Data Perception is essential to understand and correctly define your organisational data.

Data context is business user perception of its various business processes.

Business areas often combine data from different business processes like Sales, Customer registrations, Budget and Procurement. These processes need to be organised to present a unified output. When organising data, each business process needs to be evaluated to build the data context.

Sharing your business data perception with data practitioners can help improve data value and time taken to achieve desired outcomes.

Often, when data professionals are presented with a dataset, different results can be obtained from analysis of the dataset. If three possibilities are obtained, all three can be true and this depends on various factors.

These factors include the quantity of the data used, the descriptive information, and the granularity of the data that has driven the analysis. The factors are the perspective from which the data is being seen by its users.

Showing different perspectives and contexts can help the business to explore new possibilities however. Perceptions can always be improved according to business need.

Generating three different output from the same dataset can lead to users losing trust in the data when this output is not aligned with their perception. The factors which drive the output from the data should be explained to business users. This will facilitate understanding and indicate where such factors are contrary to users perception of the data.

How you perceive or view the data depends on the grain of the data and the additional information you have available regarding the data collected and historical evidence or experience.

Often data is thought of in terms of counts, sums, multiples, spread. These provide informative and familiar insights to users. After all, the business aim is often more increased success, increased profit, increased sales, reduced cancellations, #insert metric here by using current data.

There are a number of data management areas which can help improve trust in data. One important data management activity which supports increased trust in data is conforming attributes in a dataset.

Conforming attributes involves aligning definitions of business terms across the organisation.

Conformed data across the business is essential to have organised and meaningful data across an organisation. This ensures a common vocabulary is used across the business and the data fields are interpreted correctly. To get value from your data, the data team needs to understand the vocabulary. They must know how the data is represented across various business processes.

Naming conventions are critical to ensuring the various business areas use a common vocabulary. Adopting naming conventions will usually be achieved when conforming data attributes across an organisation. Resolution of naming conflicts ensure meanings are organised before making data decisions. This resolution can help distinguish between data contexts which initially appear to be the same. Developing a taxonomy and ontology of the organisation provides an advantage. These tools are used to improve the data context.

When building data solutions, it is important to pay attention to the context. The context is crucial to how the data is used in an organisation’s ecosystem. This will guarantee the data is used in the correct way. Occasionally, incorrect data is used to make important decisions. This results in outcomes that do not give a true picture of the organisation and its intended use.

It is well known in statistics that data can be interpreted differently based on the circumstances which surround the data. Statistics does not intentionally change data interpretations. Statistics highlights that data is subjective depending on the level of bias or context provided.

A statistical sample is selection of data points to represent the whole population. This sample will introduce bias in the results. It is not possible to remove bias completely but we can improve the level of confidence in our data. This results from this sample provide a view of the whole population and will be appropriate for statistical use (context).

An example is ; where a government says the economy is doing great because there is increased employment based on the numbers obtained from it’s own hiring. The conclusion is biased because the whole economy is made up of both public(government) and private jobs.

However, if the government says; the economy is doing great because we have employed more people to assist with the increased tax revenue from the whole economy, this gives new meaning to the conclusions. There isn’t a direct correlation between the government hiring and whole population hiring even though there is increased revenue to the government. Other factors may be at play. To obtain a representative and true value, the data should be collected from both public and private sectors.

The concept of data context is not new and is as old as time. However, many organisations have not successfully implemented contextual data interpretation into their data solutions.

A key task for data practitioners is helping business users understand context. Another part is getting them to relate to the data. Data practitioners need to understand the user perspective, organise the business taxonomy and ontology. This enables a common language to be used across the organisation.

Data practitioners also need to ensure data modeling activities are performed across business processes. Dimensional data modeling emphasises the need to understand the various business processes across an organisation. It requires organising the data carefully according to the context of business use. This enables the realisation of data value.

There are data management activities which can be used to ensure your business data is reflective of the applicable context.

A practical example of various processes in an organisation is ;

Business process in a Higher Education environment which typically include; Student registration, Student application, Finance budgeting, Student enrolment etc.

The registration team , Instructor team, Finance team will have different perspective on the number of students in the institution. It is important to establish the various understanding of student count and resolve any conflicts between each area requiring the “metrics”.

The metrics (called ‘facts’ in dimensional modeling) delivered as KPIs or business reports rely on other attributes (‘dimensions’) to provide meaning. The modeling of these facts and dimensions are key to representing the business information correctly.

Do your KPIs reflect your view of your business processes? Do you often have conflicting outcomes from your data? Let’s help review your data solution’s context.

References influenced by : How to Lie with Statistics – Darrel Huff , Effective data storytelling – Brent Dykes