Here is some pretty basic information helpful in understanding how to compare SEC XBRL financial filings. This exercise also helps you see how constructing a taxonomy makes things easier, or harder, to compare.
Consider this list of the 30 Dow Industrial Companies. You could look through those filings one by one, reviewing the "Document Information" section for each of the 30 companies. But what I did was put all the document information sections for all 30 companies into one single file, which you can see here.
I will assume that you have some list of companies which you might want to compare. So lets start at the level of the section that you want to compare. Here we will focus on the "Document Information" section because it is simple, it is actually rather easy to compare, but it does point out the issues.
The first thing you will need to do is determine what level you want to do a comparison. Do you want to compare networks, tables, concepts, or some other "level"? I don't want to get into a discussion about levels right now, so let us just assume that we want to compare networks. So no problem, we just go get the network name of the section we want to compare. Right?
Well, not so fast. If you scan throught the list you will notice that every company created a unique network for its document information section. They did this because SEC filing rules tell them that they need to do this. It could have been the case that one specific network URI was used to identify document information for everyone such as the URI used by the DEI taxonomy which is (you can see it in the DEI presentation linkbase here):
http://xbrl.sec.gov/dei/role/document/DocumentInformation
But one unique document information network was not used, so you have to try something else.
OK, so try something else. The SEC requires that the document information be marked with the category "Document" (as compared to "Statement" and "Disclosure"). Well, that won't work either as Travelers (number 26) used the category "Statement" category, as opposed to "Document".
But the good news is that every one of the 30 DOW Industrials used the term "Document and Entity Information". So, we can use that, as I did, to identify the document information section. Will that same technique work for all SEC XBRL filings? I doubt it, may need to adjust the algorithm.
The next thing you will note is that of the 30 companies, every one extended the root node, there is no "dei:DocumentInformationAbstract" concept, every filer created their own extension concept. One, again Travelers, did an utterly strange thing. They used a [Domain] (dei:EntityDomain) as the root of that section. That clearly should not be allowed.
Next you will note that of the 30 companies, 6 companies used [Table]s and 24 companies did not use [Table]s. Of the 6which did use [Table]s; all used the "Legal Entity [Axis]" and the "Entity [Domain]" to indicate that the information of the document information relates to the consolidated entity as opposed to a parent holding company. As such, SEC EFM rules kick in and you need to imply that the other 24 are consolidated entities, rather than parent holding companies.
If you scan through the concept names you will see that no filer created extension concepts. (This does not include that root abstract element which would never report information, it is only there to help organize the section. So while each filer reported different concepts, this information is quite easy to compare.
There are no real relations between the concepts other than the information is document or entity information. As such, there are no business rules we need to make sure are valid.
This same general approach can be utilized for comparing any information in SEC XBRL financial filings:
Easy enough steps, harder to apply because of inconsistencies in SEC XBRL financial filings, but quite doable. To make this easier, reduce the inconsistencies in the system.