Digital Financial Reporting

I was talking with a software engineer that said he had an epiphany.  The epiphany was that an XBRL-based report and the discoverable taxonomy set plus XBRL instance is actually a directed acyclic graph.

Backing up a moment; most people talk about the relations in an XBRL taxonomy as a "tree" or hierarchy.  The relations are much more than trees, they are graphs.  There is a big difference between a tree and a graph.

Further, cycles can cause problems.  The biggest problem is "loops" that you cannot escape from.  Having loops or infinite loops can have catastrophic consequences.  They want to be avoided.  Directed acyclic graphs don't have those sorts of problems.

To understand the power of directed acyclic graphs, you first need to understand the power of tree structures.  First, a tree is a special type of graph.  And so, a tree is a graph; but a tree has specific limitations.

A graph provides additional functionality than a tree.  But if you go too far with graphs, you run into the issues of cycles.  Directed acyclic graphs are the "sweet spot" where you get all the advantages of trees, some of the advantages of graphs, but without the catastrophic consequences that can be caused by cycles.

Why is this important?  Basically, if you get your fundamental model of an XBRL-based report wrong you will limit yourself.  Secondly, you can add new information in the form of metadata to the discoverable taxonomy set and expand what your system can achieve leveraging the global standard nature and functionallity of XBRL.

Want to understand more?  Read Computer Empathy.  Need something more technical?  See The XBRL Book: Simple, Technical, Precise.

When you combined the ideas of the structured XBRL-based report which is a directed acyclic graph, the notion of triple-entry accounting, and accounting process automation using XBRL you get a compelling set of features which can be turned into powerful tools.

Stay tuned!

I am analyzing the last 10-K of each of 5,734 public companies that submitted financial reports to the SEC as of March 31, 2018.  This number of filings "5,734" differs from the two sets of that I measured in my quarterly analysis of public company reports.

My analysis of the fundamental accounting concept relations is for the LAST REPORT FILED and that last report could be a 10-K or a 10-Q. That total is 5,856.  The reason I use the last report filed is because fundamental accounting concept relations could have been fixed since the companies last 10-K. (Here is a comparison of fundamental accounting concept relations measurements for the past three years.)

My analysis of the reporting checklist and disclosure mechanics is for the LAST 10-K reported, but when I ran that analysis XBRL Cloud was inadvertenly including 272 funds and trusts in that list which I did not want to work with, so that total is 6,006. (Here is a summary of the reporting checklist and disclosure mechanics measurements which I just started in March 2018.)

But for the analysis I am doing now I am working with the 5,734 10-Ks of public companies as of March 31, 2018.

Here is some information about that set of 5,734 XBRL-based financial reports.

• There are 7,948,552 facts reported by those companies in those 5,734 reports.
• The average report contained about 1,386 facts.
• The report with the most number of facts reported 29,247 facts.
• The report with the least number of facts reported 14 facts.
• The reports contained 447,715 networks which are used to organize those facts into sets.
• The report with the must number of networks provided 310 networks.
• The report with the least number of networks provided 1 network.
• The reports contained 5,553,445 presentation relationships (which I call the model structure) between the Networks, Tables, Axis, Members, LineItems, Abstracts, and Concepts that were used to represent the information in the reports.
  • 5,553,245 of those relations were as expected
  • 200 of those relations were inconsistent with expectation meaning that there was some sort of mistake, this is down from prior year measurements.

In those 5,734 10-K financial reports there were 311,000 text blocks reported using concepts from the US GAAP XBRL Taxonomy.  There were about 964 different text blocks that fell into three SEC/FASB categories:

• Level 1 Note text blocks that were used to represent entire notes.
• Level 2 Policy text blocks that were used to represent individual accounting policies.
• Level 3 Disclosure text blocks that were used to represent individual disclosures.

I am not analysing the Level 4 Disclosure Detail here. I compare Level 4 Disclosure Detail and Level 3 Disclosure Text Blocks here.

What I am doing here is to compare how public companies use specific text blocks.  Here is information about those text blocks:

• Here is a complete list of different text blocks in the US GAAP XBRL Taxonomy (note that this is the 2018 taxonomy, it is slightly different than the 2017, 2016, and 2015 which were used to submit reports)
• Here is a list of 17 text blocks that I have created comparisons for every public company that uses each specific text block to see HOW they are using it.
• Here is one comparison, the Nature of Operations Level 1 Note text block.

What is very useful about these comparisons is that you can see how different public companies use these text blocks. I did a specific analysis of the Level 3 Disclosure text block related to the disclosure of the components of property, plant, and equipment.  See the full comparison here. Here are some individual observations:

1. Most public companies use the text block "us-gaap:PropertyPlantAndEquipmentTextBlock" to represent the components of PPE which tends to look like this.
2. A good chunk of public companies include the estimated useful lives of each category of PPE in that text block, which tends to look like this.
3. A small minority use that text block to disclose ONLY the estimated useful lives of PPE categories.
4. A small minority include depreciation expense AND the breakdown of PPE components.
5. A minority include PPE components in their PPE policy text block.

Clearly all of these four cannot be considered correct.  My personal view is that #1 is correct.  #2 should be broken down into TWO separate disclosures in my view; one for the roll up of PPE components and another for the estimated useful lives.  #3 clearly cannot be considered correct, it does not match what most public companies do and it does not match the documentation of the text block.  #5 likewise is clearly not correct, PPE components go into the level 3 disclosure text block not the policy.

What causes these sort of issues is a "presentation" orientation when thinking about how to represent these disclosures.  If a "representation" of information orientation were taken, how things should be represented becomes more clear.

What do you think?  More analysis coming, so stay tuned.

Over the past years I have created a number of prototypes of organizations of disclosures.  For example, this set of disclosure best practices.  Or this resource that provides information about different disclosures. And this slightly older version.  And I personally think this is a handy example.

Here is some prototyping of a handful of disclosures.  I am using this to figure out my next version of disclosure references:

• Nature of Business
• Basis of Reporting
• Fiscal Period
• Income Tax Note

More examples to come.

Working with another professor at the University of Washington Bothell Campus, I created this resource for helping graduate students learn how to evaluate the correctness or incorrectness of the information contained in an XBRL-based financial report.

I am ultimately going to create a better organized place to get this information.  Consider this blog post an initial prototype.  Here are other good resources:

• Fundamental accounting concept relations
• High quality documented examples of fundamental accounting concept errors (see the multiple PDF files toward the bottom of the blog post)
• Tools for evaluating the fundamental accounting concept relations of XBRL-based reports
• Information about errors in disclosures of financial reports
• Analysis of the income tax note (i.e. the link above)
• Examples of XBRL-based reports that are PRETTY GOOD (i.e. don't have any detected errors)
• Blueprint for creating zero-defect XBRL-based financial reports
• Free tool

I know that the above is a lot of stuff and not well organized.  My next task is to better organize all of this so that accounting students can make use of this information.  Stay tuned!