The Algorithm Slinger

There is a type of person who is populating data science and analytic teams. These people tend to be in high demand, make great money and are often unchallenged when they voice an opinion.  They come in riding a tool set that to some seems aw inspiring.  Companies will brag about how many of them they have on their team and how many projects they are working on. They will talk about the stats background they have and how much money they are spending to buy the tools these people want. They are called the algorithm slingers, the most dangerous person you can hire in data science.

An algorithm slinger is someone who has skills at making algorithms. They can come from stats, new data science programs or just have a good understanding of deep learning or machine learning. The problem is, that’s about it in terms of what they can bring to the table. Yet they often have great confidence in their skills and that confidence helps them to gain traction.  Because they are very vocal, they tend to get the attention of senior leaders who really don’t understand what makes data science work well. Like in an old western, these algorithm slingers walk into an office and start shooting up the place with algorithms. They don’t kill anyone but they can kill your business just as easily. They will go online and find some open source tool and claim they needed it but you might find that tool hasn’t had any contributions in six months. They will ask for the latest tools, not because you need it but because it is cool and they need it to pad their resume so they can leave before you figure out what damage they caused. They don’t follow a process to get work done, in fact they may even admit, they are just throwing stuff against the wall to see what sticks.  Here in lies the problem, data science needs to be a science in order to work well, when you just spit out algorithms, it is not a science. In fact, it can skill a company if left unchecked.

I have seen this happening over the last few years. Companies will hire people who can code and make an algorithm and pay them $200k or more a year. If you live in mid markets like I do, $200k to someone with little experience and know-how, is a lot of money. The issue is, they don’t actually do data science, even though, that’s what they get paid to do, they do something that to an untrained outsider looks like data science but it is not. They often just throw stuff against the wall and hope something sticks, that’s not science nor is it a good way to run your business if you plan on staying in business. But these people are running rampant in the data science community and causing all kinds of issues.

Essentially it is because data science hasn’t been about science in many companies and these algorithm slingers haven’t figured out how to work on problems in a way that actually drives the business. As a result, they kind of get to do whatever they want.  Would you let finance do whatever or legal?  Freedom to do what they want; big egos and lack of senior leadership often results in projects that never get done and rampant costs overruns. When these costs overruns need to be paid for, other teams often pay the price with cuts of staff. But companies keep demanding to have more of these algorithm slingers because they don’t know what they don’t know.

A real data scientist actually does experiments and knows how to have a hypothesis statement and creates ones before they do any work. One of the algorithm slingers once told me that they will make a hypothesis statement after they figure out the work and can say what it means. That’s not how it works. Data science actually needs to be about science. Test ideas, validate ideas, track results, articulate the meaning of those results when possible. Not just making algorithms and strutting around with an ego about it, that’s not data science.

The other aspect is senior leadership. Often data science sits in analytics or IT. Most CIO do not understand data science, they know software, not the same.   Yes, data science does involve coding but that doesn’t make it the same as software development. Some analytic leaders know traditional BI, again, not the same. Most traditional BI teams just do reporting, while data science needs to be more proactive in how they do their work. Nothing wrong with BI, but let’s not confuse the two. Data science leadership is very new and few know it. The best thing you can do is actually be honest you don’t know and go learn it. Data strategy and data science leadership are vital to any company. Once you have good leadership, you won’t hire an algorithm slinger and you will save yourself a lot of headaches. Case in point:

I once worked on a project. Some algorithm slingers had spent years working on a model with no end in sight. They just kept adding to the project and executives thought it was great because they got to hear stat phrases that they didn’t understand and assumed that meant stuff was getting done. But the only thing getting worked over was the bank account. The algorithm slingers wanted new tools because the tools they demanded they needed 6 months ago where now, not good enough and of course it wasn’t their fault. They needed more people, just like them of course. They had no roadmap, not a single deadline and in fact one of them bragged he never completed anything on time.  They were blowing through $5 million a year and projected to double that every two years with nothing to show for in terms of a product that made money either in revenue or profits. Let’s compare that to a team that had actual data scientists using data science methods.  Another team in the same company had data leadership, a strategy, deadlines and timelines. They created a $150 million-dollar product in less than six months. Which would you rather have? If you want results, don’t hire algorithm slingers.