By Surya Mukherjee

Hi, my name is Surya and I am an addict…..of self-serve BI solutions. Earlier this year, I attended IBM’s launch conference for Watson Analytics in NYC. Watson analytics is a self-serve data exploration/visualization and business intelligence tool, which is in open beta at this moment. Over the next few weeks, I signed on for the platform and tried my hand at data manipulation using Watson. There are some unique (being a professional marketing bull-buster, I don’t use that term very often) things about Watson that I will attempt to highlight in this post.

A quick segue: I get numerous queries from enterprise clients and vendors around applicability and comparisons between BI tools (self-service included). In most of these queries, I find a line of business person trying to understand how this technology is relevant to their careers, and how to make a good business case for it. This piece will hopefully start some of us on that journey.

#1: Watson Analytics was born an adult

Unlike other self-service solutions in the market, the beta of Watson Analytics feels like a mature product with enterprise grade features. Compare and contrast this to Tableau, Qlik Sense, or Spotfire v0.1 (not the current versions but the first released versions of these software) and you immediately see the difference. IBM is late to the game, but it has brought rocket launchers to a knife fight. You could take this stuff to even the most draconian of IT shops and even they would have to say that they could use some of that – without living in fear of a data pilferage that destroys their career forever. On the other hand, show it to a business user and you may well have viral adoption. This is a great mix of social media meme meets business challenge; IBM hasn’t ever achieved this before.

#2: The data management capabilities of Watson Analytics are….really awesome

I’ve now played with the desktop versions of almost all major self-serve BI tools in the market. No one else starts with data quality. In all fairness, you could do many or all of your required data management tasks most other self-serve platforms – either directly or by tying up/integrating with 3rd party solutions.

The difference with Watson is that it doesn’t assume that you will get around to doing these things. If there is one thing I know for certain about the business user it is that he/she doesn’t think upfront about data plumbing. Great analysis and data exploration are certainly more exciting pieces, but without the un-sexy data quality bit – its GIGO. The moment I uploaded data to Watson, it took me on a data quality journey, but in a language I could understand.

Watson Analytics also provides a set of handy graphical tools to self-service users for “wrangling” data. Watson Analytics applies data-cleansing algorithms (under the covers) as soon as data is uploaded by users to identify common issues such as duplication, mismatched headers/keys, and how sparse the data is. However, the job of data quality does not stop at identification alone. More advanced features in the solution include tabular/spreadsheet views (which are a good way for humans to format structured data) which highlight semantic mismatch and allow the user to change the data so that it fits a particular schema/pattern. For example, a country name column in a spreadsheet could have both USA and United States of America, which would create problems when aggregating these entries. The solution recognizes USA to be a loose synonym/acronym for a given country name and helps the user change all nonconforming entries to one standard at the data level with one operation.

#3: Watson Analytics is automation nirvana

There are many things that get my goat; doing a lot of tertiary things before I can even get to a pie-chart will figure as one of the top ones among them. Who wants to really examine the sparsity and structure of data before analysing it? Not me. And, not many business users. Waiting for IT to do all that doesn’t play very well into the whole business-IT alignment thing. Watson Analytics, using its much fabled machine learning and artificial intelligence algorithms, basically automates much of the painful plumbing. It’s not perfect, but then, reality seldom is. Perfection is in structure, and in the real world, nothing has an apparent structure by design. Analytics is really a forced structure (better called model) on unstructured data. What Watson does is help you see the threads that hold all of this stuff together.

#4: No starting from a blank slate

If I know the data, I probably know what I want to do with it. For my experiments with self-service tools, I chose our own Ovum data – how many people downloaded and read our reports, who are our top clients, what is the average viewing time, what topics are trending – so on and so forth. With this data, it was easy to get started on a bubble diagram from scratch comparing different analysts in my team and who contributes most to our popularity as a firm/team. Instead, if you have given me data about an oil rig, I would have struggled to come up with intelligent analysis. Pretty much like the analysis about the length of skirts to the economy (, that would be prime Buzzfeed material but also entire useless and wrong.

Watson Analytics got started on interesting analysis topics the moment I uploaded my data in. This was all very rudimentary – don’t expect it to do your job – but at least it showed me 10-15 visualizations and gently suggested some interesting aspects I could look at. That to me lowers the entry barrier to analytics by several thousand feet. Again, you will not find this in any other self-service desktop/personal edition client.

#5: The predictive stuff – evolution not revolution

I like being the most intelligent person in a room. Frequently, that involves finding unoccupied rooms, or rooms with technology left-behinds. Predictive analytics sometimes lets me shine in a conference room – but one person showing off isn’t really the goal of any enterprise. What we need today is to raise the collective intelligence of the entire room. What really surprises me, even after many many years in BI, is that how little of predictive technology has actually flowed beyond data geeks and statistics majors. Predictive has been there for ages, but yet the complexity around it mandates that you have a data scientist on your team.

Newer approaches to predictive, such as SAP’s experiments with KXEN, are trying to bring predictive to the masses. Sure, confidence ranges need to be defined before analysis, and K-means need to be understood. But for the business user, getting started is more important. Watson Analytics took my data and ran a few statistical algorithms on it by default, leaving me with a interesting conclusions. You could very well go all the way in many self-serve tools and run R stat, but again, that approach assumes you will finally get around to it and that you like to code. Watson Analytics, instead, does a lot of statistical analysis (some prescriptive) and stores it for you to look at without you having to write a single line of code or even play with the data visually.

And for the brickbats

As far as first impressions go, this is what I have so far. There are obvious gaps in the offering as well as of now; after all it is in beta. Things I would like to see included are a basic option for a white slate start, faster performance (asking that from a beta cloud is probably stretching it), option to connect to live data (coming soon), and storyboarding (again coming soon). IBM has a lot of work cut out on the intuitive and visual front as well – the solution does not offer half as many features around visual manipulation (yet) that a Tableau does. Also, in the world of self-serve BI, other things that matter a lot include:

  1. The UX and viral potential of the solution – requires constant work and research into what competitors are doing and what is appealing design wise, with focus groups, A-B tests and such
  2. Number developers who are actually excited to work on this product and are doing so now
  3. The user support groups and communities built around the product
  4. The availability of ready-made templates and a growing marketplace to exchange ideas, skills, and code/methods
  5. The potential for IT services organizations to make money somehow in this mix
  6. A well thought out Freemium model, and constant work on conversion

To be successful with Watson Analytics, IBM will need to work differently than it has in the past. It will also need to work much harder, executing relentlessly on its strategy. It is a long uphill battle in a crowded market where bigger actually does not mean better. However, its one helluva start.