Yesterday I spoke at BigDataMn (video coming soon) about "A day in the life of a Data Scientist". Before my talk I heard a great talk by Nick Kolegraff (I'll post a link to his talk too as soon as I can) about how a lot of the organizations he encounters want to build a Data Science division (the insinuation being a large number of Data Scientists doing something). He said he pushes back on this thought process and made the excellent analogy that your Data Scientists should be considered your "special forces", that they should come in to do defined tasks, not general tasks. This "special forces" analogy along with my experiences the last 3 years has led me to the following conclusion:
Most organizations that think they need staff Data Scientists don't actually need staff Data Scientists.
Most of the Data Science full time jobs I have had really should have been handled as consulting projects. In fact I have left most of my full-time Data Science jobs in under a year because, essentially, the project was over. This is one of the primary reasons I formed FactorialWise instead of taking on new full time employment.
The reason that Data Science can and should often be treated as a project rather than a full time position is exactly what Nick mentioned, there are usually defined tasks to optimize rather than open ended research into new data products. For example here are some typical projects I think of:
- Optimize the workflow of fraud detection to remove (or mostly remove) human evaluation
- Optimize content creation for traffic
- Create customized recommendations for website users
These are pretty typical things companies need and they don't need to hire a Data Scientist full time to accomplish them. These are the kinds of projects that FactorialWise can take from start to finish in a matter of months (if not faster).
Now I do think there are cases where a staff Data Scientist is probably the correct decision. If your company is entirely about data, offers data products as its business model, or has the ongoing need for new data products then a staff Data Scientist maybe the right move. However I would assert that the aforementioned types of companies companies are few and far between. Most organizations need to optimize data driven things while continuing to move forward with their core business - this is exactly what FactorialWise can do for you.