What do you Mean by Hands On Experience?

“Hands on” has to be the buzz phrase of the year when it comes to jobs for data. Just about every week I hear this phrase and yet few can define it for me when I ask them. What do you mean you want someone who is hands on? Hands on about what? Often it revolves around data science and I get some very vague and undefined answers. I believe the vagueness stems from the fact that the hiring teams don’t always know what they want. They are relying on people who give them a slice of a bigger picture and that slice of knowledge isn’t often enough to define hands on. However, knowing what kind of hands on experience that is needed is often the most important step to finding the right candidate. 

Data science and data leadership are not the same, lets just get that out up front. Yet so many job descriptions combine the two roles in one and recruiters are challenged to find someone who has it all. A typical role I see for a senior leadership role might read like this:

Must mentor and hire. Also, able to communicate to various groups. Know tools like Spark, Hadoop, R, Python. Understand agile, P&L statements, road mapping, budgets, market projects and compliance. 

Right off the bat, what’s missing? The work to be done! What is the real problem the company is facing? Often the job descriptions I see are two or more jobs in one and nobody can do it all at a high level. There is a skill set to algorithm coding that is different from data engineering and different from data leadership and strategy. You will not find someone great at all three. You will find people who claim to be great at all three but I have been hiring from 7 years in this field and I haven’t met one who really is equally great at all. I have interviewed hundreds and met just as many more. If you have thousands in your pool and nobody matches, then that’s a sign the goals of skills are unrealistic.

It is frustrating to see teams hiring a great coder for a leadership role, only to see that person fail. They may have great credentials, a good degree and working at a great company, but that doesn’t make them a leader. A coder and math person is a great technical resource that can be a great technical leader who helps to train others technically, but they won’t be much beyond that. I have worked with great data scientists who beat the pants off me in coding. However, they failed at doing my job. Not because they are not smart, but because data leadership requires different skills and a different mindset. There are have been many who try to do my job and ended up going back to coding because they realized, it isn’t what they want to do and it was very foreign to them.

A hands-on coder and algorithm developer need to have a few languages but not really a deal breaker if they don’t have your ideal language, you can hire a coder if you have the funds. But they should be hands on at trying out the latest tools in the space and figuring out how they can help build models faster. They should be hands on at learning new skills. I always ask them, what conference, meetups or other groups to they frequent, to learn? What websites to they frequent? What are the books they are reading? What is the next language they want to learn and why? What is their favorite opensource tool? When would they use the cloud? These questions are good questions to ask if you need a coder and algorithm builder. But don’t bother asking them about budgets, agile methods, building a roadmap, etc…, that’s not really their wheelhouse.

Now a data leader gets a bad rap as not being hands on. I challenge you to actually do the job and tell me that’s not hands on. Who is doing the hiring, a data leader. Who needs to mentor, a data leader. Who needs to ensure algorithms are in compliance with laws and contracts, a data leader. Who is managing budget and ensuring the company is investing in the right platforms that produce a solid ROI, a data leader. Often data scientists will want what they know or like, not what is best for the company or customers, someone needs to have that discussion with all parties or it can get costly really fast. And who needs to ensure the projects are building value, via revenue, patents, NPS, etc…, a data leader. Who is setting up the right expectations, the data leader. So most of what you are looking for when you have a team that isn’t performing well, is a data leader role. Often a data leader like myself knows a few languages and the tools, yet we don’t sit around coding all day. I don’t have time!  But to evaluate new tools, yes, you get your hands dirty. You sit in code reviews, you define the actual work the algorithm needs to do, you are actively involved in defining the type of algorithm to use in what situation. Once you have that done, coding and doing the math is a very easy part of the process. The data leader is often the role missing because most think they just need a coder, that’s a misunderstanding of how to achieve success in the field.

Data leaders need to be asked questions such as, how is data science processes different from software development? There is a difference and they should know the answer. What is the make up of a team? How to ensure communication at all levels in the organization? How is the leader good at empathy with various groups internally and externally? What is their approach to mentoring? How do they evaluate tools to be used by the company and team? What is their patent strategy for new algorithms? How do they motivate their team? How do they keep their team from being poached? What were the sizes of their budgets and did they go over? (This last one is big, many companies spent $10 million USD or more on a recommendation engine that should cost a tenth of that.)

As you can see, hands on means different things. It is all about knowing what the company really needs first. Then you can source the right candidate. There are a lot more coder and algorithm builders out there so many companies think that is what they need. But often they need a data leader. If they are struggling to meet goals and numbers, they don’t need a coder, they need a leader. If they have that leader, then you know it is a coder they really need. 

Often a job will have all these and then some, but the company really just wants a coder to crank out algorithms. That’s fine but then don’t demand leadership because if a leader is really doing leadership work, they don’t spend their time writing code all that much. Writing the code takes time and that means you need to focus on the coding, not on mentoring and building a team and practice. Big difference.

When you know the difference, it becomes a lot easier to define, what is hands on. This article helps to define just one scenario, sometimes others like UX or engineering get in the mix, but this one between data science and data leader seem to be the most commonly confused one that I have seen. I hope this helps for when you face this situation next time.