This series is written for you the business user, CXO, business owner, who does not care for complex jargon but believes in seeing tangible impact on business. It is to introduce this novel concept of cognitive data-products that are set to storm the business world. I have lost count of the number of times where customer conversations veer towards “Yes; the algorithm sounds good but what decision can I take on the basis of this?”

                                    In-fact we can trace the entire history of data-science in the business world, through a series of dialogues which sound all too familiar.

In the 1980's

Technical Team: Hi there! “business user”, we have found this amazing method by which you can store all this information that you generate.  You can use it to make “informed decisions”.

Business User: Is that so?  Amazing!!

In the early 2000's

Business User: I have spent enough moolah storing all this data! I need to see real returns for having stored all this…. stuff.

Technical Team: We have your back. Look at this awesome visualization tool we made that can query databases and draw beautiful charts!  You can add it to your presentation and show everyone all this data you have stored!

Business User: Err… Ok so A on X-axis and B on Y-axis leads to what? I really don’t see how useful this is for my decision making process. This data just sits there and costs me money while I draw these charts.

In late 2000's

Technical Team: Hmmm. …. Ok we have developed these tools that use statistics to give insights on your data. Also, you can still use those beautiful charts!

Business User: Hmm what is the implication of this 'p-value' (hypothesis testing) on my business? This still doesn’t help my decision making. Buck up guys!  Give me something I can use.

By 2010

Technical Team:  Don’t say such things! We can give you all these case studies where companies employed novel creatures called “Data Scientists “who impacted their business positively.  Of-course they used these tools and something new called “Big Data”. It will be a big thing ..because it has the word “big” in it.

(Enter Data-scientist.)

Data Scientist: Don’t worry about p-values. I will help you interpret the results. All you need to do is empty your coffers a little more, to employ more of my kind.

(Cue the Indiana Jones theme as businesses hunt for these mythical creatures called data-scientists)

Business User:  When do we see something amazing from you, O’ Data Scientist. What business decisions can I make with them.

Data Scientist:  Here you go. I have created this model using a black box algorithm. It will give you correct results with 75% accuracy.

Business User: How much time will it take to hit 90% on that?

Data Scientist: Hmmm. A few more guys and a few more months?

Business User: Really …(wryly)

                This conversation shows a growing urgency in the industry to get value for having stored and analyzed information over the years. We see the industry repeatedly questioning the need for having data-science teams as cost centers when they do not show that incremental value to justify their existence.

The biggest divide that “business-science” and “data-science” (Yes! business is a science: that’s a thing now) is caused by three challenges.

1. Interpreting business problems as data-science problems: Business problems like predicting customer churn, personalization for customers through product and content recommendations can be described as a series of data-science problems that need to be solved.  The challenge lies in making the connect between an imprecise problem definition and precise solution.
2. Interpreting the business impact from data science outcomes: The second part of the challenge starts once a precise solution has been obtained. The mathematical results need to be fed back to the business audience as suggestions for action while explaining the caveats that come with the territory. 
3. Codifying “business experience” to reduce turnaround time: Increasingly managers are demanding shorter turn-around times for solutions to data science problems in order to employ insights as a part of real-time decision making.

Is there a unified method of meeting these challenges so that more business users can use data-science for their daily decision making needs? I think there is.

It is common knowledge that we become good at a task when we keep repeating it while learning from our mistakes. Also, if multiple methods are employed to solve a problem, it gives us the capacity to compare one method against the other. The faculties that help us achieve this are memory, knowledge and experience.

Cognitive Data Products try to emulate these human faculties while trying to solve specific business problems. They have an element of memory, element of learning and an element of evaluation. They are characterized by four features

1. They solve a business problem; incrementally improving performance over iterations. (learning)
2. They employ multiple data science techniques simultaneously to choose the best performance based on history. (memory)
3. They assist in the execution of actions based on the business insights they generate as an output. (evaluation and execution)
4. They are a collection of data-products that can be rearranged to solve a wide array of business problems (modular nature)

We will try to make sense of what data-products are and how they connect to cognitive data products in the next part of this blog series

 ( Ref : Images taken from : http://www.cartagenadatafest2015.org/cartoons/ )

( Blog was originally published on : DataRPM )