Digital Financial Reporting

XBRL-US and five members including Merrill, RR Donnelley, RDG Filings, Vintage, and Workiva have organized to take on the quality of public company XBRL-based financial filings to the SEC.  See this blog post by XBRL-US.

Microsoft Word and Excel are used to create an estimated 85% of all external general purpose financial reports.  But what do Word and Excel understand about financial reports?  The answer is that Word and Excel know nothing about financial reports.  What if the software application used to create financial reports did understand financial reports?

Understanding digital financial reporting and its benefits

Digital financial reporting is financial reporting using structured, machine-readable form rather than traditional approaches to financial reporting which are paper-based or electronic versions of paper reports such as HTML, PDF, or a document from a word processor which is only readable by humans. A digital financial report is readable by both humans and by machine-based processes. Because digital financial reports are machine-readable the machine can be more intelligent and helpful in understanding where a professional accountant is working in a financial report (i.e. context-aware) and therefore make the software more adaptive, more dynamic, able to provide an advisory role to professional accountants, be more proactive, and provide other knowledgeable guidance related to the creation and review of a financial report. Unlike word processors which understand nothing about a financial report, digital financial report software has an intimate understanding of a financial report.

Digital financial reporting includes the creation of general purpose financial reports under IFRS, US GAAP, governmental accounting standards, or other reporting schemes. Other reporting schemes may also use this global standard approach to creating a machine-readable digital financial report. The focus here is not specifically what goes into a financial report per one specific reporting scheme; but rather it is about the digital financial report itself. Digital financial reports can be used by any reporting scheme which might choose to express some financial or non-financial information digitally.

Digital financial reporting can be understood by contrasting that process to the process of creating blueprints using Computer-aided Design/Computer-aided Manufacturing (CAD/CAM) software . Just as CAD/CAM software is knowledgeable of blueprints; digital financial reporting software is knowledgeable of financial reports. CAD/CAM software understands what a door is, what a window is, what a wall is, and that a window goes into a wall. Similarly, digital financial reporting software understands what a balance sheet is, what an income statement is, what a disclosure is, that assets goes into the balance sheet and that assets equals liabilities and equity, per the accounting equation. CAD/CAM software is used to increase the productivity of the designer, improve the quality of design, improve communications through documentation, and to create a database for manufacturing. CAD/CAM output is often in the form of machine-readable information which can be printed; provide instructions for machining directly to a numerically controlled machine, or used in other ways for other manufacturing operations. Similarly, a digital financial report will travel through the entire supply chain which is connected via the Internet and information never needs to be rekeyed and different business systems will have the same understanding of the reported financial facts and the relations between the reported facts.

The machine's knowledge of a digital financial report is enabled by the structured nature of the information represented within the machine-readable financial report, metadata that explains business rules related to the creation of the financial report that the machine must follow, and meta-metadata which helps other participants of the financial reporting supply chain such as investors and analysts which make use of the reported financial information interact with these machine-readable artifacts to effectively and successfully exchange meaning between business systems and processes. Knowledge about the mechanics of a financial report and how to create a financial report is carefully expressed in machine-readable form by humans. This is not to say that all knowledge can or will be expressed; rather, only objective knowledge that makes a computer software program capable of helping its human operators can be expressed. Subjective knowledge, such as the judgment of a professional accountant, can never be expressed in terms that are understandable by a machine. Essentially, the machine mimics basic mechanical tasks related to the creation of a financial report. A machine gives you the sense that it is intelligent, it appears intelligent, because understands the context in which it is working.

The machine's understanding is enabled using software algorithms and machine-readable metadata and meta-metadata. Metadata and meta-metadata is stored and managed within formal and informal ontologies  which describe the things that make up a financial report, important relations between those things, and other information related to the financial report knowledge domain in machine-readable form. Ontologies are created and managed by knowledge engineers who help business professionals create and manage this metadata. Ontologies both describe the business domain and serve to verify that digital financial reports created are consistent with that description. Machines can assist accounting professionals in the creation of financial reports to the extent that metadata about the financial report and how to create the financial report is articulated in machine-readable form. Another term for these relations is business rules.  Understanding what business rules are and what they do is key in understanding the capabilities of machines to help professional accountants create and review financial reports.  (For more information about business rules, see the Business Rules Manifesto.)

