Disclosure Design Patterns

Reporting is about disclosures.  Standards setters and regulators specify disclosures.  Accounting professionals within a reporting economic entity create disclosures based on the specification.  Auditors verify that a disclosure is true and fair.  Analysts extract information from disclosures and often compare disclosures across periods or between reporting economic entities.

The work of creating a disclosure is tedious; yet at the same time there are very serious consequences from errors. The current conventional disclosure review process is labor intensive, prone to human error, time consuming, and costly.

But what if there were a better way?  What if software could be used by standards setters, regulators, and anyone else creating a reporting scheme to effectively specify required disclosures and supporting information and that information was available in both machine-readable form and also human-readable information could be generated directly from that machine-readable information.

And what if software used for constructing financial disclosures could read that machine-readable information, understand the information, and then assist accounting professionals creating disclosures for their financial reports to do so effectively.

Finally, what if that same machine-readable information could be used by financial analysts to extract information effectively from a disclosure, any disclosure in a financial report, to then use that information or compare information.

The real promise of such software for disclosure specification, construction, verification, and extraction is the software's ability to support tired, overworked, novice creators and reviewers working on unfamiliar disclosures; giving those individuals the benefits of a more experienced accounting professional's skills and experience by using machine-readable rules and software.

Software-aided disclosure specification, construction, verification/review, and extraction aided by Lean Six Sigma techniques offers very dramatic improvements in efficiencies when used by those humans skilled in some area of knowledge creating or otherwise working with such disclosures. After all, software is only as good as the operator of that software.

Tremendous, almost magical, benefits from such self-enforcing, machine-readable digital representations of disclosures can be derived. But how, exactly, are such benefits actually achieved and are these benefits really possible?

In short; "Yes", this is very possible and what I am describing is achievable using good information design techniques.  Information design patterns are logic and principle-driven guidelines convoyed by practical examples of how the patterns have been implemented in real life.  The design patterns help crafters think.

Information design patterns are logic and principle-driven guidelines convoyed by practical examples of how the patterns have been implemented in real life.

The enabler is what I call "disclosure design patterns" or leveraging the logical patterns of a financial disclosure.

Using this approach, disclosure information is not "keyed/typed" or "copy/pasted" into Microsoft Excel or Word.  Disclosures are designed, leveraging the logic of the disclosure patterns to design a disclosure that meets the business objectives of the creator of the disclosure. Software works with the semantics or "things"; not the "strings" of text describing the information.

Let me explain by walking you through one example of the perhaps between 1,000 and 3,000 common US GAAP financial disclosures.

Step 1 is to give the disclosure design pattern a name so that you can explicitly refer to the disclosure design pattern and the disclosure.  In this example, I use the name: "disclosures:InventoryNetRollUp".  I also provide a label for that disclosure, "Inventory, Net (Current) [Roll Up]".  Here is an example of that disclosure:

The next step is to look at a bunch of these disclosures to understand the logical patterns of the disclosure. Here are 63 examples of that specific disclosure. After analyzing those 63 instances of that specific disclosure, I synthesize the logical essence (e.g. the distinguishing features; allow you to identify disclosures in reports) of that disclosure.  The distinguishing features allow one to recognize the disclosure.   Here is that information:

Next, I represent that information in machine readable form using some physical format (a.k.a. technical syntax) that is expressive enough to communicate to a machine all the important details that you need to communicate.  For that, I used the global standard XBRL. Here is the link to that information for disclosures:InventoriesNetRollUp. (I am not going to bother showing you a picture, it is pretty ugly...just open the link if you want to have a look.)

You repeat this same process for every disclosure.  Yes, this is a lot of work; but someone needs to bite this bullet.  But supervised machine learning can be used to make this process easier.  I created a set of about 65 disclosures as a prototype. For more information on that, see this blog post, Disclosures.

I tested that working set of 65 disclosures, here is the result for the 2017 Microsoft 10-K.  I also created another test where I was able to identify about 94.8% of every disclosure in that Microsoft 10-K. Here is that result using working proof of concept software, Pesseract.  Here is another result using Auditchain Pacioli which has some issues, but you will get the idea. The point is that this is already proven to work using three different software applications. (Here is additional information about the Microsoft 2017 10-K testing.)

Finally, I ran this set of disclosure design patterns for the 65 disclosures against a set of 6446 XBRL-based reports submitted to the SEC.  Here are the detailed results.  And here are the results for another test run against a set of 5555 10-Ks for fiscal year 2018.

The objective here is to empower humans and organizations to achieve their goals and objectives.  Leveraging disclosure design patterns (a.k.a. logical patterns) and modern information design techniques, there are tremendous opportunities for improving current processes.

In a human-computer system where the system is composed of people, computers, and financial information; we need to ensure that the people do not become the weak link of that system.  Computers can assist in the processes of planning, constructing, testing, reviewing, and analyzing disclosure information and other such artifacts of accountancy.

The backbone of a disclosure is rarely drafted completely from scratch. Rather, a disclosure tends to be copied from some other prior use of a similar disclosure, reused as templates.  Libraries of properly worded clauses are also possible.  Being digital, this information can be organized, sorted, searched.

Design patterns are reusable models of solutions to commonly occurring disclosure problems.

Commonly occurring disclosure design patterns can alleviate the challenges, the tedium, of creating these common disclosures.  That frees up humans to focus on less common and more challenging financial disclosure problems.

Think of all this machine-readable information as an information product sold by accountants and accounting services firms. Imagine global standards based disclosure pattern libraries usable by software applications.  Crafters of these information products will be compensated for their good disclosure designs. This is a completely new business model.  Both rules-based and statistics-based approaches can be used to create and maintain the necessary machine-readable information.

What I am showing now just scratches the surface.  Imagine, perhaps, disclosure designs being distinguished between good practices, best practices, emergent practices, and novel practices.  The emergent and novel practices disclosure designs are the most valuable and require the highest skill levels and experience to create. But they are also the most empowering.

Thank you to the book Legal Informatics that helped be understand how to better communicate this information, specifically the authors of Towards a Pattern Language Approach to Document Description.

Additional Information:

• Digital Proficiency
• Informatics for Accountancy
• Disclosures
• Super PROOF
• Modern Accountancy
• Working Prototype Software for Viewing Disclosures
• Contract Design Patterns (example of using design patterns)