Winding path through woods

Photo by James Forbes on Unsplash

As we recently wrote in our first post on Serious Data Science, there are numerous challenges to effectively implementing data science in an organization. Many industry surveys warn that most analytics and data science projects fail, and most companies don’t achieve the revenue and profit growth that they hoped their data science investments would deliver. In this post, we’ll examine some underlying causes of why this happens.

In our previous post in this series, we discussed how to tackle building credibility for your data science team. Here, we will focus on the challenges of quickly delivering real value with your data science team with a platform that supports an agile approach.

Data Science Development is Often a Long and Winding Path to Value

In talking with many different data science teams, we’ve heard that it takes far too long for a team to ramp up, perform analyses, and then share those analyses in an impactful way with their organization. This makes it challenging for data science leaders to deliver value to the rest of their organization, which in turn makes it difficult to justify buying new tools, hiring new team members, and investing in their training.

Several common obstacles make it difficult for a data science team to quickly ramp up and be productive:

Finding the Shortest Path to Value

Delivering data science value in an organization requires that your team be agile. While “Agile” usually refers to a very specific development methodology, here we use “agile” to simply describe a process that has four principles:

  1. Use what you have. To quickly ramp up and deliver value, take advantage of the existing knowledge of your team and your previous investments.
  2. Collaborate regularly. The users of the product continuously meet with and influence developers.
  3. Iterate on deliverables rapidly. Developers incorporate feedback into the product in short development cycles until it delivers what the users want.
  4. Deliver results frequently. The process routinely delivers new products for users to critique and improve.

Applying these principles to the data science development process allows your team to deliver value more quickly and efficiently. They help your team overcome the obstacles noted previously, and they demonstrate commitment to a Serious Data Science approach (see Figure 1).

Obstacles Solutions
Training required on new tools Use tools many data scientists already know
Monolithic applications take too long to set up and don’t use existing analytic investments Focus on small prototypes to prove value, using tools that integrate with your existing frameworks
Slow exploratory development Rapid, code-based development
Difficult to access and share results Deliver live results directly to stakeholders

Figure 1: Use agile principles in a Serious Data Science approach to address common development obstacles.

To make your team more agile in your data science development process, we recommend that you:

Astellas’ Aymen Waqar discusses the analytics communications gap:

 

Learn more about Serious Data Science

For more information, see our previous posts introducing the concepts of Serious Data Science, and drilling into the importance of credibility. In the coming weeks, we will round out this series with a post on how to make sure the value your data science team provides is durable.

If you’d like to learn more, we also recommend: