The standard models represent typical business subject areas. However, those may not relate to all your business needs. You will have to develop and customise the models accordingly. You can figure out the gaps by mapping the data items within the data stores of the subject areas back to the data modelled in the canonical model.
However, you can still run into snags. Your data modellers must ensure that the sources of the data being used are the correct ones. There can often be more than one source of data – imagine a customer’s name in a bank, which may originate in the credit card, home loan, transmission or another division – so you’ll need to create a merged (integrated) set of that data. The next headache is ensuring that the customer’s name is correctly spelled across data sources, for example. You’ll also need to ensure that all the data are there.
You’ll also need to ensure that the source custodians, the credit card, home loans, and transmission divisions in the example above, agree to standardise the structure, collection and capture of that data in future or it throws the canonical model out all over again. That’s achieved through artefact definitions (metadata). They’re like a dictionary of definitions describing what data is collected, how, when, where, why, what it will be used for in the process and so on.
Canonical data also requires processes be clearly defined and enforced. Straying from process allows error to creep into the data collection, storage, retrieval, copying and destruction, which again throws out the canonical model.
IT resources, for example any artefact, whether a data item or process, are enshrined in the discipline of metadata management.
In summary, canonical data is standardising on the simplest form of data, across all your impacted data sources that meets your business goals. But it’s important to go beyond mapping pre-existing models for your industry to your data. You must check the source data, ensure there is a single source or ensure where there are multiple sources they are integrated and the data quality is good. Metadata management thereafter ensures on-going protection of your canonical model that delivers trustworthy data to the people in your organisation who will use it regularly.