XBRL provides an XML-based data format that enables electronic communication of business and financial information. With XBRL, every piece of information is tagged using standardized vocabulary defined in an XBRL taxonomy. For example, net revenue would have its own unique XBRL element.
XBRL International is a nonprofit organization that maintains the XBRL standard. More than 500 companies and government agencies are members, and the organization operates with more than 20 XBRL jurisdictions around the world. There are major government-sponsored XBRL initiatives in the US, Australia, Netherlands, India and many other countries.
In this article, learn about XBRL terminology and the steps for preparing an XBRL filing. From their experience, the authors include both general and specific lessons learned, and best practices, for creating XBRL format filings.
Differences between XML and XBRL
If you're familiar with XML, you might wonder what XBRL has that XML does not have. While it is true that XBRL is built on XML, and that all XBRL documents must also be XML valid, there are major differences between the two. The most important conceptual distinction is that XBRL provides a way to express and validate semantics. For example, Assets = Equity + Liabilities can be expressed using XBRL but not using native XML.
XBRL accommodates change better than XML. For example, XBRL separates the schema from the data. XBRL instance (data) documents are simple, containing normalized data immune to their sequence, and without hierarchy. Users can make changes to the XBRL taxonomy, such as changing parent-child relationships or adding new children, without having to reformat instance documents. Similarly, data in the instance document may be reordered without having an impact on the taxonomy. Normalized data also simplifies database storage.
Another unique feature of XBRL is its native support of multidimensional modeling, which is very similar to Microsoft® Excel pivot tables. Consider the situation where you want a metadata model to support segmentation of product sales by country and also by product line. Assuming a simple example of product sales by three countries (US, Canada, and Mexico) and two product lines (software and hardware), a pure element-based approach would result in 12 elements:
- sales_US_software
- sales_US_hardware
- sales_US_allProducts
- sales_Canada_software
- sales_Canada_hardware
- sales_Canada_allProducts
- sales_Mexico_software
- sales_Mexico_hardware
- sales_Mexico_allProducts
- sales_allCountries_software
- sales_allCountries_hardware
- sales_allCountries_allProducts
Adding another product to the product line adds another four elements. Add another line item, such as cost of goods sold, and that one additional line item would result in another twelve elements. Such an approach quickly gets out of control, and is extremely frustrating to model and manage.
Using XBRL dimensions, there would be the single element of sales, along with the two dimensions of country and product line, as shown in Figure 1. The country dimensions would have children for US, Canada, and Mexico; the product line dimensions would have children for software and hardware.
Figure 1. Country and product line dimensions easily reused for line items
The modularity greatly simplifies building the metadata model; the dimensions of country and product line can be reused for other elements, such as cost of goods sold, gross revenue, net revenue, and so on. It also showcases the semantic relationships between the values reported and their segmentation. For the adventurous, XBRL dimensions also support some set operations. The US-GAAP taxonomy (UGT), IFRS, COREP, and FINREP taxonomies make extensive use of dimensions.
The last noteworthy distinction of XBRL is validation. While XML validation is limited to syntax, XBRL validation confirms semantic relationships between elements (as defined in the XBRL taxonomy). For example, an XBRL validation engine confirms that the value reported for assets does, in fact, equal the sum of the values reported for equity and liabilities.
XBRL Formulas, a recent addition to the XBRL standard, provides a portable notation for the validation of other, more complex relationships between elements. For example, the FDIC uses XBRL formulas to capture the expertise of its auditors so that reviews can be performed in seconds rather than hours. The nonprofit Microfinance Exchange uses XBRL validation and formulas to examine the financial reports of thousands of microfinance organizations around the world. Using XBRL, analyst work at Microfinance Exchange that once took hours is now performed in seconds.
Key XBRL terms are:
- Concept
- Describes a discrete piece of business information. For example, "Assets, Total" label, documentation, or reference. A concept that defines a business fact, such as "Assets, Total," in XBRL is sometimes called an element—a term inherited from XML.
