Monthly Archives

September 2017

A data analyst’s learning journey

By | Uncategorized | No Comments

Andrei Deusteanu

Long time, no posting, but here we are with a guest post from Andrei Deușteanu, about his great self-directed learning journey to becoming a Data Analyst.

Read & be inspired!

 

———————————————————————————————————-

The aim of this article is to share my story of using self-directed learning to become a data analyst, even though I had brief technical and statistical background. My hope is to show you that you can truly learn anything given persistence and a structured approach to doing it.

 

What is self-directed learning?

Being a self-directed learner means that you take initiative to find out what your learning needs are, you formulate learning goals, you identify resources, you choose and implement appropriate learning strategies and evaluate your learning outcomes. This happens in a structured manner using a certain process and you might get help from other people. (Knowles, 1975, p. 18). You might wonder how it differs from a desire to always become better. I would say that the desire to learn is the core attitude while the process is the system through which you turn this into reality.

 

I first found out about it in 2014 when I became a student at the Alternative University which is centered around this idea of self-directed learning. The model I use is best described by Rocketists around 3 main clusters – design, do, reflect . Over the past 3 years I’ve used this model as guidance to help me learn things faster and achieve better results. One of the areas where I’ve applied it, is in becoming a Data Analyst.

It all started with a question – why do some Facebook posts perform better than others? I had access to a Facebook page from the student NGO I was in and I was curios how we can tune our content to make it more relevant for our followers. I sliced through various data, applied some statistical functions, but in the end I didn’t really find my answer. At that point the challenge was beyond me.

 

Fast forward 1 year later and I got into an internship on Business Intelligence – the art of transforming raw data into valuable information for business people. This is a relatively new field of work and Data Analysis is one of its subfields. Given that I didn’t have much idea of what are the actual skills required, Ijust dove into an online course from Duke University – Data Analysis and Statistical Inference in order to build some fundamentals.

The course was really good and challenging at various moments and it opened up my eyes to what I need to learn further on. Therefore I set some learning objectives over the main skills of a Data Analyst:

  • Use of descriptive Statistics
    • Be capable of explaining to others summary statistics and their meaning in context
  • Use of Inferential Statistics
    • Apply the right Hypothesis test according to the data distribution and type of comparison needed using R
    • Use linear regression and correlation to explain relationships between metrics
  • Data Wrangling & Programming with R
    • Do common operations (conditional subsetting, joins, updates, calculations by group) using R data.table with minimal need to check for syntax on the web
    • Automate repetitive tasks using R programming elements (variables, control statements, data structures)
  • Apply Data Visualization in Excel
    • Be capable of choosing the most appropriate visualization for common business requests (evolution over time compared over 2 or 3 categorical dimensions, relationships against multiple variables)

 

In order to reach these goals I used several strategies:

  • In order to gather theoretical knowledge
    • I did online courses on Coursera and edX to get introduced to certain topics such as R Programing
    • I read articles from various blogs to get a broader understanding of the domain and keep up to date with the latest technologies. A blog post on Analytics Vidhaya introduced me to the data.table package in R.
  • In order to apply stuff:
    • I shared with my manager my learning interests so that he can offer me projects on that area. This for instance has lead to an inferential statistics analysis whether people order less frequently depending on their rating of the experience with the claims process.
    • I came up with various ideas or metrics on how to respond to certain analytical questions. For instance I offered to do some Control Charting in R for a warehouse analysis my manager was interested in.
    • I tested out some prediction analysis on the data I had access to. This meant a thorough understanding of both how to do the analysis and what the data says about the business process.

 

You may notice that I was keen on creating the context or the exercises to apply what I was learning on a theoretical level through reading and online courses, even though the job description did not necessarily imply it.

 

Also in order to reach the goals I’ve designed myself some learning habits. James Clear, a great blogger, entrepreneur and bodybuilder says: “The goal is just an event — something that you can’t totally control or predict. But the reps are what can make the event happen. If you ignore the outcomes and focus only on the repetitions, you’ll still get results. If you ignore the goals and build habits instead, the outcomes will be there anyway.” For that here’s what I did :

  • I did 1h of coursework every morning, 5 days a week
  • I ran code from StackOverflow step by step multiple times to understand how it works (for the problems I could not find a solution I Googled them and applied over what I wanted to do)

***

On this path I’ve uncovered some insights on my learning style that allowed me to improve how I acquire skills:

1. If I don’t know what I don’t know about the field, but it looks interesting I find a structured course already developed by someone else that covers the basics. It’s a simple way to get into the field and become aware of what I don’t know.

2. I get bored of learning the same thing over multiple weeks. That’s why now I run with multiple goals in different areas at the same time such as Data Visualization, Reporting, Data Wrangling. They’re all under the same umbrella of Data Science and they’re complementary, but they’re different stuff so I don’t get bored.

3. I enjoy the pressure of the assignments from online courses – even though it’s Self-Directed Learning online courses offer me some external accountability that helps me stay on track for medium periods such as weeks.

4. Doing some project on real data while searching online for solutions on the problems I’m facing is extremely valuable to ensure sedimentation of what I learn by theoretical means. It shows me how others have solved problems and I get to understand the way of thinking a certain problem.

5. Repeated use leads to memorization only if coupled with understanding – there were some lines of R code that I used over and over again, yet they didn’t stick in memory. Only when I went deeper into how does that code work, I became capable of recalling that from memory rather than from Google.

6. Learning is a project with tasks just like work. Therefore at some point I put it all in 1 place instead of multiple spreadsheets. For now I use Trello with the Repeater, Harvest and Card Aging PowerUps and linked with the desktop app Pomello to structure everything into 30 minute work, 5 min break Pomodoro sessions. The logic of this whole system will be subject for another article. There’s still something missing – an integrated space where I can have both reflections on the learning area and the time I invest in it. Sort of a Learning Management System combined with a Work Management System.

7. Online spreadsheets are great for storing information – I can write key findings from articles using my phone while on the subway. Plus it’s easy to access the original links using the search box.

8. Learning something is a long term goal. In order to keep motivation high I broke it into smaller pieces and found ways to get short term gratification such as “aha faces” from business people I present my analyses to, getting paid for the work or just asking more experienced people to review the quality of my work.

***

This learning journey lasted around 1 year. At that point I got a job as a Data Analyst in some other company. From my point of view, passing those job tests was an indicator that my level was good enough on the goals I’ve selected. Next up the goal is to use the same framework and ideas to become a Data Scientist.