Product Science

2015년의 글을 이제야 보다니..!

- Product Analyst + Daya Science
- On daily basis로 Visualization을 하지만, 새로운 의사결정을 할 때는 modeling과 통계적 실험 설계하여 AB test 하는 사람?!

ㅡㅡㅡㅡㅡㅡㅡㅡㅡㅡㅡㅡㅡㅡ
https://medium.com/@sall/product-scientist-ffd1ae846172

Product Science is the central team responsible for every step of translating data into insights that improve the product.

What is the Product Scientist role?

Broadly, product scientists focus on analysis — interpreting data, sharing insights, and developing models that improve the product. We work together to analyze patterns at a macro level, answering questions the rest of the company hasn’t thought to ask yet.

On a day-to-day basis, you’ll spend half of your time with product teams: a group of engineers, designers, a product manager, and you, the product scientist, all working together in a small group to build new features. You’ll extract key data and findings, define metrics, evaluate A|B tests, identify opportunities based on the data, and share new insights no one considered before. You’ll be right there, brainstorming with the team, thinking ahead, ensuring we make the best data-informed decisions.

The other half of your time will be spent with us, the Product Science team, enhancing our general analytics. You’ll keep us ahead of the needs for those product efforts, making sure we instrument the data correctly. You’ll also build tools and algorithms (like reading recommendations, or total time reading) to expand the team’s capabilities. And — always important — you’ll simply explore Medium’s rich dataset (did I mention how rich the data is?).
ㅡㅡㅡㅡㅡㅡㅡㅡㅡㅡㅡㅡㅡㅡㅡㅡㅡㅡㅡㅡㅡ https://medium.com/alpha-beta-blog/is-data-science-really-a-profession-in-decline-e983d0c42680

as auto-ML tools mature, they make certain data science skills redundant. Does someone trying to make a prediction with a SVM model (Support Vector Machine) really need to understand its inner workings? Probably not. Once upon a time, data scientists could focus almost exclusively on building models to solve problems brought to them by business stakeholders. But in an auto-ML world, successful data science is much more about identifying the problem (requiring extensive business understanding) than model building and tuning. A consequence of this is that if company and industry specific knowledge is increasingly prized, it makes it harder for data scientists working in one industry to change to another one.

ㅡㅡㅡㅡㅡㅡㅡㅡㅡㅡㅡㅡㅡㅡㅡㅡㅡㅡ https://medium.com/@jacob.d.moore1/the-decline-of-data-science-4161b4f3595b

Data Science jobs still exist — they’ve just experienced a silent rebranding. Machine learning has been replaced with hypothesis testing; big data was replaced by SQL. But these responsibilities have been packaged up and shipped to their new home — software engineering under the flagship tole, the Machine Learning Engineer.

Machine learning aside, there are plenty of interpretable statistical learning algorithms — for example, the field of causal inference. Likewise, robust statistical models can account for causal relationships, inferring directionality and influence while controlling for confounders. This is the meat and potatoes of social science. If not machine learning, this is exactly how data scientists should be leveraged to maximize learning from data.
But oddly, these skills aren’t required of or utilized by the data scientists of this decade. A Bayesian hierarchical model gives incredible flexibility and inferential power to the statistician using it; but the upfront cost is high: Time. One could expect to invest a day towards a week to properly implement one.
Save for Google, today’s tech leaders aren’t interested in surfacing deep insights and the time commitment therein. Rather, we surface a multitude of insights, hidden right beneath the surface. And the language of these insights is SQL.
Over the previous decade, Product Managers brought deep domain knowledge to the table — data scientists, the scientific method and statistical inference. Together, they packed a powerful punch. Today’s data scientists are alleviated of the burden of understanding statistics beyond hypothesis testing in exchange for deep product and market intuitions.
And this is the essence of the Product Scientist.
A new role for a new decade. You know your value proposition of your product, the economics of your industry, and the customer segments of your target audiences the way a psychologist knows the mind and an economist knows the economy.
Moving forward, when you see ‘Data Scientist’ job titles, read ‘Product Scientist’. You’ll never be asked, ‘Does gender confound the relationship between age and spend?’ You’ll be asked to ‘Sum spend grouped by age and gender’ and let the aggregates speak for themselves, cofounders or otherwise.
For data scientists who wanted to bring the science to the field, this will feel like the death of a spectacularly interesting role. But that’s just one valid perspective; another is that now there is a seat at the strategy table for Product Scientists; you don’t need a Stats graduate degree — just hustle, product intuitions and SQL; with just these three ingredients, you can make tangible impact within some of the biggest companies on the planet — not bad, just different. Change is the only constant!