- Tuple
- A special kind of data type because it binds together a set of
concepts, like a row in a relational database where fields in a single
row are related. It used to be common usage to associate concepts
through tuples.
For example, annual reports require the disclosure of stock holdings of company officers. A tuple would be used to associate each officer's name with age, shares owned, and options held. The individual facts are meaningless without this association; you would not know which age went with which officer! Tuples may also contain other tuples.
- Dimensions
- XBRL dimensions have now taken the place of tuples for associating
concepts. Tuples are no longer permitted in the UGT or in a company's
filing taxonomy for two reasons:
- Tuples insert schema information into the instance document. Consequently, changes to tuple definitions in the taxonomy can "break" the instance document.
- Nested tuples are extremely awkward to modify, both in the taxonomy and in the instance document.
- Name
- The name of a concept uniquely identifies it in its namespace. A concept also has an ID formed by the concatenation of the namespace prefix with the concept name. For example, an element named Assets in the UGT taxonomy with the prefix us-gaap has the name us-gaap_Assets. This provides for unique identification of an element within a large, networked taxonomy (such as UGT).
- Relation
- A connection between two concepts, accomplished using the xlink
standard. The relation is always from one concept to another. It is
directional, based on xlink, with "from" and "to" as the endpoints.
The name of the relation is its role. A concept may participate in
many relations, such as a concept having multiple labels by language.
For example, first you would identify one endpoint of the relation. In this case, it is the concept for a value to be reported.
<loc xlink:type="locator"; xlink:label="IBM_NetAssets"; xlink:href="ibm.xsd#IBM_NetAssets" />Next, define the labels in different languages. These are other endpoints of the relation.
<label xlink:type="resource" xlink:label="IBM_NetAssets_lbl" xml:lang="en-US" xlink:role="http://www.xbrl.org/2003/role/label" >Net Assets</label> <label xlink:type="resource" xlink:label="IBM_ActivoNeto_lbl" xml:lang="es" xlink:role="http://www.xbrl.org/2003/role/label">Activo Neto</label>Establish the relations by creating arcs between the concepts and the two labels.
<labelArc xlink:type="arc" xlink:arcrole="http://www.xbrl.org/2003/arcrole/concept-label" xlink:from="IBM_NetAssets" xlink:to="IBM_NetAssets_lbl" /> <labelArc xlink:type="arc" xlink:arcrole="http://www.xbrl.org/2003/arcrole/concept-label" xlink:from="IBM_NetAssets" xlink:to="IBM_ActivoNeto_lbl" /> - Extended link
- A collection of concepts organized for a specific purpose. It
contains a network of relations between concepts, such as:
Assets, Total = Assets, Current + Assets, NoncurrentThere are two purposes for extended links:
- They make something big into something smaller and more modular.
- They isolate relations that might otherwise conflict. For
example, there are two ways to calculate Trade Receivables,
Net:
Trade Receivables, Gross
- Allowance for Bad Debt
-------------------------------
= Trade Receivables, NetTrade receivables, Net, Current
+ Trade receivables, Net, Noncurrent
-----------------------------------------
= Trade Receivables, Net
To capture both relations (because you want your taxonomy to be as comprehensive as possible), use extended links to isolate them from each other.
- Presentation relation
- Provides for a hierarchical organization of elements. The relations are from the parent to the child. These relations are stored in a presentation linkbase.
- Calculation relation
- Provides for summation organization of elements. These relations are stored in a calculation linkbase.
- Label
- A human-readable label for an element. Business users see labels. Taxonomy authors (taxonomists) see concept names. Labels have attributes for language and purpose. A label is assigned to a particular element through a relation established between the label and the element. These relations are stored in a label linkbase. Figure 2 shows an example.
Figure 2. The label arc links a label to an XBRL concept
- Documentation
- Describes the meaning of an element in human-readable terms. Consider
it a special type of label—a label whose role is for
documentation.
