Data Science: from Strategy to Production
Having a digital roadmap with dozens of big data projects identified is one thing, having those analytics reliably running in the cloud is another challenge.
The purpose of this article is to highlight some of the obstacles companies are likely to encounter when filling the gap between those two worlds, and how Capgemini Invent can support companies in such situations. There are mainly four topics to be addressed: the company’s strategic vision itself, the technology stack used, the structure fostering such projects and finally the delivery process. Depending on the situation, some aspects will bear more importance than others.
This introductory post will serve an overview of the four topics. Each topic will then be covered individually in subsequent articles. Although those elements are presented in a sequential way for clarity, in reality they often evolve in a concurrent fashion.
Let’s start with the strategic vision, that many companies now have, because they know that conducting a digital transformation is no longer an option. This first cornerstone defines what the company expects from data-oriented initiatives and what are the most important projects to apply AI on. It is critical this vision be aligned with the other mid-term objectives of the company, as well as compatible with an internal or external demand. For example, a company planning to build a digital platform to sell services to its customers without conducting thorough market research first, would make a risky bet. Maybe its customers are not ready yet, maybe its customers’ business is not compatible with a platform approach. The company should then focus on a different target, maybe developing an on premises solution or edge computing. Those two examples may sound silly but are drawn from real examples collected when advising our customers.
The pre-identified data initiatives then need to be supported by the appropriate technology stack. The options here are various, from creating one’s own data stack using existing bricks, to relying completely on an external provider such as AWS or Microsoft Azure. There are pros and cons for each approach and Capgemini Invent helps its customers to choose the best solution for their needs. The target here should be to make a wise decision and have a platform available quickly for developments. One temptation is to think of this as a simple task compared to one’s core business. A company building rockets may think that building a big data stack is much easier than their rocket-science. Experience proves that building such a platform requires very specific skills and can lead to huge overcosts and delays if not handled properly. Needless to say that when such a new IT stack needs to be integrated with legacy systems, as it is frequently the case, this requires even more caution.
How data initiatives are governed is also a crucial point to address. Business lines may want to run their own initiatives independently, or the company may decide to build a global analytics capability providing services to business lines. Our experience shows that the appropriate answer depends on the company’s goals and culture, and we are able to recommend the best solution for each situation.
The delivery process is the last building block, relying on the previous ones. A robust delivery process will allow to select the best use cases for a project, bring the appropriate team members to solve the problem and deliver on time robust results. All the previous words are essential because the pitfalls are numerous here. One unfortunate pattern that we have seen is companies building a team of junior data scientists without a strong technical leader, have them work on ill-defined use cases and after a couple of months reaching to the conclusion that bringing value with data initiatives is just another myth. It’s not, but it definitely takes experience, mistakes and continuous improvement.
This introduction skimmed over the 4 building blocks to successfully transform a data roadmap into valuable products. As seen, each block bears its own risks and required know-how. Capgemini Invent has developed an extensive methodology for its customers and has a proven track record of bringing AI to life for them. If you are curious about it and want to know how we could help you on your digital journey, feel free to contact the author of this post.
About the author:
Cyprien Henry, PhD is a Managing Data Scientist at Capgemini Invent France. He helps Capgemini Invent’s customers to leverage data science in order to solve business problems with high added value. You can get in touch here by checking the author’s LinkedIn profile.
