Data Science Recruiting — Part 2

In my earlier post, I focused on the key parameters that need to be evaluated in a candidate for a position on the Data Science team. This post focuses on describing the common recruitment processes that I’ve observed.

I have been an active participant in the recruiting process for all teams that I’ve worked for and enjoy creating evaluation steps that can result in the best possible talent pool. I have also been on the other side of the fence and can say that I’ve seen quite a few different styles of recruitment in Data Science teams and I will try to capture as many of them as I can.

Step 1 — Initial Resume Screening

I have typically seen all applicant resumes and cover letters are sent directly to the hiring manager with very little filtering by the HR. This may vary across organizations and allows the recruiting net to be cast wider. It may also be down to the highly technical nature of these resumes that are better understood by people on the team.

Screening is typically done by the Manager along with some senior members of the team keeping in mind the key parameters I suggested in my earlier post. I have also observed that referrals have a very high success rate, mostly because members of the team are excellent judges in determining the fit of a candidate and can recommend them based on their past experiences.

Step 2 — Short Telephone Screening

This round is required because Job Descriptions really suck! It is never able to paint the complete picture and is nothing but a laundry list of tools and qualifications with very little information on day to day activities that a person might do. While working towards a better job description is always a good idea, it also makes sense to have a quick phone call with the candidate to explain the role better and clarify any doubts and questions. This works like an important step for both sides to evaluate if it makes sense to proceed further.

The discussion involves a high-level description of what each side is working on, how the team is structured, what can a day at work looks like etc. It is also a good opportunity to understand the remuneration expectations, work location and timings and answer any questions that the candidate might have. This is not an evaluation step but a quick way to determine if the hygiene factors are met for both sides and make a decision on whether to proceed further or not.

Step 3 — Technical Evaluations

This step comprises a combination of activities that allow the hiring manager to determine the technical and logical competency of the candidate. I have personally seen several variations of these methods and many firms have a combination of these as multiple rounds in the recruitment process.

Logical ability — This is a test of the candidate’s ability to understand and comprehend data and can be referred to as data interpretation, data sufficiency, logical reasoning etc. The idea is to have a standardized test which would include simple data tables and graphs followed by several multiple-choice questions. Most of these tests are timed and also test the ability to work quickly and with high accuracy.

Quantitative Case Studies — These are the more common type of case studies that I have observed where the candidate is provided with an anonymized dataset and the objective is to build a predictive model. The evaluation metric is also well defined and the submission will be evaluated against a hold-out dataset. This is again a timed test but considering that many candidates might be working full-time, the time duration is typically a week including weekends so that there is enough time to work on the problem. For example, one of the larger hotel booking websites, provides a dataset containing various attributes of a hotel listing with the objective of predicting the number of clicks each hotel receives. This particular dataset is very sparse so the ability of a candidate to impute missing features and create a performing model is being evaluated.

Qualitative Case Studies — These kind of tests are less common but I’ve seen them being used in more customer focused data science teams. They are more descriptive in nature and can be along the lines of consulting case studies. It is still technical in nature with the candidate expected to work with real data and arrive at solutions using a mix of probability, optimization etc. They are typically conducted in a face-to-face setting and there is an increased focus on the communication and approach to the problem.

Programming Tests — These tests are more common when hiring for junior levels and are normally used when there are a large number of applicants, for example during campus recruitment. It is typically hosted online using platforms like HackerRank and test programming skills and conceptual understanding in statistics and probabilities. This is often used as a quick method of shortlisting a smaller list of candidates who can be considered for further stages.

Take Home Tests — These are shorter tests typically lasting for 2–4 hours that test a wide variety of skills like SQL, Programming, Statistics, Machine Learning, System/ Data Design etc. It is divided into several parts and some questions may even be open-ended. For example, a popular e-commerce company asks candidates to design a system that can monitor the performance of an elevator system that has been recently installed in a building with m floors and n lift cars. The assumption is that you can measure everything that you want but the candidate needs to come up with a list of important performance measures and how the data can be captured, stored and transformed so that these measures can be tracked.

Step 4 — Face to Face Interviews

This round of the recruitment process is very different and completely depends on the team and organization. The primary objective here is to get an understanding of how it is to work with the candidate and analyze whether he/she can form an integral part of the team.

A day in the office — This kind of interviewing technique is becoming more popular and is basically a real-life test of how it is to work with someone. There is normally a joint task given to the candidate that requires working with other team members. The idea is to maximize interaction by going for lunch together, participate in meetings — things that the team would do on any given day. There might be a short presentation of the task at the end of the day followed by a discussion with the team.

Interviews — There are several variations of the standard face to face interview with each organization choosing different strategies based on their culture and what has worked for them in the past. For example, a start-up that makes popular mobile games conducts a series of 7 interviews with different people on the Data Science team and across the management hierarchy. Their objective is not only to test your technical abilities but also to ensure a cultural fit with the organization. Therefore you may find yourself answering many behavioral questions even though you are interviewing for a technical data science position.

Some of the more rigorous interviews also involve solving puzzles, writing pseudo-code to implement them and other such logical assignments. Team members might also want to provide a glimpse of an actual issue that they are working on and ask for your solution approach. The idea in presenting these questions is not to solve the actual problem but to understand the candidate’s thought process, validate the assumptions he/she is making and test their ability to overcome challenges.

I have also seen some organizations conduct interview discussions in a group with multiple applicants and perform a joint evaluation. This is very uncommon and more prominent in large organizations conducting on-campus rounds.

Each organization has honed their hiring practices over several years and continuously tweaks them based on what works and what doesn’t. Many hiring managers also have their own set of criteria (for example, hiring only from certain Universities) that are also part of the recruitment process. Some other managers look for a specific personal trait or motivation in candidates who want to join their team. These individualistic attitudes are also part of the recruitment process and go a long way towards creating a unique team.

A manager is responsible for creating high-performing teams that work well together and to enable them to deliver astounding work. Each manager chooses a different approach to attract, evaluate and recruit talent that can work well with the team and fit in the organization.