3 answers
7
Questions
Asked
846 views
what is the work culture like in the data science field
#datascience
#career #computer-science
Login to comment
3 answers
Updated
Nicole’s Answer
Hi isys L. Love this most interesting question!!
I share my answer with you as someone who does not hold a data science title but I do data science work and I work with many data scientists.
Based on that, I can share that what I have experienced in the work culture includes a couple of things. 1)Most data scientists are generally excited about data. That can mean many things but in general, they are excited about what they have learned and what they have yet to learn "in the data". This excitement tends to create a culture of exploration that brings a collaborative nature to how data scientists work and share. 2)The culture tends to lean towards individuals who are independent thinkers and learners...so the exploration that I mentioned in item 1...it tends to lead the data scientist(s) into bringing forth ideas about how the data can be used (or if the data can be used) to solve a problem. 3)There are no shortage of data science tools to help in the learning and exploration process. Some are pretty deep and intricate...others are more user friendly. In an environment of learning, I have seen where taking the time to understand the tools that drive results and help data scientists (and their partners/clients) understand outcomes can save lots of time and can be helpful in communicated the benefits of a certain approach.
More broadly, the work culture can sometimes feel academic...like you are still in college...lots of drawing, visualizing, team building. It is important to remember though, that data science, on a whole, can be incredibly instrumental in solving problems so the idea is to get to a spot where individuals can use data science, explore...but eventually get to a workable solution to solve that problem.
Hope you find this answer helpful. Best of luck to you!
I share my answer with you as someone who does not hold a data science title but I do data science work and I work with many data scientists.
Based on that, I can share that what I have experienced in the work culture includes a couple of things. 1)Most data scientists are generally excited about data. That can mean many things but in general, they are excited about what they have learned and what they have yet to learn "in the data". This excitement tends to create a culture of exploration that brings a collaborative nature to how data scientists work and share. 2)The culture tends to lean towards individuals who are independent thinkers and learners...so the exploration that I mentioned in item 1...it tends to lead the data scientist(s) into bringing forth ideas about how the data can be used (or if the data can be used) to solve a problem. 3)There are no shortage of data science tools to help in the learning and exploration process. Some are pretty deep and intricate...others are more user friendly. In an environment of learning, I have seen where taking the time to understand the tools that drive results and help data scientists (and their partners/clients) understand outcomes can save lots of time and can be helpful in communicated the benefits of a certain approach.
More broadly, the work culture can sometimes feel academic...like you are still in college...lots of drawing, visualizing, team building. It is important to remember though, that data science, on a whole, can be incredibly instrumental in solving problems so the idea is to get to a spot where individuals can use data science, explore...but eventually get to a workable solution to solve that problem.
Hope you find this answer helpful. Best of luck to you!
Updated
Campbell’s Answer
Wow, @Nicole's answer is really good and thorough. I lead a team of people who are in the data field, including 2 data scientists. Some quick thoughts I have to add to what Nicole said (though she really nailed it)...
1. You will spend a lot more time gathering and cleansing data than building and performing mathematical models... in most cases (see item number 2 below). getting just the data you want and need will likely mean writing some custom SQL to get data from a database or performing some advanced vlookup functions in Excel or some other spreadsheet to get data organized in a way you can use it for your purposes.
2. Data Science means different things to different people in different companies. Sometimes they mean "Data Analyst", or someone who looks into the data to see what it is saying and how it impacts the business, but not most that is forward looking. It's a lot more like traditional reporting. Sometimes they mean "Data Engineering", or structuring and cleaning data so people can draw useful conclusions from it. This is often called an ETL Developer, a Data Wrangler, or some other term for someone who gets the data and puts it in a spot it can be used. Other times, they mean what I would consider a true Data Scientist, which is someone who takes the data, reviews it, determines predictive (or prescriptive) models to be used, determines the best model, and then puts things in place to ensure the models continue to run and stay accurate (champion challenger recurring evaluations). If you happen to find yourself lucky enough to be in the final role, remember to thank the other two data professionals who help make the job easier and more specialized. :)
3. Most of the really fun models are difficult to implement. There's a couple different reasons for this. If you ever try and explain a neural network to an executive without a background in math, you may find it difficult for them to grasp the concept. Even if they get it, you may have to talk it through with a bunch of their subordinates. yeah, it's politicking, not math that causes a slowdown here. because of this, linear regression and logistic regression (both or which are still super useful) tend to be defaults that are used. People also tend to get nervous when you get into unsupervised learning models. This is because they can miss the "human element" and can make decisions based around optimization (which they are supposed to) instead of common sense. Too many people say movies about computers taking over the world when we were younger, but there are realistic reasons that you want and need to have controls in place for things like this. Unfortunately, it means balancing the math with controls, so it is a balancing act and takes some oversight. For these reasons, it is typically easier to stick to simpler models that people understand, but you can do things like add in several models and weight them so the output can be multi-variate as the result of several single variate models. You can also add in seasonality weights (as people tend to understand these) or even use things like an XG Boost once you explain it to the customers.