Documentation should be clear enough for a business user. However, documentation is not authoritative and is not meant to offer an official or exhaustive definition. Instead, references are intended to provide authoritative documentation. Documentation relations are stored in a documentation linkbase.
- Reference
- Provides a pointer to authoritative literature for an element. For example, the authoritative references for UGT concepts point to SEC and Financial Accounting Standards Board (FASB) rules. These relations are stored in a reference linkbase.
- Taxonomy
- A collection of elements, relations, labels, references, documentation, and references to other taxonomies. It is like a dictionary or classification system. Discoverable taxonomy set (DTS) is used to define the complete network of all referenced taxonomies and their respective file-based components.
The XBRL filing process has a single, clear goal: To transform an implicitly structured financial report into a structured financial report.
Preparing a filing involves making associations between pieces of the stated and implicit data (usually in a spreadsheet or word processing document) and an explicit informational structure described in XBRL. The UGT provides a starting point. The company filing taxonomy (CFT) extends the UGT to meet company-specific reporting needs (new elements, new labels for existing elements, changes in the schema, and so on). For example, the CFT can override the standard UGT label of "Assets, Total" to match what the company uses on its reports, such as "Total Assets."
An XBRL filing is composed of two logical components: an XBRL taxonomy, which describes the metadata of the filing company's financial report (built on the UGT), and the XBRL instance document that references the taxonomy.
Figure 3. Two logical components of a filing
The ancillary pieces created during the XBRL filing process are various kinds of project documentation. The documentation is important to both the technical team creating the technical documents, and to the financial reporting specialist concerned with securing auditable information about the creation of a specific filing. Figure 4 shows a broad overview of the process.
Figure 4. Large-grain view
The map step documents the associations between your financial report and the UGT, including changes and additions. Consider this a description of the perfect filing. The task then becomes implementing the map.
The model step implements your company filing taxonomy.
Tagging is the physical process of linking your financial facts to your company taxonomy.
During packaging, you create your filing package, complete your filing checklist, review the report with the client, and deliver it to the SEC or your financial printer.
Figure 5 shows the iterative nature of the entire filing process. It also shows an optional step where a company may use IBM DB2® 9 to store its XBRL-formatted financial reports, as well as reports from other companies downloaded from the SEC site to allow for comparisons and faster analysis of financials of various companies.
Figure 5. Detailed four-step process for filing
For the Q1 FY2008 filing by IBM, an XBRL descriptive table (XDT) was created to capture the mapping information. The XDT was based on the "neutral format table" work of Charles Hoffman, the acknowledged father of XBRL (see Resources). The neutral format table captures mapping information without regard to implementation technology. The XDT adds information necessary for XBRL implementation.
The XDT is modeled in a spreadsheet. It captures the information needed both for the taxonomy and also for the instance document. Figure 6 shows an example.
(See a larger version of Figure 6.)
Figure 6. XBRL descriptive table
The XDT is captured in an Excel spreadsheet. For the Q1 FY2008 filing, UBmatrix Report Builder was used to tag the data in the XDT spreadsheet to the CFT. Other tools to tag data include Dragon Tag by Rivet Software and Snappy Reports. A tagged cell is shown in Figure 7.
Figure 7. Tagging a cell with an XBRL concept
The XDT is organized by statements (for example, balance sheet, income statement). For each statement the XDT records the values to be reported, their labels, the name of each XBRL element, the reporting period, the decimal position for numeric values, and other information. As noted previously, this information documents the requirements for the CFT, which you subsequently implement in the modeling step.
Creating the CFT can be simple or complex, depending on the complexity of the information reported. It is relatively easy to model in XBRL the three standard financial statements associated with a 10Q (quarterly statement of financial position, statement of income, and statement of cash flow). A 10K (Annual Report) is much more complex because it includes the statement of shareholders' equity, which requires dimensional modeling. Modeling is performed in a taxonomy authoring tool, and there are now several of them on the market. For the IBM filings, UBmatrix Taxonomy Designer was used. Fujitsu XWand is another option.
