Lets Talk

When assessing our approach to achieving success within our organisation, it’s essential to recognise that growth and success necessitate change. Progress is unattainable without adaptation, and what works today might not be effective tomorrow. Therefore, we need to continuously review and adjust our strategies, always keeping in mind that adaptability is key to our success. 

In essence, business operations thrive on transformation; however, it often clashes with our tendency to rigidly fix processes. Data Projects are typically expected to provide long-term solutions, spanning three to five years or more. This expectation stands in contrast to the ever-evolving demands of those who rely on the outputs of these data projects, including Strategic Boards and Corporate Steering Committees.

These entities continually reshape metrics and methodologies to evaluate performance, creating a direct conflict between stability and adaptability. It’s crucial to recognize that both perspectives have merit. We have developed a strategy that balances stability and adaptability, ensuring consistent comparison over time while enabling the business to respond to changing circumstances, instilling confidence in our approach. 

Recognizing the pivotal role of both change and stability, it’s imperative that we effectively manage and govern the pace of change within each segment of our data model. This understanding is not just beneficial, but crucial, as it equips us with the knowledge to strike the right balance between change and stability, thereby empowering us to make strategic decisions.  

As a result, we find ourselves in a situation where we must employ three distinct improvement cycles, each operating at a different pace and governed at varying levels. This structured approach ensures that we manage change effectively and maintain a balance between stability and adaptability.  

Semantic Data Layer 

It is essential to understand that this layer is the most important in the whole data program as it represents the location that establishes the “Source of Truth” (remember we must always aspire for a single Semantic Model). This means that all the report content presented to our business will come from the model and much of the ad-hoc data analysis that will be carried out. As a result, excellent governance must be established here, with all changes being assessed and approved before proceeding.

 

Though that is not a successful day-one position, it is a position that must be established as the “weight” can “gravitas” of the model increases. In practical terms, we regularly see that at initial launch, changes are made to the Semantic Model almost daily or weekly; however, as stability is applied and it becomes more business-critical, the frequency of changes decreases until they become monthly, quarterly or (ideally) annually. The more critical the model is, the less frequently it will be updated. 

Reporting and Analysis 

The stabilisation and management of the Semantic Layer mean that the report content created from there only contains visualisations or relatively nothing that should require governance. Should additional measures be made for a specific report deemed valuable, the measure should be added to the semantic model in the next update. The ability for almost anyone to create new content within Power BI or Excel directly from these semantic models and not need to copy or recreate data means that this content needs only be governed by its audience rather than centrally. This means that a board may determine the report pack they need this month and can dictate any changes required in the future without having to defer to anyone else. This method of working will come across new content that is not in the model; this can then be managed into the model or, more typically, once the model is bedded in, a significant percentage of the request can be met immediately, often altering or even removing the need for change and so saving both time and money.  

Engineering 

Managing engineering changes can be challenging, but they contribute to the stability of the Semantic Layer. There are two types of engineering changes: model-driven and process-driven. 

Model Driven Changes are initiated from the Semantic Data Model based on a data requirement. This usually involves adding a new column or table of data to meet a business requirement. The modelling team anticipates and plans for these changes, making them relatively straightforward to implement. As a result, such changes become less frequent as the frequency of model updates decreases. 

Process-driven Changes, on the other hand, are more complex and are typically prompted by changes in the source application. These changes could range from a simple version update to an entire platform upgrade or application migration. Unlike model-driven changes, the scope of process-driven changes is often unclear as they are driven by external factors. They may occur between core model updates and may require model updates outside of the regular cycle. These changes do not have a specific frequency and are more difficult to control. However, using a separate ETL space can help minimise the impact of these changes by ensuring that the structure of the previous tables used by the Semantic Model is replicated. This may involve tasks as simple as renaming columns or may require significant work, depending on the situation. 

Conclusions 

The key is to recognise that each layer of your data platform requires a distinct improvement or change cycle. Far from being an administrative burden, this variation is a strength for your platform. Achieving excellence in this area will result in a more cost-effective solution that provides necessary insights across all business levels, significantly reducing report lag (the time between event and analysis/reporting) and promoting better business growth through the standardisation of language. 

 

Geordie Consulting is a business intelligence and analytics company that helps clients across multiple sectors leverage their data and gain insights. It offers services such as data warehousing, data integration, data visualisation, dashboard development, reporting, and predictive analytics. We have a team of experienced and certified consultants who use the latest tools and technologies to deliver solutions that meet the client’s needs and expectations. Geordie Consulting is your trusted partner for help achieving your goals and objectives through data-driven decisions.