Building the data-driven enterprise

How can we improve our activities? How can we gain more markets or take better decisions? Are we struggling with lack of data to sustain our AI and analytics initiatives? A data-driven enterprise is the logical evolution of a company, aiming at improving its services and business value that will leverage data to take decisions.

Auteur : Yannick Bruck, CTO – Fujitsu Luxembourg

The evolution resides in different factors: from the methodology to be used, to the adoption across all the layers of the company, and the usage of modern tools to support the approach.

A data-driven enterprise is a company that embraces data throughout its department and activities, collaborates around data assets, and leverage insights for business value internally and externally.

The 4 steps to build a data-driven enterprise

The approach is simple from a high-level perspective.

The first step includes a Discovery phase, where it aims to find all available data relevant for the business purpose, wherever it resides. The complexity will come from the evolving IT landscape, storing data on premise, in the cloud and even on edge in certain contexts.
Another challenge is the legacy burden of having applications and data in silos. Historically, departments used to have their own applications, datasets and files, of course very specific to their business purpose, but not much open to the other department, at least technically. Leveraging a Data driven approach will lead to cross information throughout the company, and provide a single view of a topic. For example, a single view of a customer will leverage identification data from a CRM, but also financial data from accounting and marketing data.

This phase is also important to assess if the Quality of Data is good enough for the business purpose. It means checking if all records have relevant, non-empty values and if there are enough different records. This step is highly accelerated by tools available on the market, as it can be highly time-consuming if it’s done manually.

The second step is to build a proper Data Architecture. There are many technical ways of doing it but the most important point is to address the business needs and strategy with the proper architecture. Common usage will not be performant or relevant enough for IoT streaming data for example. Data virtualization solutions will greatly help here. They will enable the creation of a central abstraction layer, exposing data as business assets to be used for business outcome without the technical constraints behind it and by breaking the silos.

Moreover, as a third step an information catalog should be implemented here to let the enterprise manage centrally governance, collaboration and data protection rules (identifying sensitive information, masking part of data, etc.). This layer will also be responsible for exposing data with business terms easy to understand by end-users, and not exposing the technical names of fields. Enriched with explanations on the quality, usage and source of data, it is a very valuable source to enable adoption and trust to derived insights

The fourth and last step is to get insights out of such data. Many companies think they should start first with this step, as it is more directly linked to business outcome and value. But in the end this approach will fail, because no matter the high quality of your analysis or AI model, it still needs to rely on the quality and availability of data to bring a proper result. If your data foundations are weak, you will not get any actionable insights and will not be able to use such analytics in production.

Start small, grow fast

Our approach is different: indeed, we start small and with a business purpose, but we tackle all the steps together as we selected tools and methodology to deliver them fast and with a modular approach. This will bring the ability to directly have results and models deployable in production.

It is more important to secure the foundations, than to hire a team of data scientists with few models to be really used in processes. Technology has evolved enough to provide AutoAI features to let you benefit from it with good data. You will get more with this approach than with a better model and lack of (good) data.

As concluding words, it is now time to embrace this data-driven change, to start with small use cases and data sets (even excel files), to infuse AI and advanced analytics insights into each department of a company, supported by an internal or external competence centre to support pilot projects and manage the central information catalog. For more business value, and a more Intelligent Society.