Current tools such as disclosure checklists which serve as memory joggers to humans are not machine-readable. In the context of digital financial reporting these current human readable checklists and memory joggers are made machine-readable and therefore many tasks which can be automated using machine-based processes will be performed by computers. As previously stated, processes which will be automated are the more mechanical aspects of creating a financial report, as opposed to the judgmental aspects which require the knowledge of a professional accountant to get correct. Digital financial reports will free both professional accountants who create these reports and financial analysts and regulators who use information from these reports from tasks such as re-keying information and making sure the objective aspects such as mathematical relations of a report are correct, allowing professional accountants to focus on judgmental and other subjective aspects which cannot be automated.

The benefit of digital financial reporting is enabling machines to take over mindless and mundane mechanical tasks that are involved in the creation of financial reports.  Not all tasks, rather tasks that can be effectively achieved using machines. Enabling machines to take over tasks that had been performed by humans results in reduced costs involved in creating financial reports, reducing human errors because machines take over many mindless mundane mechanical tasks, increased quality and reduce the risk of noncompliance because machines take over these mindless mundane tasks, and less time to complete financial reports because of the assistance provided by automated machine-based processes.  Automation can be achieved to the extent that machine-readable metadata and data is provided and that software algorithms can be written.

How do you get a computer to actively help you create a financial report? Semantics teaches a computer to understand what you mean.  And so how do you construct such an application?  This video, Process Execution Through Application Ontologies, provides some answers. The video outlines five steps:

Phase 1: Gather information

• Step 1 - Delivery: Determine the delivery (in our case, the delivery is a financial statement)
• Step 2 - Events: Determine the necessary events which should happen to make the delivery
• Step 3 - Event sequence: Determine the earliest sequence an event could occur relative to other events
• Step 4 - Process states: Determine the different states of the process
• Step 5 - Create a process diagram:  Design/create a process diagram (perhaps using BPMN notation, here is a BPMN tutorial, here is a video that explains BPMN)

Phase 2: Classification of members

• Step 6 - Construct ontology: Create an ontology that represents everything gathered in Phase 1

Here is more information on this topic.  Interesting...More to come.

In order to understand digital financial reporting one needs to understand the capabilities of semantic technologies.  The document Web 3.0 Manifesto: How Semantic Technologies in Products and Services Will Drive Breakthroughs in Capability, User Experience, Performance, and Life Cycle Value, written by Mills Davis, provides a good explanation of semantic technologies.

The entire document is worth reading.  Page 4 of the document provides this concise explanation of what is meant by semantic technologies:

What are Semantic Technologies?
Semantic technologies are digital tools that represent
meanings and knowledge (e.g., knowledge of something,
knowledge about something, and knowledge how to do
something, etc.) separately from content or behavior artifacts
such as documents, data files, and program code.
This knowledge is encoded in a digital form that both
people and machines can access and interpret.
The basic shift in information and communications technology
that is occurring now is from information-centric
to knowledge-centric patterns of computing.
This is made possible by the application of semantic
technologies and open standards that enable people and
machines to connect, evolve, share, and use knowledge
on an unprecedented scale and in new ways that make
our experience of the internet better. Not restricted just
to current Semantic Web standards, the next stage of internet
evolution will encompass a broad range of knowledge
representation and reasoning capabilities including
microformats, semantic html, pattern detection, deep
linguistics, ontology and model based inferencing, analogy
and reasoning with uncertainties, conflicts, causality,
and values.

Semantic technologies will sometimes be used to bolt on additional capabilities onto existing applications and processes.  But semantic technologies will also enable applications and processes to be completely re-thought and re-created.  In other words; sure, semantic technologies will create some incremental changes but more likely than not they will be creating paradigm shifts.

These ideas can be hard to understand if you don't understand how a computer works.  It is just as much a mistake to under-estimate what computers can do as it is to over-estimate what computers can do.  I have summarized what I have learned about how computers work in the document, Knowledge Engineering Basics for Accounting Professionals.  None of these are my ideas.  All I have done is figure out the pieces to the puzzle and try and organize those pieces as best as I can in order to understand how all the pieces fit together.  The primary purpose was for me to understand how to make digital financial reporting work.

This video, The Future Internet: Service Web 3.0, is helpful.

