Overview

With the explosion of data in the last few years, data science has become an ubiquitous tool that is desired across all indsutries. In Vince Ryan’s article “Data Science: Where Does It Fit in the Org Chart?” he discusses that many businesses that are looking to become data-driven often struggle. In a recent survey by NewVantage Partners, “only 38% of respondents from the Fortune 1000 said they have managed to create a data-driven organization. Even fewer — 27% — reported success at building a data culture within their firms.” Ryan points out that a major factor of implementing data science in a company is how the data science organization is structured within the company.

Methods

DATAx in New York

DATAx in New York

Ryan touches on four approaches that Dan Gremmell, vice president of data science at Plated, discussed at a New York data science event DATAx.

approaches = c("Functional", "Centralization", "Embedding", "Structured Embedding")
approaches
## [1] "Functional"           "Centralization"       "Embedding"           
## [4] "Structured Embedding"

Functional

One approach of organizing data talent in a company is to structure data scientists based on function. Each department within the company has its own hierarchy of data scientists. While this solution certainly appears pragmatic, it can isolate these groups of data scientists, decreasing efficiency. Gremmell mentions that this setup really only works for large companies.

Centralization

Centralization takes an opposite approach to the inefficient functional one. The idea is to isolate anyone working in data into one group. The main pro of this method is that it is beneficial for companies looking to do experimental work, such as “conduct[ing] a lot of research [or] advanc[ing] a technology.” Unfortunately, this tends to detach the data science team from the business application side of data science.

Embedding

This third approach attempts to strike a middle ground between the functional approach and centralization. In this method, data scientists are grouped into functional teams, but they have some specific business goal in mind, such as “growth, for example, or product.” Embedding seems to balance efficiency with domain knowledge. The one downside to this approach is that individual employees can get isolated in that they are too caught up in the business side of data science. Hence, Gremmell argues that this approach works best in small startups.

Structured Embedding

Structured embedding is similar to embedding in that it groups data scientists, analysts and engineers into groups for specific business tasks. However, the groups have a more rigid hierarchy or “structure.” It is essentially a more centralized version of embedding.

#This code produces an error, but I can prevent it from evaluating with code chunk options.
4 + b

Article Info

Publication.Date Views
Jan 7 5677