The future relies on data. In 2019 and with the rise of AI as omnipresent recipe for success, this is a common fact. Data-driven organizations are faster in decision making, more efficient in production and achieve higher process quality.
The reality in most companies, however, looks different. Data management is highly complex, and organizations did not yet adopt or actively maintain a data culture. In addition, many systems cannot cope with the non-functional requirements of modern data management. These include data security in cloud computing (especially in financial services), national regulations in integration of global business units or frequent M&A activities as well as carve-outs.
To start with, organizations have to break mental silos within the people they work with and make data a priority. This can be achieved with the next steps.
Central version of truth
Data is scattered across the organization. Multiple versions of the same data exist in various sources, which constantly poses the question for the version that is the ultimate truth. Decisions may be taken based on outdated information, compliance audits may fail due to use of the wrong data and trust in the data erodes completely with internal as well as external stakeholders.
That is why the most important reference points, i.e. master data, reference data and meta data have to be managed in a single, easily accessible source. Data from various systems have to be consolidated and harmonized in that reference system. All business initiatives need to draw their data directly from that source in order to build a common ground for decision making.
Data governance executes the operating model, establishes global data structures and standards, a global data organization and preparation of data for strategic initiatives in alignment with the corporate vision and strategy.
Governance is established on three levels: operational, tactical and strategic. On the strategic level, decisions on data initiatives and co-initiatives are being decided upon, such as new data initiatives and collaboration with other domains such as analytics and BI. The tactical level is the executive of new initiatives and escalation of operations, for instance, deciding on new standards for data with wider impact. The operational level consists of data stewards that approve new data items and their lifecycle and deal with day-to-day escalations.
The structure of the data governance bodies has to be clearly communicated to all stakeholders of the organization. Governance also has to make sure that people are informed about important key performance indicators (KPIS) of data and its performance.
A data dictionary describes data objects in semantic context and provides additional meta data as well as information about ownership and data quality metrics. Data lineage is made transparent to the extent of how a business data object is represented in various systems on database level. A data dictionary is made available to everyone in the organization as codified central reference and lookup. However, data definitions should be defined and agreed on based on a bottom-up approach and getting buy-in for common definitions is a challenge per se due to the existing different definitions across domains and business units.
Mature data organizations embed their business data glossaries directly into the intranet landing page for quick lookups. Any initiative, be it marketing campaign segmentation with customer data or large financial transformations involving multiple data objects, has to be able to get data and data about data easily to adhere to its structures and standards.
Broad data access
Broad is a broad term per se, and such is access to data. Various factors, mainly information security, play an important role when access is granted to various stakeholders. But access is not only a data dump or an API (application programming interface) endpoint - access is also visualization of data in form of dashboard or integrations with existing applications.
It is also no longer be enough to provide data with focus on local, i.e. customers only in a specific market. Advanced analytics through machine learning technics rely on as much data as possible to refine their results. Global centralized data lakes provide various forms of consuming data, e.g. through APIs or linked data endpoints, that can be made available locally.
Not everybody will be a data scientist, but everybody should be enabled to have a basic understanding of how to make use of the data provided and take the right decisions. The three major dimensions are statistical skills that can reach into machine learning territory, data visualization and data story-telling.
Number-crunching using statistical methods and models is essential for most data workers nowadays. But also simple A/B testing can provide valuable insights compared to more advanced predictive models in non-critical business processes. Visualization skills are important to make insights comprehensible by a broader audience. From choosing the right chart type in Excel to creating complex infographics, visualization is one of the easiest forms to describe complex results and scenarios. Eventually, bringing both together, data story-telling is the art of creating analyses and visualize them to tell the story behind data, e.g. to get buy-in for an innovative project.
Even though most essential in the century of AI, data-driven cultures are not readily available and often enough corporate culture hinders organizations from establishing them. Personally, I have seen many data initiatives fail due to the fact that organizations do not foster a culture of data sharing and caring. Taking the right steps as outlined above with a clear vision will support the process significantly though. To spice things up, create challenges like internal data hackathons to see what people come up with based on the data provided to them. After all, getting a hands-on grip on data makes it less of a theoretical exercise.
Thanks for taking the time to read this short article. Contact us if you would like more details on how to put this into practise.