What can Data Science do for you?
Data Science has been all the buzz in the last few years due to the rise of “Big Data” and “Machine Learning”. The Harvard Business Review named Data Scientist “the sexist job of the 21st century” back in 2012. According to Glassdoor, Data Scientist Leads 50 Best Jobs In America back in 2019. It seems like everyone and their mother wants to get in on the data science hype, hiring data scientists to tackle their company’s biggest problems. Although data science as a field does have practical applications to most fields, there still exists a lot of misconceptions and confusion regarding the field, primarily perpetuated by mass media and business people who aren’t precise when using nomenclature pertaining to data science. As a Data Scientist, I’d like to clarify some of the confusion and help business leaders understand how data science can help your company grow and succeed.
Data Science can be loosely defined as a field that leverages various datasets to scientifically extract insights for business decision-making or apply additional value to software product features. Data Science is a fairly broad field with various sub-disciplines, each focused on solving specific problems with slightly different technical skills and knowledge.
Data Analytics
Data analytics is a popular field that leverages data to directly extract various insights to evaluate performance or assist in making decisions. Under data analytics, there are generally two types of analysts: 1) business analysts and 2) product analysts.
Business analysts are tasked with unearthing problems or opportunities within a business department that can be observed through empirical data. Doing so can help business leaders make decisions for ways to optimize operational efficiency, increase conversion, lower costs, improve profits, etc. Business analysts can exist in Marketing, Sales, Finance, Operations, or any business-related department.
Product analysts on the other hand are slightly more technical analysts that work alongside product managers and engineers to figure out how to best improve and optimize the product for user growth, engagement, retention, and user experience. Product analysts are paying close attention to how users are interacting with their products, how the product affects them, what frustrates users during usage, what users find useful and useless about a product. Product analysts play a crucial in deciding the direction of a product, as product managers need data-backed feedback in order to iterate the product in the right direction.
Data Architecture/Engineering
Although data might be readily available to be used, it isn’t always easily accessible or in a desirable state; this is where data engineers and architects come in handy. Data architecting is the process of designing a system that is able to ingest, process, and store data for purposes beyond transactional/production usage. Data architects will first understand the business or product requirements. After that, they will translate the requirements into schemas, which are the blueprints for the entire architecture. Once data schemas are aligned with business leaders, data engineers will collect missing data, collate data from various sources, develop data pipelines, or ETLs, to feed streams of data into analytics storage, clean data by imputation or other algorithms to ensure completeness, and format data to be analytics-ready or other advanced purposes.
Although data architecting seems like the tedious part of the process, it is very crucial in laying the foundation for more advanced analytics and data science projects. Companies with a haphazard data architecture will often end up with inadmissible results, inevitably jeopardizing business decisions and business performance in the process. On the contrary, companies with solid data architecture will often extract more value than the initial effort put into it, creating more opportunities for growth and innovation.
Machine Learning and “AI”
Last but not least, one of the most popular, if not the most popular field right now is Machine Learning. Machine learning put simply is traditional statistical modeling married with advanced computational algorithms and massive datasets, or “big data”. Simple machine learning models can help make predictions inferred from training data, while more advanced machine learning techniques such as deep learning, reinforcement learning, or neural networks can make decisions or perform advanced tasks such as playing Go. In order to implement machine learning, companies must build the aformentioned data architecture, as well as define a specific problem to solve. Data engineers or data scientists can then begin preparing training data through a process know as feature engineering, whereby data is labelled and organized into different variables to help identify patterns in the data that correlate to a particular outcome. This process is know as supervised learning. Some problems don’t require labelling and can be used directly to identify patterns or insights; this is know as unsupervised learning. Data scientist are usually the folks that lead machine learning initiatives, specifically those who have a strong academic training, as machine learning done right usually requires a solid understanding in scientific methods, advanced mathematics, and computational algorithms.
Some companies might give data analysts the title Data Scientist, whereas Data Scientists are call Research Scientist. However, it is my experience that job titles are not the determining factor of what you will work on, as titles are just something to identity different personnel in a company.
Conclusion
Although there might be massive amounts of data in your company, a lack of understand how data is used, processed, analyzed may very lead to not possessing the data in the first place. Finding the right data person for the right task is also highly important, as well as knowing what stage of the your company is at in terms of data maturity. Hopefully, in an age where data is so ubiqutious, we can make data work for us, instead of us working for the data.
