In a private group email related to the creation of the Standard Business Report Model (SBRM), a colleague from the Object Management Group (OMG) mentioned the notion of "fail loudly" which is one of the most brilliant statements I have heard in years. And the notion of "fail loudly" is connected to other notions related to failure:
- Fail Loudly
- Masked Failure
- Prevent Failure
- Define Failure
- Control Failure
- Cost of Failure
- Avoiding Failure
In the rest of this article I will elaborate on each of the above ideas as they relate to the work performed by professional accountants in the creation of financial reports or performing internal or third party financial audits.
Fail loudly
"Fail loudly" means that instead of allowing mistakes to accumulate and show up unexpectedly later in a process; software proactively monitors and detects invalid inputs, unexpected states, and violations of assumptions and expectations in real time. If something goes wrong, software throws an exception. Additionally, "failing loudly" ensures that when something breaks, it breaks in a way that a reporting team or audit team can quickly diagnose and fix the issue because they are made aware (loudly) that the issue exists.
Silent failures or suppressed mistakes lead to problems that are harder to track down. By proactively notifying users of software by providing notifications and asserting undesirable conditions; software makes such failures visible proactively such that they can be responded to proactively.
The first screen shot below shows an example of failing loudly. The second screen shot shows a baseline where everything is OK. In this first screen shot I induced an error by putting a "1" in a report for the period "2024-12-31" and the line item "Cash and Cash Equivalents" (see the black box in the screen shot below). Notice that doing that caused a red outline to appear in the human readable rendering of the report for "Current Assets", the container of the right for "1101 - Statement - Balance Sheet" to turn red because of that defect, similarly four other containers turned red because the defect impacted that information as well, and then in the lower left corner of the screenshot in the black bar, a red bubble appears with the number "5" which if clicked on would show more detailed information about all the rules being checked and the five failures caused by this one mistake:
In this screen shot below, everything is back to being OK because I fixed the flaw, changing the "1" back to "0" in this working proof of concept. And so, you see none of the glaring indicators that there is some sort of failure within the system. Why? Well, because there are no failures to show.
And so, that is an example of "failing loudly". A software user is proactively made aware by a software application of a mathematical flaw. I am not saying this notion of "failing loudly" can only be used for mathematical flaws. In fact, the Seattle Method currently supports eight categories of flaws, eight theories related to the fundamental mechanics and dynamics of how financial statements work that both the software application and the human using the software application can understand.
Masked Failure
Lets face it, traditional electronic spreadsheets, like Excel, mask failures many times. Excel is a beloved tool of accountants; I am an accountant and I love Excel. When I started as an accountant with Price Waterhouse in 1982 I created spreadsheets on paper. When I started using electronic spreadsheets, I started with VisiCalc. Then I switched to Lotus 1-2-3. Then to Excel when it was introduced by Microsoft.
But traditional spreadsheets are not the solution to every problem that has ever or will ever face humanity like most accountants seem to think. Electronic spreadsheets are a tool; even somewhat of a Swiss army knife. But how many professional chefs do you know that use Swiss army knifes for all their cutting needs?
Traditional electronic spreadsheets are a stepping stone in the progression of tools for accountants. Traditional electronic spreadsheets have their advantages; but they also have disadvantages. One of those disadvantages is that they mask failure in many cases, proof of that is the reputation traditional spreadsheets have for containing errors that no one knows about. That alone makes spreadsheets difficult, or really, impossible to scale.
Prevent Failure
The best way to deal with failures is not to have failures occur in the first place. That is what Lean Six Sigma is all about, eliminating the possibility of errors. While eliminating the possibility of all errors when it comes to accounting working papers, audit working papers, financial statements, and financial analysis models is not always possible, it is possible to prevent many failures altogether by creating good processes and using techniques such as poka yoke "mistake proofing" and other Lean Six Sigma principles, practices, and techniques.
Poka yoke is a simple idea with profound ramifications which can be applied to accounting working papers, audit working papers, financial statements, and financial analysis models.
Informatics, cybernetics, and industrial engineering can be used to build better software.
Define Failure
The first step in preventing failure is understanding all the things that can go wrong.
Epistemic risk relates to the risk of being wrong.
Requisite variety relates to every possible way of being wrong. To manage the risk of something being wrong, your "inspection process" or methods of checking, validating, and governing risk must have enough variety to match the possible errors which could occur. For you to be able to detect the maximum amount of failure, your system must be
complete for your area of interest. Completeness is not produced by reasoning;
completeness must be declared or governed. Completeness must be defined. I have defined "complete" (for now) as being the eight theories of the
Seattle Method that
define the mechanics and dynamics of a financial statement. All those eight theories are necessary; I am not holding them out to be sufficient. You will very likely also need humans involved but there are thousands and thousands of additional rules that can be used to define the things that can go wrong, the "perturbance variety", and then control those possibilities using "control variety". Way, way more automation is possible.
