Office Hours
The 4 biggest mistakes of data science job applicants
I’ve interviewed over 90 candidates for data job positions. Here are the most common mistakes of those how didn’t succeed
In the past 12 months, we’ve hired about 10 new data professionals for Hotmart’s Data Science team. Ranging from interns to senior levels, you can imagine all sorts of different profiles applying for data analyst and data scientist jobs. I’ve learned a lot from those candidates, and I hope to have taught something new to each one of them. During this period, I identified some common points in those candidates that, unfortunately, didn’t succeed in this hiring process.
The Maslow’s Hammer
When I apply technical assessments to candidates, I never define one single tool or programming language that she must use to solve a problem. I think that by leaving the options open, candidates will be able to use creativity and experimentation to test different approaches. Most of the time it works, but in some cases the opposite is true.
There were many times when a candidate tried to use one single solution to solve all the problems. When he failed, said that he had no way to solve that situation. An example of this situation is when one thinks that every clustering problem is solved with K-means; or that if you can’t create a model using Facebook’s Prophet, you can’t forecast a time series; or simply thinking that every problem can be solved using machine learning.
How you can solve this: you need to focus on the problem, not on the tool that you are using. On a daily basis, people will come to you asking for “an AI to solve this” and “a model to predict that”; but your first question should not be “what algorithm should I use?”, but “what problem are they trying to solve?”.
Outstanding code skills, awful communication
Sometimes —most of the time, actually — we need to translate a technical concept to a non-technical person, and that is not easy. It doesn’t matter if we are explaining how a metric is calculated or why the model made one prediction, much of our work is in communication.
In some candidates, I’ve noticed a good technical capacity, but they failed to communicate the results. When we asked questions about their solution — questions that a business person would normally ask — they missed the point, failing to justify why they chose a particular approach.
How you can solve this: you must have a good reason for picking the solution you are proposing, which can be benchmarks, previous experience, or experimentation — for instance, you can say “Among the solutions tested, this one is the better because of reasons ABC”. Also, you must know the pitfalls of your solution, what are its disadvantages, and how it can be improved.
The Snippet Machine Gun
This topic is basically an extension of the problem above. One of the reasons why people fail to communicate is to leave storytelling aside and focus on spitting out as much code as possible in the solution. In extreme cases, I received reviews with codes that mixed camelCase with snake_case; with comments mixing different languages, and so on. This is not enough to eliminate a candidate, but it showed a lack of rigor.
In most cases, there was a lack of storytelling: candidates did not take the chance to explain their approach while alternating with codes. In companies like Hotmart, where the technical test comes before the initial conversation with the team, this can either be the difference between moving on to the next phase or being eliminated from the process.
How you can solve this: there is no problem copying a snippet from StackOverflow. I must do this myself a couple of times a day. However, you must be discerning and remind yourself to tell a good story. I have my own reservations about thinking that “good code doesn’t need comments”, and when we talk about the work of Data Scientists and Data Analysts, tools like Jupyter Notebook — widely used in hiring processes — offer us something extraordinary: the possibility of merging stories with code.
They don’t know what the company does
This is the biggest mistake that a candidate can make, especially one that is applying for Data Science positions. I’m not saying that you need to know everything about Hotmart, but you must know, at least, what the company does. But, why is this important?
A couple of months ago I wrote about Hotmart’s hiring process for Data Science positions — written in Brazilian Portuguese. One of the things that I emphasize is that our technical test simulates the daily life of this professional in our team. We create some business questions to be answered, and the more you know about Hotmart’s business model, the better.
It’s not only Hotmart that suffers from this problem, many companies have the same problem. You will act in a position that is quite related to the business and know how the company makes money, who its main competitors are, and what its business model is the minimum you need to have a competitive advantage against other candidates.
