5 years ago

Rethinking Data Science Teams: Part 1

Data science teams can be likened to an assembly line. One specialist is in charge of collecting and gathering the data, another is going to analyze it.

Much has been said about businesses and data science. There are currently several ways in which data science can add value to an enterprise.

  1. It gives management insights and information that lead to better decisions. Data scientists act as a trusted advisor and strategist for the C-suite. As they measure, track, gather, and analyze data, your executives can have more updated information to work with when they are making decisions.
  2. Understanding trends that can, in turn, help define your business goals. A data scientist can explore your enterprise data and recommend avenues for better customer engagement, performance, and profitability.
  3. Encourage staff to adopt best practices. Data scientists can ensure that your team is familiar with your analytics product so that they can use it and get valuable insights to do their work and help them address business challenges. It also helps them identify opportunities, learn new methods of doing things, and improve on processes.
  4. Decisions are no longer based on emotion. Data science fuel business decisions that are backed by solid evidence. It can also simulate results of possible actions, letting key players see expected business outcomes.
  5. Find new target markets. Businesses are always gathering customer-related data. They do customer surveys and get customer feedback, even from digital footprints of Google Analytics. Data science can help identify specific demographics that you may have missed. For instance, you might find out that most of your Web site’s visitors come from California, or you might discover that certain products are popular with millennials.
  6. Finding the right talent. Data science can help you identify various resources that you might have missed. For one, it can help you hire the best employees for your company. Data science can mine information available on corporate database, job search sites, and even social media.

Even with all of these benefits, however, not everybody is jumping onto the data science bandwagon. In 2017, Forbes.com found that only a little more than half of the companies, or 53 percent, are looking into big data. Two years before that, only 17 percent of companies expressed interest in big data analytics.

The most popular use cases for data science is data warehousing optimization, customer analysis, and predictive maintenance.

Getting your data science team together

One of the first steps to tapping data science for your business is to invest in a good data science team. To outsiders, data science teams are composed of — no surprise here — data engineers. Data engineers are excellent programmers and have hardware skills that allow them to build your data infrastructure.

In real life, however, data science teams are more complex and have different professionals working for them. A typical team would consist of data engineers, research scientists, contributing inference scientists, machine learning experts, and so forth. You have all of these specialists working for different goals and are coordinated by the product manager.

Division of labor

Division of labor has been a staple in all productivity scenarios. You have an assembly line of workers all doing different functions. Each of these workers has their own task to do, and they do it repetitively. In time, they get very good at it.

What happens is that the entire line is very efficient, each part of the product is well made, and when taken together, you have a durable and well-made finished good. The assembly line is an excellent way to mass-produce consumer goods.

For instance, a bottled beverage. One worker will be tasked to clean the bottles. Another will be filling the empty bottles with the drink. Yet another will screw on the bottle cap, before sending it to another worker who has to seal the cap.

The assembly line gave rise to very efficient factories that in turn, led to shortened production periods, lower equipment costs, and more skilled workers. As you can see, data science teams can be likened to an assembly line. One specialist is in charge of collecting and gathering the data, another is going to analyze it, while others put it to use depending on the use-cases that you have.

But is that necessarily a good thing?

Why the โ€œassembly line wayโ€ of doing things is not a good fit for data science teams: data science is not for production and execution

There are several reasons why we should not strive for a data science team that performs like an assembly line. For one, data science teams should not be optimized for productivity.

In manufacturing, you have an end product in mind. In our example, you know that you want this quantity of bottled beverages to sell. For that, you have a clear view of the requirements you have to make one bottled beverage. Workers can execute on each one of these requirements as efficiently as they can.

Data science is slightly different. You do not just execute the requirements. Data science is there to give you insights on and develop new enterprise capabilities. There are systems, such as client engagement, style preference classification, fashion design systems, seasonal trend detection, logistics optimization, size matching, and other algorithmic services and products that you cannot design upfront.

Most data-based services and product need to be learned. You do not have blueprints or concrete plans to get them. There are a lot of variables, models, parameters, and other elements that will need to go through experimentation, iteration, and trial & error.

Unlike in production, data science requires you to learn as you go. You cannot learn about the products and services that you can get from data before you start anything.

What’s more, the assembly line thinking kills improvisation and critical thinking. Workers in an assembly line are not expected to think out of the box. They are just required to do what they are supposed to do. Success is measured by how efficient each cog in the machine is.

Also, in an assembly line, you need to be consistent. This is why there is no room for you to improvise.

But you cannot expect the same kind of thinking with data science. Data science evolves. Your goal is not to produce another finished good that is similar to all the other products you have before. Instead, you strive to learn.

Stay tuned for the second part of this article.

Photo courtesy of NASA HQ PHOTO.

Scroll to Top