This video, Intro to the Semantic Web, is helpful.

This video, The Next 5000 days of the web, is helpful.  It points out that the internet is really one machine and the devices we have are windows into that single machine. In 5000 days the web will not be "only better", it will be something different.

This video, "Why the Sematic Web Won't Work", is helpful because it explains WAY certain things do not work and what DOES work.

This book, Systematic Introduction to Expert Systems: Knowledge Representations and Problem Solving Methods, helps you understand expert systems.

XBRL UK has published a paper, Company Reporting in the UK – an XBRL Success Story, which discusses the use of XBRL-based financial reporting by HMRC (the tax authority) and Companies House (the business register) in the United Kingdom.  Approximately 1.9 million companies are said to be using this system.

The document is worth reading and the implementation approach used is worth understanding. XBRL UK and others have clearly put a substantial amount of thought and effort into what they have created. Anyone trying to implement XBRL using any approach could learn a lot from their publication.

CoreFiling, a software vendor, seems to be making this information available in an organized format.  You can see what CoreFlings has provided here.  I have not figured out how or where to get this information from the Companies House web site.

The document says that companies report using UK GAAP and sometimes IFRS.

While I think this UK approach is worth understanding and the document points out very important considerations when creating a system for exchanging information, I think that the approach is limiting in that it does not support reporting entity extensions.

There are two significant limitations that the UK approach has:

• Adding new information:  Say a reporting entity wanted to provide some specific disclosure that the UK taxonomy does provide for.  For example, what if a reporting company wanted to break out revenues by some criteria not provided for in the taxonomy.  This could not be handled.
• Changing or amending disclosures such as qualitative disclosures:  Say a reporting entity had a complex derivative that they need to provide important qualitative information for that did not fall into the existing organization of the UK taxonomy.  This could not be handled.

These sorts of very company specific disclosures are not provided for under the UK taxonomy approach.  However, the only gap between the UK approach and say the US GAAP XBRL Taxonomy approach which does provide for those sorts of very company specific disclosures is a very thin layer which could be exposed by the UK taxonomy.

That thin layer exposes the taxonomy construction rules and taxonomy building blocks, opening up just a minimal amount of very specific flexibilty for those who need to create extensions.

Providing these openings or what I call "slots" does not have to turn the fairly simple and successfully implemented approach created by the UK into something that is complicated and which cannot be handled by business professionals creating this information.

Using known patterns, templates, metadata, well articulated rules, and other means such as those pointed out in this document, Understanding Blocks, Slots, Templates and Exemplars, the complexity of creating extensions can be moved from being handled by business professionals to being handled by the software applications business professionals use.  It is hard to see this because the complexity involved in creating extensions is being judged by today's poorly created software.  Today's software tends to force business professionals to deal with technical syntax and complicated knowledge engineering issues.

And so, while I think that what the UK is doing is a very reasonable moderate step for a regulator to take; I also believe that the institution of accountancy need not be seduced by an approach that tends to try and simplify the problem by making the problem more simplistic.  Rather, empowering business and accounting professionals with good software and basic knowledge engineering skills and the ability to therefore create rock-solid extensions can also work, I believe.  XBRL was engineered to be extensible for a reason.

This is why I believe what the SEC and FASB are doing in the US, taking a significantly bolder leap, is still the right path.  While it is true that public companies are struggling with this approach currently, I believe a lot is being learned by going through these struggles.

Again, this is my vision of digital financial reporting.  I am not the only one that seems to have this vision.

I think that XBRL implementation in the US could learn a lot from the UK approach; but I also think the UK and others can also learn from the approach used by the FASB and SEC.  As the UK approach points out, collaboration and cooperation of all those in the financial reporting supply chain is critically important.

This article, Justifications for semantic technologies – revisited, is worth reading.

The bottom line is this.  Semantic technologies allow for:

• straight application of knowledge in the production process through application ontologies; there is near to no entropy involved
• description and management of processes
• unambiguous descriptions through glossaries, context, knowledge models whereby concepts are explained through their relation to other concepts
• knowledge engineering
• knowledge representation
• reading by men and machines
• interoperability constructs and formats
• version control on the level of descriptions and operations
• knowledge generation on top of asserted knowledge
• support the growing attention for openness in public administrations (open data)
• coping with the above mentioned issues