(See a larger version of Figure 8.)
Figure 8. IBM Q1FY08 Taxonomy in UBMatrix Taxonomy Designer
Lessons learned from creating XBRL format filings for the SEC
For compliant and correct XBRL filings at the SEC, you should focus on acceptance criteria, lessons learned from experience, and best practices. The acceptance criteria, as defined by the SEC, are in the deceptively small document "Public Validation Criteria" (see Resources). XBRL.US provides a "Preparers Guide" (see Resources) that offers ideas for a workflow in creating a filing and some best practices.
However, experience demonstrates that experience is the best teacher! The authors have collective experience creating multiple filings for IBM and other large companies using three different versions of the UGT (2005, 2008 beta2, and release 1.0). This section outlines some of their lessons and best practices (all directed to creating SEC filings).
Table 1. General lessons learned and best practices
| Lesson | Best practice |
|---|---|
| If you've ever tried to learn the folk guitar, you know you can play pretty quickly, even if badly. Creating an XBRL taxonomy is quite similar. It's easy to create a taxonomy, particularly a bad one. "Bad" means it fails to apply the architecture rules of the UGT, selects the wrong UGT concept for a financial figure, fails to accurately represent your company's financial reporting model, renders statements that do not match your other data formats, and so on. | Have at least two people on the team, one an expert in XBRL and another in the financial reporting practices of the specific company. Ideally, each expert should know something about the other's area of expertise. |
| It takes longer than you think. Even with the help of someone experienced in this process, and with a good process to follow, the preparation of a 10Q will take from 32 to 40 hours (not including training time). | Establish a reliable internal process. A reliable process is especially important because the SEC will soon require concurrent delivery in all data formats. Imagine the fallout for your company if it delays the release of its quarterly figures by a week or two because the XBRL filing is taking longer than anticipated. You should never have to start from scratch except on your first attempt. Make that first attempt while you still have time to make and learn from mistakes (before the end of 2008). |
| This is not a linear task. Each step involves iterations of testing and revision. The iterative process should be done for each financial statement. You might think you're near the end of the project only to discover (surprise!) that you've picked an incorrect UGT concept for a financial figure. | Plan for iterative development and lots of reviews. |
| Your filing is not perfect. Currently, there are many
opportunities for human error, including the human-dependent steps
in examining the quality of your taxonomy and instance documents.
The service industry arising from the SEC mandate is slowly
recognizing the challenges in detailed tagging of every piece of
data in a financial report. There are technical issues and process
issues. But, most critically from the perspective of a CFO, who
carries personal liability for the accuracy of the data,
there are accountability challenges. For example, how do you know you've made the right mapping of figures to XBRL concepts? How do you know that you have complied with UGT architecture guidelines? How do you know that cross-reporting-period relationships are correctly reported? At this time, nearly every company filing XBRL reports is using eyeballs to ensure accuracy. As the number of reported facts climbs into the hundreds, and the complexity of the data model grows, human verification of data accuracy will clearly be an unacceptable method. | Plan for perfection by understanding your requirements. Then push on the XBRL services industry, tools vendors, and UGT steering committee to create the tools and XBRL formulas that you need to support the universal requirement for demonstrably correct filings. |
Table 2. Taxonomy lessons learned and best practices
| Lesson | Best practice |
|---|---|
| The first attempts to create our CFT involved modifying the UGT. XBRL provides a clean mechanism for extending other taxonomies, so the technology is well-suited for SEC filings. Every filing company begins with the UGT, then customizes it (customization is nondestructive to the UGT). Yet some customization was so extreme, it was more efficient to create a new network of relations rather than customize an existing taxonomy. | Create company-specific extended links for each statement unless
the UGT provides one that's a near-perfect match. Think of the
company-specific extended link as a definitive statement. Rather
than seeing how the company deviates from the UGT (GAAP is
Generally Accepted Accounting Principles), simply
ensure your company-specific extended links make a clear statement
(pun intended). Once you've completed your company-specific extended links, the next best practice is to remove the references to the UGT equivalents. They are no longer needed. |
| It was a challenge to create presentation relations that represented the correct data model relationships between concepts, and that would also support a desired rendering (sequence and indentation) based on the presentation's parent-to-child relations. This is an overloading of the presentation relations. They cannot serve two masters. | Use the presentation relations to secure the best data model, while sacrificing the data display. In the near term, particularly with the 10Q, making this sacrifice is not likely to present problems. |
| The UGT is big. Comprising than 150 taxonomies, more than 550 files, and over 12,000 concepts, it is over 55MB in size. In the early work, a UGT entry point was used that pulled in nearly everything. The team thought that was smart, but turned out to be wrong. Performance became a problem (especially as the files are loaded over the Internet), and they had to sift through networks of relations that they did not even need. | Identify the exact statements and disclosures you need for
reporting, then reference those taxonomies (a subset of
the 100) in your CFT.
