Analytical Diaries

Whenever I speak in front of expert communities about the holy grails of Business Intelligence and Sales Analytics the buzzword most often heard is single source of truth. But whenever I look at my daily practice with any of our clients the truth is rarely single. And usually the multiplicity of truths is due to one of the following:

• expertise along a certain function
• responsibility along a certain geography
• broken value chains that still honor Taylor’s ideas in the digital age
• ignorance of the big picture / complete business model
• the fear of losing independence or flexibility

So do we need to put the ideal of corporate Analytics to rest? I do not think so. Instead we need to find analytical practices that preserve the benefits that private truths offer while at the same time, unify data, calculations and processes.
We have come across a new practice recently that aims at merging corporate integration and functional or personal flexibility. The main architectural components building the framework of this ideal are described in a compact mode in this and in a more detailed fashion in the sub-sequent posts.

• 4 Layer data model – build up your data architecture along a four layer fashion from the start. While the first layer mirrors the source systems in an original and 1:1 mode the second layers starts to form certain concepts of a common denomination. For examples this second layer may form one common concept of clients or products based on the multitude of client or product representation that are living in the operational systems of records. The third layer then builds one integrated and common data model based not on the operational layers but on the second layer of abstraction. This will help to keep both the common data model and the operational data independent and flexible from each other. The fourth layer in turn is used to translate the “single truth” semantic into detailed and function specific views, replacing the former private truths with new but integrated versions of expertise, geography or process steps.
• Common Master Data – a data governance and data stewardship framework that aligns the main dimensions of the company along agreed standards. While you may continue to have multiple views and flavors of clients and products a common and central logic needs to exist, that links the private to the corporate truth.
• Federation capability – instead of developing and clinging to hard and inflexible ETL processes and logic, the entire transformation logic from operational data store to the final report (across all 4 layers of data models) will be entirely transparent and non-persistent. Instead of copying data around, the federation component will translate and access on the fly and create an evolving enterprise virtual data warehouse.
• Central Data Store – while federation might take the place of traditional ETL there still is the need to create a central analytical data store over time. To preserve reporting, KPI and performance management history data and the point in time snap-shots of shared data the former data warehouse will evolve into a more operational role – both looking forward as well as looking into the past.
• Open Reporting Layers – and at last we should refrain from trying to standardize and centralize reporting in any medium to large scale organization with distributed responsibilities. Instead we should make sure, that the important logic is not refined to the reporting tools but made re-usable by applying transformation logic within the third or fourth layer within our data architecture. And have any tool use the same baseline.

Obviously even this advanced approach towards a truly integrated data management policy will take lots of effort and time. But it can at long last bridge the gap between tactical benefits and strategic direction. And it will let us work towards a single truth yet supporting all the private truths on top. Feel free to discuss…