Extend current tool stack of data science team

Question

I recently started to lead a data science focused company with two departments that has uses SPSS Modeler extensively. To be honest nobody there has close to the skills of a data scientist. My expectation of a data scientist is in a nutshell that he/she can apply/test/develop complex models to data. As this skill was basically missing in the past, the existing company became a reporting and business intelligence department that is basically providing data in excel to other departments within the overall organization.

The usual process is the following:

1. Somebody from the overall organization needs some data
2. He calls the department and we do SQLs within SPSS Modeler to provide them the data
3. If they are not satisfied we do this over and over again, which created extensive work for all team members

This process is horrible and time consuming!

I recently hired someone who is closer to a "real" data scientist. He can program python/R and has extensive modeling experience. He currently implements a semantic search for email scanning as this is one of the models that we are currently planning to implement. However, he basically does not get accepted by the team as everybody thinks that Jupyter, Python etc. is worthless. They do not need it and are happy what they have and also - to my surprise - with what they are currently doing.

However, our companies roadmap is to become more of an IT/data science unit. I want that as a company we are focusing on creating analytical models, automating dashboards and NOT selecting data and sending excels around!

I know this is a deep cultural topic and I am currently trying to solve it by:

1. In the future only hiring employees that only know R/Python and extensive analytics to drive cultural change
2. Offering for existing employees Python courses and help them come up to speed

I recently communicated that we will still be having SPSS Modeler, BUT for the right problems as python is the right tool for other problems.

I was also thinking to communicate that from now on we only develop models in Python/R, as some of our existing models were only created in SPSS Modeler and are not performing well.

I also feel that I would like to convince the team to use Jupyter even more and they then see the possibilities.

How would you drive this cultural change use Jupyter/Python to come closer to our goal of becoming a IT/data science company!

Answer 1

They do not need it and are happy what they have and also - to my surprise - with what they are currently doing.

There you have it. Your team is happy with what they are doing and their tools work well enough by them. You must prove to them your methods are better or necessary. Heliocentrism had this problem too— it was the better theory, but it took along time to actually be proven useful.

You have to prove your point to them, otherwise all the training and dictates will ineffective. You have to be respectful of your employees and solicit their input. Learn the benefits they see in their current tool. Then, quantify what is broken or inefficient with the current system and why your tools are superior. Then, guarantee your support for them as they learn the new tools.

This is definitely a pattern in the work place: a superior system / new people come in, but struggle to gain acceptance. Reach out to seasoned leaders for advice.

Answer 2

From what you have described the problem is not with the tooling.

SPSS modeler uses Python and you can code in Python for it. But the tool is intended to reduce complexity of coding and ease of deployment.

Switching to Python/R because lack of how to use the full features of a tool is not a good reason to switch.

If you want buy in from management you would need to prove that the cost of switching and the running of the new tool is going to dramatically reduce costs and time.

This factors in (not an inclusive list)

• Phasing out existing processes
• Converting to new systems (time and complexity)
• Upskilling of current employees
• Expectations of employees after receiving new skills.
• Cost of ownership of new process.
• Cost to find/hire new employees
• Any other organizational change.

As I understand you, this is more computer IT/Development team. Are you the manager of that team?

If so one approach to take is as follows.

• Look at where process automation can be improved through custom nodes in SPSS
• Have team members build those custom nodes in Python
• Look to have 1-2 of team focus on learning Data Science skills + Python.
• Have them convert 1-2 process flows to Python (first using SPSS, then native python)

This way you can get a feel if such an organizational change is worth it. You will also potentially get the team on board to switch. Worst case they become more experienced and process improvement reduces annoyances.