The client was under great pressure: his infrastructure was so outdated the vendor wouldn’t provide anymore patches nor security updates. To add extra pressure on IT management, the information infrastructure was a poster child of the late seventies MIS architecture and governance processes.
If the business partner needed a report, he had to ask the MIS department who would scramble a few tables in the source systems, create or update a semantic layer and give the user access to the report. Needless to say that this way of working was unacceptable for the business and led to a chaotic unmanageable information model. To add insult to injury: absolutely nothing was documented.
Emergency measures needed
The first task was to migrate the entire infrastructure to a recent software version, thereby eliminating as much as possible the anomalies in the DDLs and queries without changing the basic tables used by the semantic layer in order not to break things.
Four steps to build a roadmap
Lingua Franca follows these four steps to build a roadmap:
· Document analysis: studying the existing reports
· Observations: how people use the reports
· Workshops: to reach a consensus about handling certain issues
· Interviews: get the requirements straight
Analysing structured data models
Examining the existing reports it soon became clear that there were not too many facts and dimensions to be distilled from the data: about 20 facts and 45 dimensions would model the entire process of the organisation. Of, course that was the “as is” situation. Observing the report use provided clues for lacking functionality and information. It added marginal information about attributes so this was not really a change agent in the analysis process.
Extending the scope to a modern data architecture
The users staid in their comfort zone and were not too ambitious with their requirements.
Therefore, a workshop was needed. With the various managers involved we made them dream about new analytical opportunities without restraints like time, resources, budget,… Even then, these new requirements didn’t seem too farfetched.
Eliciting new requirements
Enter the interviews: in one on one conversations around simple themes like: “What do you need to accomplish in your job and how can information help you with that?” and “Wat are the decisions you need to take? With which frequency? What data would support your decisions better?” This led to a comprehensive wish list and helped us to make an inventory of internal and external data sources that would need integrating in a modern data architecture. This would be a mixture of classical star schemas for structured data and a medallion architecture for unstructured and semi structured data with the potential to support a data mesh architecture.
Drawing the roadmap
Using the data models for the star schemas and examining the APIs for accessing data for a lakehouse resulted in a comprehensive list of 187 tasks, each with their execution and throughput times as well as the capital expenditures.
Based on that detailed information, a PowerPoint for management and the executive summary was the easiest deliverable of this 55 day project.
If you need a roadmap for a modern data architecture, don’t hesitate to get in touch with us!
Commentaires