The Data Science Recruiter Marco Hoefler was interviewed by me about his job and the selection process for Data Science positions. He tells me what qualities applicants need to have in order to succeed in the process and which skills are of secondary importance.
In most cases, the first contact with a company for applicants of any experience level is through a recruiter or talent acquisition partner. For Data Scientists, this is the person to whom resumes and other documents are placed. What does a recruiter actually do to fill Data Science positions? What does he pay attention to and what influence does she or he have in the application process? How does the recruiter see Data Science as an external specialist? I was able to discuss these questions in an interview with Marco Hoefler. He is Talent Acquisition Partner at METRO AG, a global wholesale company based in Germany. In his role, he is responsible for all vacancies in the areas of IT and digitization.
Marco, how long have you been recruiting for Data Science roles?
Since my start in professional recruitment three years ago, these positions have been one of my focal points. As a working student, however, I also had a few such roles to fill.
So, how many applications for these positions have you received over the years?
I estimate that I have done about 125 interviews and looked at the application documents of about 300 to 400 people.
So you are also in the function of a gatekeeper who makes the decision about a first selection?
This property necessarily results as a consequence. However, I would define my primary role here in total differently. In the first instance, I am a consultant who mediates between interested candidates and the needs of our company. In a joint dialogue with candidates I find out whether the step to METRO is the right one for both sides and offers exciting development opportunities. It is very important for me to provide our candidates a good further information flow about potential opportunities, tasks and the team-environment to enrich the first impression after seeing a job advertisement. That ensures a good expectations management for both sides before entering an in-depth recruitment process.
If, in the course of the first review of documents, it should already stand out that there is no suitable intersection, then the gatekeeper effect you mentioned would occur.
How would you describe the candidates who apply to you?
The nice thing is that there is often not the classic candidate and I am in discussions with a variety of different characters and backgrounds. A tendency is, of course, an academic background in mathematics, statistics, and physics. In several cases, I have also had conversations with candidates who, after graduating, initially worked in the research environment of a university.
The bottom line is that the whole spectrum can be divided into two extremes, and there is a variety of hybrid forms in between or even complete career changers: On the one hand, there are personalities who have primarily specialized in solving problems with academic demands for precision/accuracy. Their goal is to find the most 100% valid solution, prediction, etc. and now would like to switch freshly into the business world. On the other hand, there are the candidates who have already worked in the private economy and therefore live the approach to not only strive for maximum precision rather than plan and invest time efficiently. This range is also reflected in the ability to communicate. In Data Science, this is the bridge between what the customer needs and what can be developed in this field.
Could you describe the application process, please?
Our current acquisition process begins with a pre-selection interview. Questions such as the candidate’s current job, his/her strengths, and wishes will be discussed. What motivation does the person have for a change and I ask for first professional skills such as the knowledge of agile methods, which play an important role for us. I also pay special attention to communication skills and whether the candidate can prioritize and react to customer wishes. I like to carry out this step as a video interview, because the costs are low, but you can still find a personal level.
If I discover in this first step that there is a common basis, the next step would be a first discussion with the department. It depends on the current location of the candidate whether this takes place in our company or remotely.
If this more technical discussion is also successful, a next dialogue follows, which in any case takes place on site. In addition to a further technical discussion, the main focus is on the values and culture of the company. If both sides then have an agreement the process leaves my area of responsibility and proceeds to the formulation of the employment contract. This can be be finalized with a quick processing within two weeks.
As a Talent Acquisition Partner, how do you influence the final hiring decision?
Of course, the final decision is made by the Hiring Manager. However, there were of course many conversations between the people involved in the process, such as his/her employees, and myself. As a facilitator of the entire process who moderates the calibration conversations between all involved parties, it is natural that my view on the candidate at least indirectly also influences the decision by discussing.
A very relevant part for the Hiring Manager, however, is also my opinion on whether the candidate can present and explain complex topics from Data Science and Machine Learning in an understandable way for colleagues from outside the field. After all, this is an essential part of the future job and the hiring manager just has the technical knowledge so that my view can deliver added value.
How do you see the meaning of the following attributes or skills: PhD?
In my experience, the PhD is of minor importance for us in Germany in the corporate environment. I do not know of any hiring manager who would have explicitly preferred a candidate on the basis of a postgraduate degree.
Industry experience, in your case wholesale?
This knowledge is very important or at least definitely a great advantage. The quantitative methods are well known to most candidates. However, the background knowledge from the industry allows the analysis to be carried out more target-oriented by the employee. In retail, for example, these are certain perishabilities or special marketing occasions.
Company-specific financial figures?
My personal statement is that this information is rather not important. This interests me less. That the candidate has tried to understand the business model and to derive the associated challenges is, however, of definite interest to me.
Online certifications for technical skills?
