Search This Blog

Wednesday, August 31, 2016

Power BI with Narrative Science: Look Who's Talking?

[2016-Aug-31] Someone said a picture is worth a thousand words and this is true that in most cases a visual representation of a fact is absorbed better compared with a verbal or textual version of it.  However sometimes just looking at a “picture” or analytical report it also would be beneficial to have some narratives along the way. 

It’s just like when I take my family to an art gallery to observe various paintings then it helps to look at some exhibit notes on the museum walls or tiny descriptions beneath an artist painting; and then it comes, that “Aha” moment, after which I can tell to myself, that’s what he (or she) meant by drawing that piece or art.

In a business world of analytical reports, we understand this very well; we don’t just design and create reports filled with all sorts of visual elements, we provide report titles, metric descriptions and sometime we may add a text box with words that describes what users can see and interact with.

Working for more then 5 years in a marketing department of one company in the past I’ve seen many PowerPoint slides with graphs and tables where additional value was placed on what marketing managers could derive from visual elements of their slides and put into verbal sentiments besides them. And I know that sometimes it had been a very time-consuming task; I wish they had some help them during the time preparing those slides.

There is one company that actually does it, yes, you prepare your data set then create your report visualization and after adding its external component it begins textually telling stories about your data and users can see that live and interactive, isn’t that amazing!

This company is Narrative Science (https://www.narrativescience.com/) and it recently released a report component for the Power BI that does exactly this (https://powerbi.microsoft.com/en-us/blog/get-natural-language-narratives-in-power-bi-reports/).



This component only requires a data set with numeric measures and corresponding textual dimension attributes; after this its internal engine creates textual narratives in a form of a free text of set of bullet points. And the great thing is, if you start interact with your report visuals (filters, slicers, parameters, etc.), the Narrative Science component will reflect those changes and update its narratives in real time.



At the recent Toronto Open Data book club we were discussing a Learning Opportunities Index that is calculated for all the elementary and secondary schools of the Toronto District School Board. Here is their data set for 2014 that I had used to create this Power BI (http://www.tdsb.on.ca/research/Research/LearningOpportunitiesIndex.aspx). The higher this LOI score is for a particular school the more this school is needed in additional support to provide better learning opportunities for its students.

If you notice at the top right corner of the report there is a text box with a footnote “Powered by Narrative Science”, what you see inside is a product of this component, and I didn’t write any of those words.

Here is a slightly changed version of this text box after a clicked a bar that corresponded to the Ward 11, which holds the lowest LOI score value (which is very value good by the way).



and here is a data set that I had used for those narratives:






















Where only School name and associated LOI Score were used:











After this I just had to set visual attributes for this Narrative Science component. Where Type could have four different values (Discrete, Continuous, Percent of Whole, Scatterplot), Structure could take a form of Paragraph or a Bullet list, and Verbosity could be High, Medium and Low.












And the final step would be to adjust the Language settings (all other settings I left unchanged):















Now all numeric measures are properly displayed on my bar chart, report map component correctly displays school locations and Narrative Science component explains and talks about this all at the same time. I strongly believe that there is great opportunity to use this component in various analytical scenarios and user will appreciate this very much, because it does help when someone tells you a story behind a data that you try to visually analyze.

And now you don't have to guess who's talking, it's Narrative Science! :-)

Happy data adventures!