive service that provides business professionals access to predictive and visual analytic tools.
Power BI was also introduced by Microsoft in 2014 with a release date to the public in July 2015. The first release of Power BI was based on the Microsoft Excel-based add-ins. Power BI offers Data warehouse capabilities including data preparation, discovery and interactive dashboards. However, this was all via a software application that was downloaded onto your Windows desktop. Only in March 2016 did Microsoft release an additional service called Power BI Embedded (which is what we’ll be using today) on its Azure cloud platform.
The growing simplicity of data usage has been a trend in business intelligence tools. Many business professionals don’t have much experience with databases languages. Thus programs like this seek to offer options that do not require deep programming knowledge. Making data easily accessible and understandable for everyone.
What question are we asking? I’m using the KISS (keep it simple stupid) method here and asking both Watson and Power BI (and Hyper Anna) the following:
What is the trend of sales over year for different income description type?
7 clicks to get outcome
60 second loading time
Descriptions (discoveries) not in depth or in plain language. Extra steps taken to get more insights.
30 second loading time
No plain language insights .
Introducing Hyper Anna:
Hyper Anna are new players in the game. Using natural language Hyper Anna excels at giving its users shortcuts to decisions and actions. Interacting directly with data, Hyper Anna allows anyone in any position to gather and use insights into customer behaviour. How do we do this? Anna is ‘invisible’ software and, like Watson, can be accessed via a web based platform. There is also the option of simply cc’ing Anna into email conversations where Anna will read the context and understand what information is required and then email it to you.
10 second loading time
Plain language, easy to read descriptions and breakdowns of data per income group. Helpful insights that are written in plain language and are already available with no extra steps needed.
How accurate does it understand Natural Language Processing?
Does it display the right chart?
Does it have insights that are easy to understand?
Does it perform analytics for the question?
The end result: From the perspective of myself, a person in marketing that has little experience reading data, Hyper Anna makes the data accessible instantly with ease and little effort providing instant analytical insights in plain language. While Watson did provide me with a similar end result, it was much harder for me to navigate and understand the dashboard and insights without requiring extra steps to be taken to get the insights and correct chart needed. Power BI had an easier dashboard navigation but the lack of plain language insights made it more difficult to understand the data presented.
However, whatever way you look at it, these programs are providing a platform that makes data accessible and easily utilised. All these types of programs are changing the way we look at and read data. What makes Hyper Anna different though is the machine learning technology of the system. Allowing users to ask a question and have insights delivers that are sculpted to reflect the real intent of the question.