How do you find the right data scientist?

Why it's hard to find the right Data Scientist for your organisation

Recruiters and managers generally don't really know what to do with data science. Even people from fields such as IT or Business Intelligence tend to struggle. Frequently a data science vacancy can be a headache. This is proven and then compounded by the fact companies struggle to get the stellar value out of data science that they have been promised. Only 62% of companies indicate they're getting value from their Big Data and AI investments.

In practice it largely comes down to two reasons:

• Organisations don't know how it's useful to them
• And/or they don't have the proper (data)infrastructure

That's a problem for both the Data Scientist looking for a job, and for the company that wants to improve itself. No need to explain this regularly leads to the wrong person being in the wrong place at the wrong time. And while there is no easy fix, we can help you get the basics right.

"Having spent several years in the industry as a specialist, department head, and now as a consultant, I am increasingly struck by how much difficulty recruiters, HR people and managers have in hiring the right Data Scientists. "

- Alex de Ronde, Data science consultant, Digital Power

What organisations are looking for in a data scientist

To figure out who to recruit and how, you first need to figure out what you actually need. Data science sits at the intersection of (mathematical) modelling, IT, and organisational or domain knowledge. Data Scientists therefore use all kinds of mathematical or technical techniques to do things that are of concrete use to the organisation. Here, however, you can go in any direction. That is why there are also many different types of Data Scientists belonging to different use cases. We list the most important ones.

The main types of data scientists

The Advanced Analytics guru is someone who can do more than an ‘average’ analyst. This person can handle the most complex analyses and has the knowhow to get the data and transform it into useful visualisations. An advanced analyst has significant statistical knowledge and probably some technical background or is self-taught, but doesn't need to be a data engineer or a statistical genius. In fact, this person ‘just needs to get the model working’, preferably the simplest one that does the job. Often this role will be the first one you need in your journey to being data-driven or AI-powered, since they can deliver a lot of value in a short period of time.

The second type is someone who can build end-to-end data science solutions, ones that include the engineering to get a model running on a server and giving live results. This is a profile that is more technically and mathematically oriented, since both are required to get models working reliably and properly in a more advanced context. A pitfall you often see in vacancies is that people recruit for this profile, while not actually requiring these skills. That's a shame, since these people who are more technically and mathematically focused tend to be less interested in understanding your customers or logistics, potentially leading to a poor fit.

A great example that shows the difference between these profiles is when you do personalization, a hot item in the market. Companies that succesfully apply personalization, will generally first move to segmentation, and only then to personalization.

Advanced analytics experts understand the customer and are close to marketeers and managers so they can dive into the data, understand exactly what's happening, and use algorithms and statistics to determine and test potential segments.

Over time you'll want to move from general segments to more specific smaller segments in specific circumstances. At some point a human can't handle those anymore and then you need the end-to-end data science solutions with proper Machine Learning Ops to make these decisions automatically and real-time, for which you need the second type with a different focus and skillset.

There also exist more specific data science roles, that we'll only mention shortly.

• The Analytics Translator understands data science but doesn't actually build models, this person helps people from the business or operations understand the data/tech side and vice-versa, acting as the glue that holds the machine together and preventing miscommunication.
• A Machine Learning Engineer is like the end-to-end data scientist but focusses purely on the engineering side. This becomes relevant in situations where you have extraordinary demands on latency or processing ability (think of banks), or when you have many data science models running side-by-side.
• There also exist experts on specific sub-fields such as someone specialized in computer vision or natural language processing. We can't go into all these here.

The more specialist knowledge, the better? Think again!

Besides (technical) skills, seniority is also important. This is often asked for in vacancies. Bear in mind that the kind of seniority is essential to make data science a success. The difficulty in many organisations is not building something technically - this can be solved with tools and people - but implementing it in the organisation as a whole.

Can your end users actually work with what you build for them? And what process do we follow to determine whether a model is still up to date after six months?

Seniority in data science is therefore not just about specialism in the technical domain, but about overview, organisational knowledge and leadership. Of course, this does not alter the fact that specialists are extremely valuable when they are deployed in the right place, at the right time and in the right way.

The 3 most important selection criteria when looking for a data scientist

So unless specific tool knowledge is absolutely essential, select the most suitable Data Scientist on:

• Type: what kind of actual work will this person do, and do you need a specialist?
• Degree of seniority: are you looking for someone with many years of work experience or do you opt for a recently graduated data scientist who is trained with the most recent developments in the field?
• All the other requirements you have! Such as culture fit and personality.
  Make sure you clearly state these criteria in your vacancy. That way, you avoid disappointment for both parties.

What does your data scientist vacancy look like?

Our question to you, the reader, is now whether you can find the most suitable Data Scientist for your organisation in the usual way:

For example, you check for 'a data science role' on at least 3 years of (relevant) work experience, must have experience with AND R, AND Python, AND cloud, AND Spark, AND MLOps, and why not a few more. Furthermore, you test for keywords, at least a completed master's degree in econometrics (but actually a PhD), 2 years of experience in the financial sector, and 'must get all excited about digging in a big bucket of (unstructured) data'.

So what role are you looking for here? Ask yourself this and take another critical look at your Data Scientist vacancy.

Find the right Data Scientist

Of course, we at Digital Power did our best and made mistakes. In hindsight, we sometimes failed to hire a good candidate, but this way we did build a data science team that many organisations can only dream of.

With the old-fashioned method, we certainly could not have built this team. Are you looking for a suitable Data Scientist? Then ask yourself this: do you want someone who meets your checklist, or do you want the person you need?

We are happy to think with you about how to approach this. Contact us if you want to share thoughts on this.

Receive data insights, use cases and behind-the-scenes peeks once a month?

Sign up for our email list and stay 'up to data':