Back in 2010 I did some rudimentary but comprehensive testing on public company XBRL-based financial reports and was able to find 92 (about 1% of the total) which I dubbed "All Stars" who were getting a certain specific set of information in XBRL-based financial reports correct.
In March 2014, for fiscal year 2013 public company filings, I discovered a set of minimum criteria for reliably reading any information within a public company XBRL-based financial report. This document helps you understand those criteria, and this document shows how public company XBRL-based financial filings stacked up against those criteria. Testing revealed 1,281 All Stars (19% of total filers) back in March.
Yesterday I did my first testing run using commercially available software against those minimum criteria and I can proclaim that the number of public company digital financial reporting All Stars has grown to 3,936 which is 56.7% of the total 6,953 SEC filers.
So over half of the public companies get 100% of these minimum criteria correct which makes their XBRL-based financial filings fundamentally useable and reliable. This is NOT to say that all of the information in the digital financial report is correct. The information is only reliable and meaningful to the extent that it can be tested and proven to be reliable and therefore meaningful and usable. Remember this truth:
The only way a meaningful exchange of information can occur is the prior existence of agreed upon technical syntax rules, domain semantics rules, and workflow/process rules.
If you don't understand why the statement above is true, go back and watch this HL7 video. Given that truth, here is the logic which shows why the minimum criteria are necessary:
- Technical syntax is readable: Software works with some technical syntax, XBRL in the case of SEC XBRL-based financial filings. Software reads the syntax and therefore has access to the information expressed. No problem here what-so-ever. In my test set, only 1 filing had an XBRL technical syntax error. That filer is excluded as the information is unreliable because the technical syntax is unreliable.
- Edgar Filer Manual (EFM) automatable rules are conformed with: While not all EFM rules relate to the fundamental usability of the information, some do. Breaking these out is not that useful because so many filers pass all of these tests anyway. The specific EFM rules that are fundamental to reading the information are addressed below. And so, the 137 reports which do not pass automatable EFM rules are not considered trustworthy.
- Business report level model structure is conformed with: An XBRL-based financial filing is structured information and structured information, well, has a correct structure. This "model structure" is the relations between the different structural component categories: Network, Table, Axis, Member, LineItems, Concept, Abstract. (These rules are documented in the US GAAP XBRL Taxonomy Architecture. This blog post explains model structure in detail.) If you put an [Axis] within a [LineItems], what exactly do you mean? As such, the information in digital financial reports which does not follow the model structure rules is at a minimum ambiguous, at worst unreliable. And so, the 167 reports which don't follow these rules are not considered trustworthy.
- Root accounting entity discovered: If you can read all the information correctly thus far, you then need to discover the accounting entity or economic entity or the entity of focus of the report. That is all that this minimum criteria tests, that "root entity". Other related legal entities, business segments, geographic areas, and other such information breakdowns might be provided, but that is not what I am testing. Just the basics, the accounting entity of the report. The EFM has rules for articulating this entity of focus. This is easily discoverable for all public companies, except for 52. Because the accounting entity is ambiguous, the information in these 52 reports is unreliable.
- Current balance sheet date and year-to-date income statement period are discovered: Once you understand the accounting entity you need to work with, you need to understand the period within the report that you need to work with. Again, the EFM rules provide guidance on expressing this information, and all but 110 reporting entities follow that guidance so their current balance sheet date and the year-to-date income statement period is discovered. Again, clearly you will want to work with other periods, but that is NOT what I am testing, only the current period. And, if the current balance sheet date is not discovered and the year-to-date income statement is not discovered, the report is ambiguous.
- Reporting units discovered: (MISSING MINIMUM CRITERIA) So I made a mistake in my minimum criteria. The SEC also made a mistake and does not require a filer to indicate the monetary reporting units starting point. This is not a big deal because the vast majority of public companies report in US Dollars. However, not all do. This is a flaw in the minimum criteria, I don't know the extent that it impacts reported information, but it is low. As such, some entities that you might see are NOT in US Dollars. My bad, I will fix this. SEC might need to fix this also.
- Fundamental relations are intact: The fundamental trustworthiness of the information is checked using fundamental relations between reported facts. For example, "Assets = Liabilities and equity"; "Net income (loss) attributable to parent + Net income (loss) attributable to noncontrolling interest = Net income (loss)". For a comprehensive list of these fundamental relations see here. Now, that list needs to be tuned for the different ways entities report and for different accounting activities reported. For example, a bank reports differently than a retailer. As such, the fundamental relations are adjusted for this reality using what I call report frames. Here is a list of such report frames. This document helps you better understand report frames. If all of these expected relations are intact, that is evidence that the meaning of the financial report is being interpreted correctly. But if the relations are not correct, that means software cannot fundamentally interpret this information because of an error in representing the information or because of some ambiguity.
- Basic primary financial statement roll ups exist and are correct: There are four basic roll ups which must exist: on the balance sheet (assets, liabilities and equity); on the income statement (net income (loss)); and on the cash flow statement (net cash flow). If those exist and if they roll up correctly, you can be confident that the primary financial statement information for the root accounting entity and for the current balance sheet date and year-to-date income statement date (which both the income statement and cash flow statement use). But if the roll ups are not articulated or they are articulated and they don't actually roll up, then you cannot trust the information.
And that is the minimum criteria. It is not a lot, but it is necessary. I am not even saying that it is sufficient. What I am saying is that it is necessary. It is a foundation which can then be built upon. It is something that can be measured to determine how public companies are doing right now. It is something which can be used to manage additional improvements in XBRL-based financial filings of public companies.
And now half of the public companies are getting this correct. This is awesome. But what is more awesome is that you can see the 43.3% who are getting these minimum criteria wrong so that they can be corrected.