How to improve data science-enabled systems?

 

system_engineering2

Data science in commercial settings is fundamentally about translating and connecting business questions into computer and data science faculties and then translating it back into business processes.

So, in order to connect all the “dots” in as efficient way possible, one has to have deeper understanding of dots itself, sequence in which they appear before one can attempt to connect them.

And these “dots” are nothing else but components of data science lifecycle - business questions, data relevant to these questions, technology that would be used on this data to extract the answers and skills needed. But wait – what skills? Skills needed just to run technologies, skills needed to run tech teams, skills of methodologies and processes, or all this together and above?

And if these “dots” are connected in an optimal way - only then “water would flow” in full steam. And if there is nothing coming out but a trickle – we know well that there is some blockage going on. As any plumber would tell you - finding the blockage, understanding what causes it, and removing blockage and underlying causes  - is the way to get lasting water flaw back in system.

But it gets whole lot more complicated than just understanding dots, and connecting them all in right sequence - just like plumber would do with all the pipes and connectors. In data science, system is much more dynamic and much more "alive”. And with all things “alive” - there has to be natural match between all the components. Just like when you going on a date – everything has to play a part.  If car breaks down to the restaurant, if service is atrocious, food is horrible, if what is planned doesn’t happen and what is unplanned does – date may be memorable for all the wrong reasons. Same thing can happen in in data-science use-cases when things don’t match between people, processes and technologies.

It is whole system that needs to work together. And to make it work one needs to know both; its sum and all its parts, so that they can be connected in best way possible. And without knowing how to put it all together – it is hard to know how to break it down to its building blocks in order to “remove the blockage”. But without controls pegged throughout processes - it is hard to tell to which degree system is functional and if there is block and how big it is.

Without deep knowledge of systems and value chains enabled by data science any improvement can be achieved only by chance–alone, unfortunately data science system engineer jobs don’t exist yet. And that is why data science success stories are still far and few between for many who have invested in it.

No Comments Yet.

Leave a comment