My VizChitra Proposal
There is a cool data visualization conference happening in Bangalore next month, and I've proposed a talk on "Self Service BI Serves Nobody". Reposting my proposal here
Next month, in Bangalore, one cool conference is happening, called “VizChitra”. It is a conference around Data Visualization, organized by Hasgeek.
For starters, I love the name. It is a multiple pun, bringing together “Viz” (for visualization), “Chitra” (word in several Indian languages for “picture”) and “Vichitra” (word in several Indian languages for “strange”).
In March the organizers approached me to help canvass the event, and bring in speakers. I enthusiastically agreed, and decided that I’ll submit a proposal myself. Now, as things happen with me, NED took over.
The deadline for submission was 15th April, which was the day after I returned to India from a 2 week trip to the US, and so I didn’t have the time or energy to write a full proposal. The organizers suggested that I put in a placeholder and edit it “later”.
Life and work happened, and I that “later” happened to be today. Looking back, I don’t know what took me so long. And I understand that due to my late (real) submission (among other reasons), my proposal may not get selected.
In any case, I’m happy (now) with the essay I wrote as part of my proposal, and so thought I need to put it here, irrespective of whether my talk gets selected or not.
Self Service BI Serves Nobody
Summary
The last 20 years have all been about self service BI. The understanding here is that you allow the user to slice and dice the data in whatever way they want, and they are able to get the insights for themselves.
However, business users don’t want to serve themselves. They are busy managers, and don’t need an additional job of interpreting data to get insight. They want to be served.
What we need is opinionated BI that directly delivers the insight without the user having to do too much work.
Synopsis
For the last 2-3 decades, business intelligence has been all about allowing “self service” by business teams. Analytics and BI teams, to the large part, have considered job to be putting together data pipelines and models, and building dashboards that “allow any business user to get whatever insights they want”.
To this end, the focus has been on expanding the set of questions that the business users can answer by themselves. For example, early BI tools such as Tableau required one to download data into local tables (or Excel sheets) off which the dashboards would run.
This limited the frequency of update of data, and the amount of insights one could get. The next generation of BI tools, led by the likes of Looker, connected the dashboards directly to the data warehouse in the cloud, expanding the range of questions that could be answered.
The question, however, is whether such “self service BI” actually serves the people it intends to serve. A survery of 100 CXOs I conducted last year confirms that most of them don’t want to look at dashboards, and only want insights (fewer than 5 out of 100 actually look at dashboards on a daily basis).
There are a few common objections - the interpretation of the data is left to the user; one needs too many “clicks” to get to the insight (navigate to the right page, right tab, right dropdown); it looks the same every day so needa a lot of attention to detail to get the insight..
It need not be like this. With AI we can do better. What the executives care about is the insight, and we can bypass the “BI layer” (which is but a conduit) to deliver them the insights they need.
Yes, beautiful graphs can make a difference. Sometimes you get the joy of crafting the perfect visual that once someone sees it the message is crystal clear. However, that is more the exception. For the “rule”, it is best to give the insight directly to the business user.
Expected key takeaways:
Stated versus revealed preferences - managers want control, but prefer to be told the insight than figuring it out for themselves
Insights might literally be staring the user in the face, but it is possible that the user may not discover it.
The focus of BI should be on delivering insights - design, visualization and any other artifacts are just a means to getting there.
Who is the audience for your talk/session?
The primary audience for this session is twofold - business managers and analytics / BI professionals. Actually given the nature of VizChitra, I’m crafting it such that it is most insightful for analytics / BI professionals (of all levels).