What happens when you saddle one person with the work of three? Often, none of the work is done well, even when that one person is capable and conscientious. The human body has only so much stamina.
Lately, I have seen many job descriptions for “Data Scientists.” This new title is famously undefined and each job description shows real imagination – describing what the writer hopes a Data Scientist to be. One thing that these fantasies tend to share is an emphasis on data management.
What happens when you saddle one person with the work of three? Often, none of the work is done well, even when that one person is capable and conscientious. The human body has only so much stamina.
Lately, I have seen many job descriptions for “Data Scientists.” This new title is famously undefined and each job description shows real imagination – describing what the writer hopes a Data Scientist to be. One thing that these fantasies tend to share is an emphasis on data management.
Over the years, I have spoken with thousands of data analysts working in industry, government, nonprofits and academia. None of these people called themselves “Data Analysts,” at least not until quite recently, and few, if any, I think, would be acceptable candidates to the recruiters who bombard me with these checklist-filled inquiries. In general, though, they were capable analysts, analysts who viewed data with real insight.
When the question of data preparation has come up, everyone says that it takes the majority of their time. It is not at all unusual to hear of over 90% of a data analyst’s time going to data preparation. With that in mind, when I read job descriptions calling for one person to be a database expert and take responsibility for data management as well as an analytics expert responsible for data analysis, I wince. By the time all the data management work is done, there won’t be much left for serious data analysis. Dumping a load of responsibility for data management and programming on the same people you expect to be experts at data analysis isn’t realistic.
This week I’ve heard that the latest is a demand for Data Scientists who program in C and C++. That’s not a total shock; quants, data analysts in finance, have made heavy use of C for decades. But working with them as clients, I have seen that they generally restrict themselves to a very narrow set of analytic methods. This is due in part to the conservatism of the industry, but that’s not all. If trying a new analytic technique requires you to develop a program for it in C, you have a lot of work to do, and need a lot of time to do it, not to mention the help of a quality assurance team and other experts to make sure you do it right.
If you do find magic people who are experts at all those areas, and they generously accept the role you offer, there are still only so many hours in the day. More likely, though, the person you hire will be an ordinary human being, perhaps expert in one area, but not all of them. Most often you will end up with an IT jockey who knows a few analytics buzzwords (because so many recruiters and hiring managers know something about IT, and nothing about analytics.) A great human being, perhaps, someone conscientious, able to produce reports on schedule, but not someone who leads your business to great heights through meaningful data analysis.
Employers who are hot to get on the analytics bandwagon need to start accepting that obtaining valuable insight from data calls for a team of specialists, each with the resources and the time to focus on a specialty.
©2012 Meta S. Brown