How to ace the take home data science assignment?

Let’s say you just got off the phone with the recruiter about an open data science role. It looks like a great fit based on your discussion, and now you’re awaiting the next steps when you receive the following email:

“We are excited to move further in the interview process with you. The next step in the process is a data science assessment that you will take remotely. Please find the instructions enclosed.”

So how do you now prepare for it?

In this article, I will walk you through what to expect in these take-home exams and give you tips on how to crack them.

Take-home exams are a popular screening technique and sometimes are used even before you have a live interview with someone from the company. They are often part of the technical screening process of the interview and are designed to mostly mimic the type of work you may be expected to perform. The company will likely evaluate you against the key skill sets they need in their data scientist.

The format of these varies from company to company. Based on my research, I have found the following two formats popularly used:

1. Online assessment:

Depending on the company, you may be asked to complete a coding challenge on interview platforms like HackerRank, CoderByte, and Code Signal.?These tests vary in duration but typically take between one to three hours, and you have the flexibility to choose the programming language you want to code in. Often, the platform tests you against the completeness of the code, correctness of the result, and the speed at which you developed the solution.

2. Custom take-home assessment:

Some companies may ask you to complete a take-home data science project. As part of this, they usually give you a broader business statement to tackle and an accompanying data set to use to solve the problem. These projects can be completed from anywhere between six to ten hours. Again, most companies give you the flexibility in choosing the programming language you want to work in.

What is an interviewer looking for?

I spoke to a couple of senior data scientists and when it comes to evaluating performance, this is the rubric they mentioned the most:

• Problem-solving skills: This is one of the key elements of the rubric you may be tested against. The evaluator wants to see your ability to comprehend the problem statement, break it down into smaller tasks, and develop a solution and recommendation.?
• Code-writing skills: These exams are designed to test whether you have the basic code-writing skills expected of a data scientist.?
• Speed: In an online exam where tracking of completion time is possible, you may also be evaluated on the speed with which you can code.?
• Completeness: In the scenario that you aren’t able to fully complete the test, you may be evaluated against what portion of the test you were able to complete.?

Pro tip: In cases where getting through the whole challenge appears difficult to you, address the key questions first and tackle secondary questions later.?

What should you expect in a take-home exam?

If you’re asked to complete such an exam, you can typically expect a few tasks types:

• Query language: You may be asked to perform data operations like joining two tables, aggregating key columns to generate metrics, and other basic operations such as filtering, sorting, case when, and if/else statements. Some tests may look at your know-how of advanced SQL techniques like use of window functions in SQL.
• Predictive modeling: You may be asked to build a predictive model, either classification or regression in most cases, and be asked to score a test data set.?
• Clustering: You can be given an unsupervised learning problem and be asked to develop clusters that are statistically sound and that have the desired business application.?
• Probability and statistics basics: In some cases, you can expect to see multi-choice questions testing you on probability and statistics basics. For probability these could be around conditional probability, Bayes’ theorem, etc. For statistics, these could be around p-values or z-scores.
• Data interpretation: I have seen product and consulting companies ask data interpretation questions, especially if the role has a heavier “data analytics” component to it. You may be given a raw or summarized data set and be asked to answer specific questions, or to generate insights from the data set. Most of these questions are posed in the business context and may require you to make a formal recommendation to the business on the problem you’re asked to solve.?

How should you prep for take-home exams?

Here are a few things you can do to prepare before you are even asked to take an assessment at home:

• Create an often-used code library: You can build out basic functions in a programming language of your choice, so you have them handy when you have to work on a take-home project. If you’re assigned an online assessment and aren’t allowed access to any outside material during the exam, building out these functions can be great practice and will help you code faster during the test.??Here are a few functions you can consider building out:

1. Impute missing values?
2. Plots to do exploratory data analysis
3. Building out summary statistics on a data set?
4. Train and score a classification and regression model?

• Practice coding exercises: You can practice coding challenges on online platforms like CoderByte, HackerRank, LeetCode, and CodeSignal. Most of these online platforms offer free exercises for you to practice. Doing this well ahead of time will help you evaluate where you are when it comes to tackling these tests and address any gaps that you see beforehand. It is also very helpful to familiarize yourself with at least a couple of these platforms so that you know how you can write, run, test, and debug the code on these websites. You may lose time learning to navigate the test environment if you see it for the first time on the day of your test.?

Pro tip: Before you are given the assessment, check with the recruiter on what format you can expect the areas you will be tested on, and the rubric you will be scored with. Your recruiter may have already given you this information, but in case they haven’t, you can always ask this question so that you can adequately prepare for it.?

To ace these tests, you should practice beforehand and make sure your fundamentals in programming and data science concepts are solid. Although they may seem daunting initially, with sufficient practice, you will be able to ace them.