Continued from - Art of Data Science part 1
During the execution of a Data science initiative, one person has to constantly think big and about the business application of the project. May be this is needed in any IT project. But the catch is that in a typical IT project, the teams are large. Hence, 1 big thinker means 1 in a hundred member team and hence may be they are not feeling the pain of finding such members. 20 big thinkers in an organization of 2000 may not be that difficult to find. However, in a data science project with 2-4 members, one big thinker per team means 25-50% of the entire team should be big thinkers. Finding that many is difficult. Also, a typical IT support project offered by the IT head demands a lot less business understanding than a consulting data science project offered by the CEO or COO offices.
Whenever I employed the senior data scientist of the team for this role, results were anything but great!
Let me take an example project that did not end too well. It had all the elements that a data scientist would swoon over. In the first three months of execution, our engineering team set up a complex big data environment, extracted millions of sparse data records, ran fairly sophisticated machine learning models and designed interactive dashboards! I know for sure that this could be a poster project in any big data event. It was a brilliant effort by any yard stick.
But, I, despite being the primary client contact, got too involved in the engineering. My slides were purely technical (how we extracted a complex object with an equally complex query in record time kind of stuff!). During the quarterly review, it became painfully obvious to me that I failed to set the right expectations, contextualize the results or communicate the business benefits of the work constantly to the users.
Client said they understand that the team was working real hard and solving some very complex things. But, they simply did not know what that would mean to them!! After three months of truly cutting edge exceptional engineering, ouch! It hurts!
I also tried a three member team approach for a few projects. One of the members would be a Math Statistics expert, one would be a Programmer/hacker/Big Data Expert. The last one would be a domain expert.I expected that the domain expert would be the story teller in this case and also bridge the gap between the data scientists and the Business users. However, these experiments also failed commonly.
The cause was complete break-down of internal communications. A typical domain expert I would find would be a “retired expert” with over 30 years of experience in his/her area. Typically the other two members would be much junior (<10 years of experience). This senior-junior combo with a vast gap in age, exposure and experience never was very effective with just one way communication! It never yielded any results.
I exhort every organization that’s keen on establishing a center of excellence in data science to appreciate this difficulty. I have been talking to experts and am formulating a few strategies.
First one is creating good data story tellers. It is as big if not a bigger challenge as training people in math and programming.
We cannot and should not expect a domain expert or a senior data scientist to automatically take up the job of a big thinker. A dedicated specialized design thinking training is required for shaping professionals to fill this role. They should become good at,
A close friend of mine who is an expert in design thinking believes that we should be able to train any inquisitive and creative person (with sufficient quantitative bent of mind) for this role. They should be able to quickly grasp any domain and become the bridge. Through INSOFE, I began working with a group of MBA schools to design the curriculum and incorporate these skills in fresh MBA grads. Early signs are encouraging!
I shall explain the other aspects in a follow up article.
Article idea & guidance by - Dr Dakshina Murthy Kolluru
Script, Design & Edited by - Suman Malekani