|
Tools will get better. But for now, this is our recommendation for how to identify the exact statements and disclosures you will need for reporting.
Table 3. Instance document lessons learned and best practices
| Lesson | Best practice | ||||||||||||||||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Financial reports often have figures in millions (for example, from the cash flow statement), others as whole numbers (such as shares outstanding), and some with two decimals places (earnings per share). | For handling the millions, report these values as whole numbers and indicate that they've been rounded to the nearest million. If your financial report provides, for example, a net revenue of $26,311 in millions for Q4 of 2007 (IBM is a big company), the value in the instance document would be reported as 26311000000 with a decimal attribute of -6 (which is substantially different from reporting it with decimal attribute of 0). | ||||||||||||||||||||||||||
| Instance documents also contain other details, including definitions for the numeric unit, context, and footnotes. Complex filings may have a hundred or more of these details. | Establish a naming convention so you'll know how to quickly
distinguish one context from another. For units, we recommend
U-[unit]_[decimal]. For negative decimal values, a minus sign (-) or
dash (--) replaces the underscore (_).
For nondimensional contexts, [period type]-[relative fiscal year]-[quarter+number, or annual] is recommended. Period type is either I for instant or D for duration. Relative fiscal year is C for current year, P for prior year, PP for prior-prior year, N is for next fiscal year (needed for some disclosures), and year-to-date reporting period is YTD. Assuming the reporting period is a third-quarter report as of 9-30-2008:
Assuming the reporting period is for a fiscal year ending 12-31-2008:
Dimensional contexts are more complex to name because of their
length. Include the domain name in full, then an acronym
representing the domain member. For example, an annual report
disclosure about business combinations has the name "Business
Segment [Domain]" and a member "Global Systems Technology
[Member]." If the duration reporting period is current fiscal
year, year, combined with the domain and member information
noted above, the dimensional context would appear as: |
Thanks to Bill Hoffay at IBM, who was and remains an integral part of producing these filings. And, thanks to Charles Hoffman and Timothy Randle for sharing their design experience and insights on preparing SEC quarterly filings.
Use of XBRL is growing worldwide, largely because of its powerful capabilities. An equally compelling reason for XBRL's growth is its adoption by regulatory agencies and the SEC mandate that all filings be in XBRL format within the next two years. XBRL will certainly disrupt many current methods of financial reporting. There is already significant XBRL activity in the service companies and organizations that support filings to the SEC, including the accounting industry, financial printers, and financial reporting departments of the largest companies. In this article, you learned from the authors' experiences about the opportunities in this challenging area.
Learn
- XBRL standards are maintained by
XBRL International, a not-for-profit
consortium of approximately 550 companies and agencies worldwide working
together to build the XBRL language and promote and support its adoption.
- Nonprofit
Microfinance Exchange, the world's
leading business information provider for the microfinance industry, uses
XBRL validation and formulae to analyze financial reports.
