What is Data Science?
... and will we know it when we see it?
TL;DR
Data Science designs, builds, and maintains Data Products that solve a need or a problem in a company.
Data Products are applications of data to facilitate achieving a business goal.
Data Products are built through an iterative and structured process that starts and ends with business goals and end users – the Data Product Cycle.
Data Science is a relatively new field. We spent the last two decades trying to define what Data Science is or what a Data Scientist does, and we still do not have a definition everyone agrees on. The discussion will probably continue for many more years, just as it did when other multidisciplinary fields were created – like Operations Research after WWII or Statistics after WWI.
Even if we don’t agree on a definition for Data Science, we can answer a more useful question: what does Data Science do? More precisely: what are the outputs of Data Science and what purpose do they serve?
The answer is surprisingly simple: Data Science designs, builds, and maintains Data Products that solve a need or a problem in a company.
By extension, Data Science is a tool to run a business better.
Will I know a Data Product when I see it?
So, what is a Data Product? Put simply, Data Products are applications of data to facilitate achieving a business goal. Let me give you a little taste of the four classes of Data Products you may encounter:
dashboards – think of the graph you consult to get real time updates on sales for each one of your stores
answers – think of all the questions you needed answered during the pandemic: did TV viewership increase because lockdowns? (causes), has streaming activity increased linearly or exponentially (explanations), what would ad revenue be if lockdown ends in one, two or six months? (scenarios), how many TV viewers should we expect for the next upcoming football games (projections)
deployments – think of the Excel spreadsheet with updated sales forecasts that is emailed to you every Sunday night
custom-built apps – think of the app that analyzes video of amateur soccer players and reports on their strengths as professional players in a mayor league team
All of these are Data Products, and all of them are outputs of a Data Science Shop to address a particular business need or problem
The playbook to build Data Products
While they day-to-day of most Data Science professionals focuses on the technical aspects of building Data Products, the process to create, build and maintain Data Products is much more than a purely technical task. Data Products are designed and built through a process that starts and ends with business needs and end users. We call it the Data Product Cycle.
The Data Product Cycle is kicked off with a crisp definition of the problem to be solved and an assessment of the appropriateness of available data in a diagnosis stage. Three more stages follow: a development and implementation stage where the Data Product is designed and built, an adoption stage where obstacles to embracement are identified and dealt with, and an evolution stage where new or overlooked needs are uncovered and incorporated in the Data Product roadmap.
A Data Product is not a one-time output, but a solution that evolves constantly. For that reason, the evolution stage at the end of the Data Product Cycle opens the door to a new diagnosis stage that starts the next cycle. Because you always want a better product that is synced with change in your business, the process must be cyclical.
In a nutshell, the Data Product Cycle is the playbook to build impactful Data Products that create value. The intuition is distilled in this animation:
What Data Science is NOT!
For some time, everyone with the title “Data Scientist” argued that what they do – and only what they do – can be called Data Science. So, for a while, there were almost as many definitions as there were Data Scientists.
Inevitably, that left many misconceptions we should dispel:
Data Science is not just Machine Learning or “Artificial Intelligence” (AI) - Machine learning and AI are some of the tools used in Data Science, but there’s a lot more to Data Science. Just as we would never say that the secret to a Michelin-starred restaurant is the pots and pans it uses, there is no good reason to say that Data Science is the tools it uses.
Data Science is not Analytics - think of Analytics as the sonar in a ship – it warns of immediate conditions and dangers. Think of Data Science as the navigation system in a ship – it uses advanced science to chart the optimal route to your next ports and takes care of logistics. The questions they ask are different. The methods they use are different. The outputs they produce are different.
Data Science is not external to a business - Data Science receives purpose from a business, uses data generated in the business, responds to the business strategy, and evolves with the company and the industry. To be effective, Data Science must take part in any process that powers a business so it can design and build appropriate Data Products that evolve with your business.
Data Science is not a silver bullet - bringing Data Science to a business will not magically solve problems. Data Science creates value when it operates through a blueprint anchored in critical thinking: that’s why the Data Science Shop was created.
Did we spark your interest? Then also read:
Data Science Shop: why this Project? to learn more about the story behind the creation of the Data Science Shop
What is the Data Science Shop? to learn more about the roadmap for the operation of Data Science in a business