Control Failure
Failure can be controlled. Step one of controlling failure, as pointed out just above, is understanding all the things that can fail. Certain aspects of failure can be controlled with automated processes. For example, an accountant with a calculator can be used to control mathematical mistakes. But not all failures can be caught by automated processes. But you don't want automated processes and human processes that perform work to be separate, in separate silos. You want humans and machines to be able to work together, I call this a "
mindful machine". Creating a general machine that does everything is extremely hard work, and that might not even be possible. But, it is very possible to build a
specialized machine, a "
mindful machine for accountancy". This mindful machine for accountancy enhances the capabilities of an accountant making use of that machine.
Another name for this type of machine is a
knowledge based system. What if we were to
rethink the electronic spreadsheet and create another type of electronic spreadsheet. The traditional electronic spreadsheet would still exist, but we would have a new additional tool in our quiver. Imagine a new type of electronic spreadsheet. Imagine a global open industry standards based, model-driven, semantic-powered, artificial intelligence enabling electronic spreadsheet that was also a knowledge based system.
If we are honest with ourselves, we see that
financial reporting and financial audits are showing signs of stress. Information is becoming more and more complex, the pace information flows is increasing, and the volume of information keeps growing and growing and growing. But the same technologies that are causing these problems can be used to solve these problems.
Things like Sarbanes Oxley address the symptoms of what is going on, Sarbanes Oxley does not address the conditions that cause the problem in the first place. A new type of global open industry standards based, model-driven, semantic-powered, artificial intelligence-enabling electronic spreadsheet can be a big step in addressing the conditions that are causing the problem.
What if accounting working papers, audit working papers, financial statements, and financial analysis models were databases?Cost of Failure
Failure has a cost. Costs multiply by a factor of 10 each time a defect goes unnoticed. The
1-10-100 rule highlights the exponential cost increase of addressing quality issues at different stages:
- $1 to prevent a defect by fixing a system,
- $10 to correct a defect once detected, and
- $100 or more for dealing with the ramifications of failure to catch a defect before that defect reaches the customer.
The 1-10-100 rule emphasizes that proactive prevention by fixing a system, say by using poka-yoke and other Lean Six Sigma techniques, is significantly more cost-effective than later correction or dealing with the fallout of a significant failure, thus promoting investment in quality control early in any process.
Lets face facts. Today's accountant is what amounts to a "
data janitor" and a "
transaction chaser" that often has to resort to techniques like "
the plug" to get their jobs done. The systems they have to use are "
hairballs". The systems focus on the "
end of processes" (i.e. they don't consider the full system). And most accountants think that this is normal. (graphic
from here)
There is a better way. Can artificial intelligence contribute to solving this problem? Yes, artificial intelligence can help. But, as
The AI Ladder points out; first you have to fix your data and get the right skill set. Artificial intelligence can help avoid failure but artificial intelligence cannot perform magic. A hybrid combination with the proper mix of rules-based and probability-based artificial intelligence can help avoid failure. (From
A DARPA Perspective on Artificial Intelligence)
Avoid Failure
There are lots of ways to avoid failure. One is to throw a lot of expensive people at a problem which is how most people get financial reports created and perform audits. There are two consequences of this approach. First is a higher cost to prevent failure. Why the high cost? Well, because of all those expensive people.
The second consequence is that human error is hard to eradicate. Humans make mistakes and making mistakes is part of their humanity. Therefore, it is hard for humans by themselves to remove all failures.
But what if you could enable humans and computers to work together as a team? This is not a new concept; the medical profession uses
robotic surgery.
Robotic surgery (a.k.a. robot-assisted surgery) uses computer "robots" to enhance a surgeon's capabilities in the performance of minimally invasive procedures during complex operations with enhanced precision, vision, and control through tiny incisions.
Think computers enhancing an accountant's capabilities.
A new era of human/machine teaming has arrived. Machines will free accountants from the drudgery of repetitive processes like the rekeying of information and augment the capabilities of accountants and auditors; amplifying there intelligence. Some task or process currently performed by humans that, if measured, would achieve a sigma level of 3 which is a defect rate of 6.7% (about 67,000 defects per million opportunities) could be improved and could achieve a sigma level of 6 which is a defect rate of 0.00034% (about 4 defects per million opportunities).
You are hearing me right, defects go from a whopping 67,000 down to 4!
Obviously we want to avoid failure and for the lowest possible cost. But currently, most businesses are stuck in the paradigm where electronic spreadsheets and electronic spreadsheet-type tools are the only alternative. People have been beating on this approach for years trying to make it work better and better; that is a dead end which simply cannot be made to work at scale.
What if there were a better way, a completely new paradigm that was not a dead end? What if there were a new paradigm that provided a new foundation, or an entire ecosystem, upon which to build?
The global open industry standards already exist via the XBRL International
Open Information Model (OIM) and Object Management Group's
Standard Business Report Model (SBRM). The Seattle Method is based on OIM and the Seattle Method is a large part of the basis for SBRM. And so, the Seattle Method, OIM, and SBRM are in alignment.
Additional Information:
Comments
Post a Comment