Spoiler alert: It doesn’t have anything to do with learning math!
If you’ve ever searched “learn data science”, or “data science”, you will find tons of articles listing all the detailed steps to learn data science: from math & statistics, programming languages to machine learning, etc. Those articles are definitely helpful; however, if you are a beginner, it’s easy to get lost when reading many new terms and complex concepts. You may wonder: Where should I begin? What should be the best way to get start with data science?
Believe me, the first step to start with data science is not to learn math, statistics, programing languages, or machine learning. What you need to have is a data-oriented mindset. You need to be comfortable with using data to make decisions. After that, you will be more confident in learning more complicated and quantitative elements of data science.
This article helps you point out the areas that you need to focus on as a beginner.
Learn how to think in business metrics
A business metric is a quantifiable measure of tracking and assessing the status of a particular business process. For instance, if you are planning to run a marketing campaign, you may need to ask yourself these questions: What are the purposes of running this campaign? What are the positive changes that I want to see in the business performance? What is the biggest problem that I need to solve right now? Without those critical questions, you may realize that you have spent time and effort to take an action that is not meaningful. Thinking in business metrics also helps you sharpen your data mindset.
Let’s think of business metrics as a funnel. The bottom of a funnel (a.k.a. primary metric) is always the most important metric that you are focusing on. This may vary across industries, companies, and departments. However, there are a few general metrics that are normally used regardless of which industry and firm you are working at. Pick revenue or profit for example. The ultimate goal of every for-profit organization is to maximize revenue or profit. Another metric that is likely to be correlated with revenue/ profit is number of items purchased (this can be number of products sold, bookings, or orders, etc.).
Looking at a higher level, there are secondary metrics that you might be interested in. Those metrics largely depends on the industry of function that you are working at. For example, if you are working at Marketing department, you may want to care about traffic sources (E.g.: direct, search, and referral), action rate (contact, lead, bounce, etc.), or SEO keyword ranking. If you are working for an e-commerce firm, you need to monitor average order value, cost of acquiring an user, or conversion rate.
A tip for you to is to spend a decent amount of time in the project planning stage to think about business metrics and ask yourself the questions that I listed above. Every action should closely link to the key business metrics and to improve the business performance. The next step is to monitor the metrics every day or week when you are running your project. Finally, after the campaign finishes, let’s spend time thinking of what have you done well to improve the business metrics? What can you do to improve the metrics next time?
Break down the problem by different aspects
If you suddenly see revenue decreased last week or last month, what will you do? No fancy data science model needed. What you can do is to break down the problem into as many aspects as possible. For example, what locations experienced the biggest down in revenue? What product categories suffered the highest loss? How about the top performing products? What channels saw the poorest performance? Doing so helps you narrow down the problem into small areas.
After analyzing, you may find out that the paid channel is the biggest loss driver. Then, you will do an audit of business process for paid channel. Did anything happen during that period? Did we wrongly allocate the budget? Did competitors spend a lot more investment on paid channel?
If you only see a general down trend and couldn’t find any particular problem, that’s fine. At least, you spent time thinking of the problem in a logical way. The more you think in data, the more you develop a data mindset and data sense, which will be helpful when you learn more complicated data science concepts.
Be comfortable with charts and graphs
Charts and graphs are the easiest ways to look at data. Before learning data science, let’s create a habit of visualizing data in a nice and meaningful way. Think of what form of data visualization that you can use in each case. For example, if you want to see business performances over time, a line chart is the right one. If you don’t really care about trends or daily/ weekly/ monthly performances and you want to see the total number or the difference between groups, a table or a bar graph is a good choice. Depending on your purpose, you may choose the right way of visualizing data to start with.
You will not only need to be comfortable of reading charts and graphs, but also be able to build the right data visualization. Let’s develop a habit of visualizing data whenever you can: build charts and graphs for weekly reports, build business performance dashboards, or drawing visualizations to support your arguments.
Conclusion
Those points above may sound obvious, and of course you may know about business metrics, breaking down the problem, and data visualizations. However, these are the most basic steps to start with, and you can use those points as a checklist to build a data mindset. Once you master these areas, you are ready to get to the core of data science and start your learning journey.
Image credit: American University
Data Science How To 
