Office Hours

We spend 50% or more of our adult life on our Careers. We make many decisions that have a direct or indirect impact on our careers. It could be about switching jobs, moving to a different city, choosing a hobby project, or a data science topic to learn. Well-thought-through decisions ensure a successful career. It is natural for every one of us to make some mistakes. But there isn’t enough luxury to make too many wrong decisions as each wrong decision costs time and money. Here I am going to show some simple tools that you can adopt to ensure a successful career in data science.
Mental models are an amazing set of frameworks that help in clear thinking. It helps in making a better decision. In this article, I am going to share some mental models and about embracing them.
How can mental models help in enhancing your data science career?
Adopting a mental model will help to train your brain in critical thinking. This helps us in better understanding the problem and coming up with a better solution. This is the secret to the success of a Data Science project. Yes, your understanding of the problem is more important than your technical capabilities.
Mental Models also help in better knowing ourselves and our goals. It helps in getting answers to important questions like,
- What should I up-skill to increase my earning potential?
- What data science role is more suitable for me?
- Do I have sufficient opportunities in my job to grow as a data scientist?
- What should I look forward to in my next job?
- What types of data science role should I apply for?
- What are my strengths and weaknesses?
We are going to see in detail about adopting the below 5 popular Mental Models,
- First-Principles Thinking
- Second-Order Thinking
- Inversion
- Map vs Territory
- Feynman’s Technique
First Principles Thinking
This framework helps in a clear understanding of a problem to its fundamental elements. I got introduced to this concept while reading the book Elon Musk by Ashlee Vance. Ashlee explains about Elon had an option of buying a rocket for SpaceX at about $65 Million. But, he chooses to use first principles thinking to build it himself at one-tenth cost.
To adopt first principles thinking follow this simple approach
- Define the problem to the best of your ability
- Break down the problem into its fundamental elements
- Try to change one or more of the fundamentals
- Build a new solution from scratch
Another popular example of first principles thinking is the invention of the rolling suitcase. People have always been carrying weights over their heads or on their shoulders. Around 400BC, we started using bullock carts for transporting heavy loads. But only in the 1970s wheels were attached to a suitcase to carry personal belongings
First-principles thinking can be handy to solve most data science problems. It helps in better understanding the problem and coming up with the best possible solution. To learn about applying first principles thinking to data science problems read the article below
How to use First Principle Thinking to solve Data Science Problems?
Second-Order Thinking
Most people try to just solve the immediate problem. There isn’t much thought process about the consequences. Anyone can come up with a solution to a problem. But to come up with an ideal solution one needs to think about the consequences too.

Second-order thinking can help to see beyond the solution. Amazing results can be achieved by seeing what the majority are blind to. It will be fine to not adopt other mental models but failure to adopt this could be costly. It is very easy to adopt this, always ask more questions and think what next?
One famous example of failure to not use second-order thinking is the Titanic tragedy. There weren’t enough measures in place to handle complications. The ship which carried more than 1500 people had only 16 lifeboats, just enough for 1/3rd of the people on board. The crew was not equipped with proper lighting and didn’t have access to binoculars. So failure to use second-order thinking could result in a really bad consequence.
In the data science context, while working on recommendations to business teams. Always think about all possible outcomes. Some outcomes would be immediate and some could take a lot more time. But, the final solution should be chosen after considering the outcomes. For example, one way to increase customer retention is to offer discounts to customers with a higher propensity to churn. Though, it is most likely these customers might continue to stay. But we need to understand the consequences like, What impact it might have on other customers? How long will this discount help in retaining customers? Will it result in a price war with the competition? Is this strategy sustainable in long run?
Inversion
This mental model might not lead you to the best solution. Maybe not even close to it but it will ensure that you avoid the problems. I find this approach can sometimes be enlightening.
Let us consider a more relatable example, your plan to learn a new data science topic. There could be multiple paths to reach your goal. You can sign-up for a course, you could learn from a friend, you could work on a project and learn by doing things. It might be confusing to choose the right one, like what are all the different courses? Alternatively, what hobby projects can help me to learn the topic? The inversion method is about stepping aside from the normal way and start thinking about the opposite. Like, what might stop you from learning this topic, things that could bore you soon. Thus inverse thinking not only helps in identifying the obstacles but also helps in eliminating less suitable options.
This method can be adapted to any data science problem. For example, let’s say the problem statement is, increasing the customer base. We should think about strategies that could help in acquiring new customers. But along with that, we should use inverse thinking to know about the triggers that could cause customer churn. As otherwise, even though new customers are being acquired there could be many existing customers churning. So the recommendation to business must include directions to acquire new customers and strategies to prevent customer churn
Map vs Territory
This technique is about recognizing the difference between a map and a territory. A map provides a higher-level view based on a snapshot at a point in time. The territory shows the actual reality.

Here when I say map I mean
- The strategy we design based on the experience
- Recommendations from the case studies or white papers
- Step-by-step guides
And when I say territory I mean the reality in execution. A strategy’s previous success doesn’t guarantee future success. Also, just because many people are following a step-by-step guide doesn’t mean that it will help you to achieve your goal.
It doesn’t mean we need to disregard the strategies based on previous experience or from books. The key is we need to understand that there is a need for customization based on reality. For example, the popular step-by-step guide to learn data science might not work for you. Maybe because of your personal commitments, objectives, and level of expertise. But if you pick the useful insights from the guide and adapt them to your plan. There is definitely a better chance of being successful.
Below are some set of questions that will help you to adapt the Map (Strategy/Guides) vs Territory (your reality) Strategy
- Who created the map? And Why?
- When was this map created? Is it too old to be used now?
- What can go wrong with the map?
- How to make the map more usable?
- What should be ignored from the map?
- What are the assumptions considered in the map?
These questions will help you to understand if the map we are talking about is still usable or what changes it would need to make it more adaptable?
In a data science project context, let’s say you have been successful in reducing customer churn for online merchandise. It doesn’t mean you would be able to easily fix the customer churn problem for all merchandisers. It depends upon the customer base, their profile, touchpoints, and other patterns. The learning that comes from success is good but you need to adapt them to the new territory to ensure success
Feynman’s Technique
Feynman is one of the greatest physicists of all time. He is known for breaking down a complex problem and explain it in very simple terms. This technique can be used to better understand any complex topic.
In a data science career, there are always plenty of scenarios where Feynman’s technique could be applied like,
- To better understand the new topics you are learning
- To arrive at a simple narrative for discussion with the business stakeholders
- To design a data science training program for the new joiners in your team
Below are simple steps from Feynman’s techniques that can be applied to your scenario,
Step 1 – What do you know about the topic?
- Write down everything you know about the topic.
- If it is the first time you are learning about the topic, spend some time reading about it
- Write down your understanding of the topic.
Step 2 – How would you explain the topic to a tween?
- Keep things simple enough so that they could be understood by a tween
- Keep your narrative short because kids have a shorter attention span.
- Don’t have any jargon and refrain from having any complex terms
Step 3 – Understand the gaps
- Based on step 2, understand the areas where you lack understanding or need more reading.
- Work on your gaps until you are comfortable to explain or talk about them in simple terms
Step 4 – Refine your narrative
- It is time to put things in the right sequence and come with a narrative.
- If you don’t have an audience then share it with your friend and get their feedback
To stay connected
- If you like this article and are interested in similar ones, follow me on Medium
- I teach and talk about various data science topics on my YouTube Channel. Subscribe to my channel here.
- Sign up to my email list here for more data science tips and to stay connected with my work