XBRL 2.0

Digital financial reporting that actually works

THE XBRL FILES: CONSENSUS TAGGING

In preparing XBRL filings, the process of tagging a company's line items to the standard US GAAP taxonomy is arduous. There are, after all, more than 20,000 standard items to choose from. But if the process is informed by the millions of tags that comprise the past 9 years of XBRL filings, the task is much less daunting. In fact, it's crucial to the usability of XBRL data that companies use this knowledge-base and make a reasonable effort to conform the with the group consensus. If consensus tagging resulted in greater tagging uniformity, it would begin to fulfill one of the principle goals of XBRL - comparability. Data that is more comparable is more valuable to investors and will ultimately lower the cost of standardized data generally, another goal of XBRL.

Consensus tagging is important for all tagging decisions, but particularly for decisions to use extensions, or custom tags. Extensions sabotage XBRL's purpose of standardizing financial data using a common language with a fixed set of concepts. The concept list is not fixed; it’s ever-growing due to extensions. The dictionary changes every day. Extensions are allowed by the SEC, but as a last resort, when no standard item comes close to the meaning of the company's item.

Unfortunately, every company interprets this admonition differently. Take the example below. Psivida Corp. (using Donnelley Financial software), created an extension for ‘Amortization of Debt Discount’. That’s odd, because it’s a very common item. And as you can see, two search methods (exact and partial matching with frequency counts) and an expert system (using block terms) arrive at a consensus standard tag for this item. Had consensus tagging results been integrated with the filing software, it’s unlikely that a custom tag would have been created.


This is not uncommon. In fact, further down the statement is another extension – Deferred Rent – that also has a strong consensus tag.

Most extensions can be avoided in this way. In addition, the flexibility of XBRL allows filers to: 1) choose a standard tag with approximately the same meaning as their item, then apply a label to convey the exact meaning, or 2) create a new axis off a standard line item and represent their item as a domain member. Both approaches are within the rules and both preclude the need for extensions that break the standardization.

So XBRL tagging shouldn't be done in isolation. Indirect collaboration via consensus tagging can substantially improve the usability of XBRL datasets.