With XBRL, the two primary issues have always been quality and comparability. Data quality is an important concern, but it's fixable and has improved over time. The lack of complete comparability, however, is built into the SEC's XBRL specification. Though not encouraged, the practice of adding company-specific items to the reporting taxonomy is certainly allowed. These custom items are called extensions and they're part of the original architecture of the "Extensible" Business Reporting Language. So the suggestion that XBRL is standardized is a myth. And it's the reason that almost 10 years after its inception, the biggest beneficiaries of the SEC's digital reporting initiative are not consumers of financial data, but compliance software vendors who assist in the preparation of filings.
Consider this example. McDonalds Corp reported its 2017 income statement using XBRL and tagging its line items to standard US GAAP line items...except for the four items shown in red below. Those are custom items created because McDonalds accounting department using Workiva software couldn't find comparable standard items.
So how can McDonalds be compared with their peers when significant items are non-standard? Well, here is what 3 XBRL-based data vendors did to standardize this income statement. Brands A (Intrinio) and B rolled up the detailed items into broader items, removing information that could be of value to investors. They did this for both custom and non-custom items (ie Revenue). And they arrived at very different answers. Brand C (Calcbench) didn't even attempt to produce a standardized income statement, instead offering customers to chose items for comparison that ARE standardized.
BRAND A BRAND B
XBRLogic's solution is to map the custom items to the most nearly comparable standard items in an automated process that leverages the millions of tags created by all companies for the past 6 years. Reasonable people can disagree over these mappings, but what's important is that no data is lost via rollups. The granularity remains and the underlying as-reported data is linked. This is the approach taken by the large data aggregators (Bloomberg, Factset, Thomson and Capital IQ). And it's the approach that the SEC should have taken when it implemented XBRL. There are ways that the SEC can repair this design error - requiring 1) that extensions be linked to standard items, or 2) that the existing dimension protocol be used for extensions. Comparability requires at least one fixed axis - the line item. Until the SEC's implementation is changed, the best solution is to map custom items to the standard taxonomy, illustrated by the example below.