My leader asked me in a 1:1 recently, ‘Where will data science be in 5 years?’ I somewhat sheepishly replied, ‘I don’t think there will be data science in 5 years.’

 

Oooff. Not to talk myself out of a job here, but for as many success stories around data science as I hear, it (albeit subjectively) feels like I hear 5x as many stories of frustration that data science jobs are mostly analytics jobs, complaints that data science job seekers are either too qualified or under-qualified (no Goldilocks moments of being just right), or that data science work continually gets shelved and unused by the business. There must be a lot of wonderfully trained classification models sitting on computers out there. About a month ago, I saw a LinkedIn blog about how data science is essentially dying. Provocative, but I had a moment where I realized, this is what we’ve been dancing around behind closed doors and don’t say. But to see it there, in plain print, seemed a revelation. And I ponder, if data science isn’t around in 5 years or so, what will be around?

 

It’s openly talked about that ‘data scientist’ is often a catch-all term that could mean any job function ranging from data intake, data cleaning, data warehousing, data transforming, advanced algorithmic development (a lot of work that I think many data scientists would consider on the boundary of data science but not really data science), dashboarding, client reporting, metric generation, etc. And by the way, you need to be an expert programmer, write error handling, logs, object-oriented style, be skilled in deployment strategies, scaling, write documentation, understand bash and PowerShell. Oh yes, and also, we need you to, you know, build models and be an expert statistician too. No wonder it’s messy. 

 

I’ve been trying to come up with a word for what data scientists, including myself, would use to describe what they do if the term ‘data scientist’ was suddenly unusable. Then I came across it: algorithmist. I don’t really even know if this is a word. A quick consultation of Google turned up a blog, a GitHub account, a few other sundry items, but not much. The earliest use of this term I can find is from a 1976 conference where they define algorithmist as ‘a person skilled in the art of devising, expressing, verifying, analyzing, and testing algorithms.’ Yes! Isn’t this what we’ve been waiting for!?

 

I posit that at least part of the reason data science fails often, especially in growing tech cities, is because businesses aren’t looking for data scientists to come in and be algorithmists, although this is the expectation of the data scientist. They are looking for someone to be a partner to them in learning how to use data science tooling in their products and services. And as much as I don’t want to admit this, I think often, complex modeling isn’t the business answer for many problems, and the value gap between fundamental use of data and analytics and predictive modeling is closer than we might think. It’s not a magic market differentiator, although it can be. Data science has its place in business, and that’s where the art truly comes in—identifying when and where you invest in algorithmic modeling and deployment and where you don’t. As Measure What Matters indicates, decide and commit. And that means making a deliberate decision that you won’t commit data science (i.e., algorithmist) resources somewhere else. Data scientists who master the art of partnership with their business and understand things like EBITDA, revenue targets, shareholder and valuations, and perhaps most importantly, the users and business sector, will more successfully become algorithmists. It takes time and building trustful relationships, like anything else.

 

Until then, here’s to the rise (again) of the algorithmist.