Sign up

So you want to be a data scientist?

July 3, 2017

Evo Pricing is hiring

It’s often recruitment time at Evo Pricing. This month we are assessing candidate Data Scientists for our HQ in Turin, and we would like to share how our process works.

One thing we can say with certainty: there’s a lot of talent out there!

We set the bar very high, yet we received over 30 excellent applications for the final round of interviews next week.

So, how did this group of hopefuls get to this stage, and what selection process are they facing?

Fishing in the right pond

They say that when you go fishing you need to go where the fish are.

Turin is an ideal fishing ground for data scientists. The University of Torino pioneered Italy’s first M.Sc. in Stochastics and Data Science, entirely taught in English; July 2017 will see the first of what will no doubt be many graduates from this innovative programme.

The University understands the importance of developing productive relationships with the business world, and so our CEO Fabrizio Fantini was invited onto the steering committee of this course.

This academic/commercial relationship is mutually beneficial. It helps the university develop a course that is directly relevant to the real world, while companies like Evo Pricing get a chance to interact with and assess the potential of post-graduate students.

It is the word ‘potential’ that is the most crucial to the Evo Pricing recruitment process. While we welcome high academic achievers graduating with top marks and/or having a Ph.D. in Mathematics, Physics or Natural Sciences, we actually put huge emphasis on potential and willingness to learn.

The Evo Pricing recruitment process

When we are ready to hire the next batch, we  ‘invite to apply’ suitably qualified graduates, and also publish a job post on our Careers page. Our lovely Cristina applies an initial screening filter, to make sure that the required academic qualifications are in place: completed a 3+ years university degree, with relevant quantitative focus (not necessarily just Data Science, but also Engineering, Physics, Mathematics etc), high graduation and/or exam marks, and at least some non-academic interests and experiences.

In this latest recruiting round, 34 applications were appropriately qualified and so made it to the first stage – the test.

We feel that it is in the best interests of both the company and the candidates that the test is a fair reflection of the work that we actually do.

No surprises. If you get the job, you’ll be doing more of the same!

First, the test requires some basic understanding and ability to use the R language. Anyone with some programming skills should not see this as a major obstacle, in fact learning R from scratch just to take the test is a perfectly viable option.

R is a high level, expressive language, with an explicit focus on data manipulation and analysis. As is the case in real life, those who do not know how to use this, or how to perform a particular task, should at least know how to learn (fast).

There is no time limit, though we were a little concerned about one candidate who estimated it would take him 9 months to complete this test! It does not matter much if the job is completed over a day or one week (but a couple hours to half day of actual work are often enough), and the answer could be 1 page or 20 pages (but explaining and showing the logic in detail is important).

What is different about the test is the way we assess it. Of course, we use an evaluation grid and give a grade for each skill involved, but we assess mainly the WAY the test is approached. Does the candidate really like this job? Put in the effort required? Follow a logical approach? Have common sense? Explain the work done clearly?

About half of the applicants, even some with excellent CVs, never complete our test (or, sadly, even bother replying to it). However, of the 18 tests returned, 11 passed it – it truly is not a hard step, if the ‘willingness to learn’ is there. Even a short and straight answer can work. So this step has been completed with flying colours, and we are now at the next stage – the role play.

The role play

Programming and predictive analytics are just a part of a Data Scientist’s toolkit.

Being able to communicate our results effectively to clients is a must. That means explaining approach and recommendations to managers and C-level executives in plain English. They do not know data science, but they do have burning business questions that need answering.

So we arrange a 30 minutes Skype or phone interview, where candidates must convince our chief Data Scientist, Giuseppe, that the work is solid. The catch is that Giuseppe acts as a client CEO, so we encounter all the traditional challenges of communicating to non-experts: using simple but specific language; building rapport while handling complex content at fast pace; showing common sense; thinking on one’s feet as unexpected questions may be asked; estimating the time and effort required to build solid production-level answers.

Does the candidate think about outcomes rather than process? Is model accuracy assessed rationally? Is the approach used sensible from a business perspective?

(BTW, the official working language of Evo Pricing is English, and proficiency with the language is also assessed during the role play).

Of the 11 completed tests, 8 were invited to our recruitment day for the final evaluation.

Still with us?


The final round of interviews

Next week we will hold the final round of interviews in person. This is always an exciting but time consuming moment for everyone in Turin. We want to give each candidate a fair opportunity to shine, but at the same time we have hungry clients who require constant feeding and therefore must attend to them also during recruiting day 🙂

So, each candidate is assigned 2 slots, typically of 30 minutes each and in sequence, so we try to be efficient both ways. Each slot has 2 interviewers, so each candidate will see 4 interviewers in total. The goal is to get as many of them as possible excited, but at least one of them converted to a true advocate (the hiring ‘sponsor’). Even a junior member of our team can be the ‘sponsor’, so it is important that candidates persuades each interviewer appropriately.

The interviews have a fairly predictable structure, similar to a consulting interview (for which plenty of prep material is available online): a ‘case study’, then personal questions, and last but not least the chance to ask questions to interviewers who in turn take great care to answer them with as much detail as possible.

‘Cracking’ the case study requires the ability to think in a structured manner (decomposing the problem into its main parts, often as a tree), analyse a problem logically, and perform pen-pencil calculations in real time. Often there is ‘no right answer’, but having pen and paper ready (and used) really helps! Showing the logic, steps and calculations helps to keep the interviewer involved and actively participate in the thinking process.

Personal questions aim at exploring the candidate’s experience, both academically and professionally, to assess fit with the team and long term ambition. We do not necessarily need people who aspire at being with Evo Pricing for life, but on the other hand we would like to understand why are they interviewing with us in the first place. And if we would like to work with them.

Last, the chance to ask questions is a great way to assess how much thinking and research the candidate has done into their application. All questions are welcome, and it may be that there are none.

The reward

We are looking for at least 3 candidates in this round. It is a challenging and highly competitive process, but we have an amazing job to offer. Glassdoor calls this the “best job in America” (and Italy, the UK, Spain, Germany, China, Australia) . . .
Welcome to your data science career 🙂

Visit our Careers page.

About the author

Martin Luxton is a writer and content strategist who specializes in explaining how technology affects business and everyday life.

Big Data and Predictive Analytics are here to stay and we have only just begun tapping into their enormous potential.

Hey! Was this page helpful?

We’re always looking to make our docs better, please let us know if you have any
suggestions or advice about what’s working and what’s not!

Send Feedback