The certification of online education providers is of secondary importance to me. I clearly prefer when candidates can convincingly and comprehensibly explain to me how they have worked with technologies on which issues and in which projects. As a Non-Data-Scientist, I need to understand how they came up with practical solutions. Due to the large number of certificate providers, it is also not always easy to classify the value behind each certificate in detail.
Are use case challenges used in your company when selecting Data Scientists? If so, what do they look like?
Such challenges are always applied in our company. We send the candidates a case study in advance, which contains realistic data. They should then work out a solution and consider a presentation. We will ask the candidate what he or she has learned from his or her analysis.
Is there a final solution in these cases or is the task rather kept open?
Our cases are always open for discussion. This reflects our daily lives, in which assumptions have to be made at the end and results have to be discussed. There is not only the simple truth, the everyday life at work is too complex for that.
However, we expect certain core arguments to be used in the presentation. We coordinate these general requirements for certain results in advance with the persons involved in our organization.
Are there any hard knockout criteria you wish to see met?
Not in that sense. Unless someone comes up with a very poor evaluation. This mostly refers to clear technical shortcomings. In my experience, this is, for instance, dealing with statistical outliers, how they are treated and, above all, how this is then justified argumentatively. However, the cases are not equipped with conscious pitfalls that are hidden and that the candidate has to find.
What does the Data Science team look like in your company?
Organizational speaking, it is more a kind of hub than a closed team. This means that the Data Scientists commute back and forth between the departments with which they work on projects and the direct Data Science colleagues. At this point, we have a matrix organization so that the Data Scientists are assigned to both, the department and the Data Science unit.
In terms of personalities, we have a very broad spectrum. The disciplines range from mathematicians, statisticians, and physicists to an increasing number of computer scientists. We have both university graduates, aged roughly 26, and colleagues with about seven years of specific work experience in their mid-40s. By the way, our gender ratio is almost 50/50, which makes us very happy.
How could a Data Scientist make any career in your organization? What are the ways of further development?
Our aim is to enable our employees to develop both vertically and horizontally. This means that a positive development does not require a higher job title following the next. Everyone should decide for him- or herself how he or she wants to develop within the organization without this decision being subject to any evaluation. As an example, this can classically mean that someone from the technical area wants to develop in the direction of a manager. But there can also be a development towards the products. Thus the role of the product owner can be a personal development and a career step. However, it is always possible to take the step back to a different focus.
So this also means a data scientist who wants to do exactly the same job for many years is not viewed critically?
That’s exactly right. We don’t stick to automatisms that dictate a certain development or change. Of course, a general interest in something new is necessary, but this can take place within one’s own domain.
How do you view the fluctuation of Data Scientists in your company and in general?
In my opinion, the fluctuation of Data Scientists at METRO is very low. However, I also think that low fluctuation is a rather German phenomenon. In my experience, the need to stay longer with an employer is strongly established in German culture. That’s why I would generally conclude from my observation and without any valuation that the more international a team is, the higher the fluctuation compared to a team recruited locally, mostly in Germany.
For Data Scientists, however, there is also a high demand on the market. This means that market mechanisms show that Data Scientists have a higher fluctuation rate than other disciplines with a greater supply of skilled workers or lower demand.
Has anything changed in the course of your work in the active approach of possible candidates for the Data Science area?
The market has become much more competitive. Even highly individual communications often do not lead to success, as this is done by many companies and agencies. This means that you don’t get a lot of attention from people because of this. Other methods are more promising in which a great touch is achieved, so to speak: technical events in our offices, joint open source projects with our employees, and so on.
Finally, it would be nice to hear your personal view on a much discussed topic: As someone who works closely with Data Scientists, but is not a technical expert in this field, how do you assess the future development of the Data Science, Machine Learning and Artificial Intelligence complex? Do the positive feelings outweigh the fear of serious negative consequences?
First of all, I find the speed of development very exciting. The spectrum of possibilities does not develop linearly, but exponentially. Personally, I notice in daily life that automation running in the background is being used more and more frequently, and then in implementation at lightning speed. Where this makes our lives more comfortable, I welcome it very much. Without wanting to formulate a fear, I am curious to see how society will take up the trade-off between this comfort and the loss of anonymity. I am convinced that there will also be very different tendencies and developments in the different nations, which of course can already be seen today.
I don’t believe in “terminator scenarios” and I also don’t believe in the fact that mankind abolishes itself. However, I think that the most exciting question in the future will be how we can avoid or compensate social distortions due to an increasingly automated environment taken over by intelligent machines. Will social protests increase significantly? Will social classes drift apart or will sufficient new fields of activity emerge on the labor market? So, we are talking about nothing less than a world order in the age of digitization, artificial intelligence, complete automation and quantum computers.
Are you questioning your own job in Data Science? Then why not use CRISP-DM for this review? You can read how this works in my article concerning Data Science career reflection.
