Digital Financial Reporting

Today, many important XBRL-based taxonomies are being created as what amounts to human-readable "pick lists".  In fact, that term "pick list" was used by people at the FASB to describe the US GAAP XBRL Taxonomy.  Here are three examples of XBRL taxonomies that are essentially "pick lists":

PICK LIST BASED APPROACH: 

• US GAAP XBRL Taxonomy
• IFRS XBRL Taxonomy
• UK Charities XBRL Taxonomy

So, what is the alternative to these "pick lists"?  Why is the alternative better? 

The alternative is to use a model-based approach to representing the XBRL taxonomy so that the information in the XBRL taxonomy is machine-readable but also readable by humans.  The benefit of using a model-based approach is that you can control reports created using the XBRL taxonomy ensuring that they actually work effectively.  Here are examples of XBRL taxonomies using a model-based approach.

MODEL BASED APPROACH: (model driven architecture)

• Not-for-profit (US GAAP) XBRL Taxonomy (working prototype) (Also see this for full machine-readable and human-readable documentation)
• Other model based approach examples

So, what is the actual difference between the "pick list" as contrast to the "model-based" approaches?  Here you go:

1. Consistency:  The model-based approach is extremely consistent whereas the pick list based approach is far less consistent.  For example, the model-based approach uses XBRL International's best-practice guidance related to the use of XBRL Dimensions.  Every disclosure is consistently represented using a uniquely identifiable hypercube.
2. Smaller, identifiable pieces: The model-based approach results in many more smaller easily identifiable pieces as contrast to the pick list approach which tends to create fewer pieces and crams "stuff" together in a far less organized manner.
3. Guarantied quality: The model-based approach provides far, far more machine-readable rules that have four roles.  First, the rules are used to verify that the XBRL taxonomy has been created correctly.  Second, those same rules provide explicit guidance as to how to construct disclosures using the XBRL taxonomy by those creating reports.  Third, those same rules are used by automated processes to verify that the report has been created consistent with expectations.  Fourth, those same rules are used to effectively extract information from the XBRL-based reports.  It is the rules that both provide and enforce the guarantee!
4. Easier to use:  The model-based approach is easier to use because complexity is moved from those creating reports to those creating the XBRL taxonomy.  Further, that complexity is hidden from the user by software applications that leverage the well created XBRL taxonomy and all of the rules.  Essentially, business professionals need only concern themselves with one thing: business logic of the information being represented.  That, they understand.
5. Easier taxonomy maintenance: With the model-based approach taxonomy maintenance is easier.  Why?  First, the pieces are smaller and significantly easier to manage.  Second, all those rules make it virtually impossible to inadvertently break the taxonomy.  Essentially, taxonomies that use a model-based approach have a built in test harness.
6. Significantly improved taxonomy functionality: Model-based XBRL taxonomies are a web of information.  Just look at all the relations for a disclosure.

Pick list-based and model-based XBRL taxonomies are not mutually exclusive.  Contrast these two example disclosures for the components of inventories for my Not for Profit prototype and US GAAP.  You can turn a pick list XBRL taxonomy into a model; or you can turn a model-based taxonomy into a pick list if you so choose. 

There is no impact on the actual terms.  The impact relates to structures, associations, and rules.  Models have identifiable structures, clear associations, and complete set of rules.  Pick lists don't.  Pick list based taxonomies leave a lot of information out; that is what causes the quality problems experienced by such XBRL taxonomies.

Here is an example of what you can achieve using a model-based approach.  Consider this: 

• Have a look at this report.  Human-readable | Machine-readable
• Note that the report is comprised of 77 individually identifiable set of information.  For example, "Document information" is one set, the "Statement of financial position" is two sets (i.e. Assets roll up, Liabilities and Net Assets Roll Up); count them all you will find a total of 77.
• This validation report shows that all 77 of those individually identifiable sets of information is consistent with expectation. (Note that the list has 58 items, but the Level 3 and Level 4 disclosures are listed together on one line in that validation report, but individually count as different information sets)
• This validation report and this validation report show that the mathematical relations are consistent.  Here is another version of the same information.
• This validation summary provides a pretty good dashboard that helps you understand that the report is properly functioning.

And so, while software is not where it needs to be; a person that understands WHAT to look for and WHERE to look can confirm that a model-based report is properly functioning.  You fundamentally cannot do this with a pick list based report; the information is simply not there.  So what do you do?  Add the information you need.  Then, you will have a model based XBRL taxonomy for your report.  But that does not make the underlying base taxonomy model-based; only the report.

Here, I created the rules that cover 95% of the 194 individual sets of information in the Microsoft 2017 10-K.  Why only 95%?  Because the other 5% I could not effectively represent because of extension concepts. (see the complete analysis here) 

• SEC Filing Page
• XBRL Instance
• Human readable validation report
• Disclosure Mechanics validation (partial validation tool you can use, before I added the additional rules)

This video provides additional information.