4. If you want to set yourself apart from the rest of the Data Scientists, understand when something is good enough. In academics people get warned about overfitting models, but then get taught methods of evaluating model optimization. In Business, there is another metric to be aware of: the point of diminishing return. If you are asked to predict the number of Website clicks per day something gets (or the propensity to buy and item, or the lifetime value of a customer, or...) then the difference between 562 and 567.493526 is typically not worth the effort. Even if it is exactly correct, another 5 clicks (or dollars, etc.) is not generally enough to move the needle. And the effort that goes into getting that much closer is often less valuable than the output gain. We are taught to be the best we can, and mathematicians in particular tend to be perfectionists. We just need to remember when "good enough" is good enough. :)
Again, these are all things I see in my day to life life and may be a bit more detailed than you want. But they are tips to make yourself more competitive in a field where it is difficult to be called "wrong" as most of the time the real answer is that the analysis is just different. that said, Nicole's high level answer is a great one. ")
1. You will spend a lot more time gathering and cleansing data than building and performing mathematical models... in most cases (see item number 2 below). getting just the data you want and need will likely mean writing some custom SQL to get data from a database or performing some advanced vlookup functions in Excel or some other spreadsheet to get data organized in a way you can use it for your purposes.
2. Data Science means different things to different people in different companies. Sometimes they mean "Data Analyst", or someone who looks into the data to see what it is saying and how it impacts the business, but not most that is forward looking. It's a lot more like traditional reporting. Sometimes they mean "Data Engineering", or structuring and cleaning data so people can draw useful conclusions from it. This is often called an ETL Developer, a Data Wrangler, or some other term for someone who gets the data and puts it in a spot it can be used. Other times, they mean what I would consider a true Data Scientist, which is someone who takes the data, reviews it, determines predictive (or prescriptive) models to be used, determines the best model, and then puts things in place to ensure the models continue to run and stay accurate (champion challenger recurring evaluations). If you happen to find yourself lucky enough to be in the final role, remember to thank the other two data professionals who help make the job easier and more specialized. :)
3. Most of the really fun models are difficult to implement. There's a couple different reasons for this. If you ever try and explain a neural network to an executive without a background in math, you may find it difficult for them to grasp the concept. Even if they get it, you may have to talk it through with a bunch of their subordinates. yeah, it's politicking, not math that causes a slowdown here. because of this, linear regression and logistic regression (both or which are still super useful) tend to be defaults that are used. People also tend to get nervous when you get into unsupervised learning models. This is because they can miss the "human element" and can make decisions based around optimization (which they are supposed to) instead of common sense. Too many people say movies about computers taking over the world when we were younger, but there are realistic reasons that you want and need to have controls in place for things like this. Unfortunately, it means balancing the math with controls, so it is a balancing act and takes some oversight. For these reasons, it is typically easier to stick to simpler models that people understand, but you can do things like add in several models and weight them so the output can be multi-variate as the result of several single variate models. You can also add in seasonality weights (as people tend to understand these) or even use things like an XG Boost once you explain it to the customers.
4. If you want to set yourself apart from the rest of the Data Scientists, understand when something is good enough. In academics people get warned about overfitting models, but then get taught methods of evaluating model optimization. In Business, there is another metric to be aware of: the point of diminishing return. If you are asked to predict the number of Website clicks per day something gets (or the propensity to buy and item, or the lifetime value of a customer, or...) then the difference between 562 and 567.493526 is typically not worth the effort. Even if it is exactly correct, another 5 clicks (or dollars, etc.) is not generally enough to move the needle. And the effort that goes into getting that much closer is often less valuable than the output gain. We are taught to be the best we can, and mathematicians in particular tend to be perfectionists. We just need to remember when "good enough" is good enough. :)
Again, these are all things I see in my day to life life and may be a bit more detailed than you want. But they are tips to make yourself more competitive in a field where it is difficult to be called "wrong" as most of the time the real answer is that the analysis is just different. that said, Nicole's high level answer is a great one. ")
Updated
Ghazi’s Answer
Hello Isys, your question is a great one and the two people who have already answered have been spot on in their answers. I would say in my experience where you work really sets the culture. You can work places where there is a lot of support and collaboration, or you might end up working at a place where your on your own and expected to provides results as a task. Do your research on the companies/ places you would like to work and that will help you a lot! That being said, the data science field is a great one. It can be incredibly fun and rewarding when building your "story" of what the data is showing. You could make a real impact for the business with your findings, and when it comes to presenting you can be creative in the way you visualize the information! There are a lot of online free courses that can help you peak into what the day-to-day tasks might be like as well, and that way you can get a feel for if you like it or not. I would suggest going to Udacity and similar websites to find free courses and start to learn. You might even be able to pick up basic skills in SQL programming, Python Programming, and Tableau visualization software this way. Best of luck to you!