- Charles Hoffman, the widely acknowledged
father of XBRL, provides information on
the neutral format table and other
XBRL information.
- The SEC publishes XBRL
Public Validation Criteria
that reflect SEC staff’s current views on appropriate validation criteria
for XBRL tagging to improve the consistency and quality of XBRL documents
submitted to the SEC. The list will help filers and data-tagging software
companies improve existing products or create new products to automate
testing for these criteria.
- The
XBRL US GAAP
Taxonomy Preparers Guide
is a good starting point for preparing XBRL filings. It explains XBRL
terminology and concepts so preparers understand how to format financial
statements in XBRL.
- Learn more about
IBM DB2,
which offers industry-leading performance, scalability, and reliability on your
choice of platform from Linux® to z/OS®.
- Read about Fujitsu
XWand,
an application development and runtime environment for building and
deploying full-feature, industrial-strength XBRL applications.
-
Rivet Software
develops financial communications and analysis software, including XBRL
creation and viewing applications, and Crossfire Analyst, a
Web-based financial aggregation, analysis, and publishing solution
designed for accounting, finance, and other business users.
-
UBMatrix, Inc XBRL
solutions increase operational efficiency and financial transparency, and
ensure reporting accuracy and regulatory compliance.
- The United States Securities and Exchange
Commission
XBRL section
provides information about interactive data, news items, and more.
- The
SEC Viewer
lets you view different taxonomies.
- Browse the
technology bookstore
for books on these and other technical topics.
- Browse the
technology
bookstore
for books on these and other technical topics.
- In the
Architecture area on developerWorks,
get the resources you need to advance your skills in the architecture
arena.
- Find resources to help you architect enterprise and software systems in
free IT architecture kits from IBM.
- Stay current with
developerWorks
technical events and webcasts.
Get products and technologies
- Download
IBM product evaluation versions
and get your hands on application development tools and middleware
products from IBM, DB2, Lotus®, Rational®,
Tivoli®, and WebSphere.
Discuss
- Participate in the
IT architecture forum to exchange tips and techniques and to share other related information about the broad topic of IT architecture.
David Newman is an advisory applications developer. He was the technical team lead for the IBM XBRL filings in the SEC's Voluntary Filing Program in 2007 and 2008. He successfully prepared the 2007 10K XBRL submission per the more detailed filing requirements from the SEC's XBRL Mandate.
Dean Ritz is an independent XBRL consultant. He is currently helping IBM and other companies in their preparation of XBRL format filings to meet the mandatory filing requirements coming into effect January 2009. He formerly was a senior taxonomy developer in the professional services group at the XBRL technology vendor UBmatrix, Inc.
Murali Vridhachalam, an Open Group certified IT architect, has been involved with XBRL since early 2005. He was the lead architect for IBM's first ever submission of financial reports using XBRL, as part of the SEC's XBRL voluntary filing program. His current interests include SOA and Software as a Service offerings built using the IBM enterprise software portfolio.
Comments (Undergoing maintenance)
Table of contents
Next steps from IBM
With the no-charge DB2 Express-C, pureXML, Zend Core PHP, and Java you can develope advanced Web 2.0 applications quickly that matter.
- Try: Download a free trial version of DB2 Express-C, which includes Zend Core PHP and pureXML technology that is ideally suited for managing both XML and relational data.
- Article: See how DB2 pureXML schema capabilities can evolve schemas without making any changes to your applications.
- Tutorial: This tutorial shows you how to use DB2's pureXML to expose the native XML capabilities of DB2.
- Buy: DB2 9 for Linux, UNIX and Windows
My developerWorks community
Tags
Use the slider bar to see more or fewer tags.
Popular tags shows the top tags for this particular content zone (for example, Java technology, Linux, WebSphere).
My tags shows your tags for this particular content zone (for example, Java technology, Linux, WebSphere).
Popular article tags |
My article tagsSkip to tags list
Popular article tags |
My article tags
