How to Ask the Right Questions As a Data Scientist?
Have you ever thought about how to pose the right questions while working on a data science project? I got this question after joining my current company, where I closely work with about 25 data scientists. I have noticed that some really great data scientists, apart from having strong “hard-skills”, are also good at asking questions. These great data scientists tend to attend to the problem of why and so whatbefore diving into how.That is why I decided to learn the skill and share it with those, who might find it helpful as well.
This blog post explains a tool on how to scope a project using a framework from social science named CoNVO, which helps to structure and systematize our thoughts. Structure gives us room to think through all the aspects of a problem and come up with knowledge and interesting insights from data.
All stories have a structure, and project scope is no different. Like in a story, a project has an exposition (context), some conflict (need), a resolution(vision), and hopefully the happy ending (outcome). Let us dive into the storytelling in data science by understanding each part of the project scope.
First of all, mnemonic CoNVO stands for
Every project has a context, which comes from understanding who we are working with and why they are doing what they are doing. We learn it from continuously talking to people until we understand their long-term goals.
This department in a large company handles marketing for a cosmetics manufacturer with a large online and offline presence. The department’s goal is to convince new customers to try its cosmetics and to retain existing customers. The final decision-maker is the VP of Marketing.
Overall, context provides us with a larger view of a project and helps us to keep a focus. Also, the context includes larger relevant details like deadlines, which is helpful in prioritizing our work.
As we know, any data science project aims to create knowledge. Hence, a data science need is a problem that can be solved with that knowledge,not a lack of a particular tool.
It is important to distinguish the difference between needs and tools for solving problems. A tool is used to accomplish things and does not have a significant value except as an academic exercise.
Examples of tools: a dashboard, a predictive model, and etc.
By contrast, a data science need is a need that can be met with data, which is essentially about knowledge, about understanding some parts of how the world works. Data fills a hole that can be only filled with better intelligence, a better understanding of how things work. When we correctly define a need, we can clearly find out what can be improved by having better knowledge.
Example of a need:
The marketing department at the cosmetics company does not have a smart way of targeting people on social media. Currently, it is selecting targets based on intuition, but they believe there is a better way. With a better way of targeting people, the department expects sales will go up.
Note that the need is never something like, “the decision-makers are lacking in a predictive model”. This is a potential solution, not a need.
There is one point in defining a need, which I think is challenging and I personally want to master. The point is framing the need that way that you may get particular actions out of that. For example, instead of saying “The manager wants to know where users drop out on the way to subscribe to a service,” consider saying “The manager wants more users to finish their journey for a subscription. How do we encourage that?” Answering the first question is a part to answer the second, but the action-oriented formulation opens up more possibilities such as testing new designs and performing user interviews to gather more data. Of course, it is not always possible to state a need that action-oriented way, but at least it should be related to some larger strategic question.
