Web Content:
How to Get it Right

by Michael P. Voelker

Content management systems have universally attacked the problem of "webmaster bottleneck" by moving Web content creation and publication responsibilities from the Web team to content creators themselves. Unless businesses that implement Web content management (WCM) systems are careful, however, this new workflow can create its own problems.

"When businesses print materials on paper, they go through endless iterations to make sure the content is correct," says Christopher Harris-Jones, principal at the London-based consultancy, Ovum. "There's a danger with Web content management systems that content does not go through the approval it needs because it's so easy to publish."

Therefore, the WCM "users publish" model requires well-designed approval and staging processes to ensure that the right content gets published in the right format at the right time.

The approval process in most WCM systems actually begins before content is ever created. WCM workflows regulate who can create and modify content via user/role access controls, and in what format that content should ultimately be displayed via templates. These access controls must recognize that documents contain many pieces of content, and a single piece of content can be linked to any number of documents on or off the Web site in either electronic or printed form.

"You need a granular permission system that gives rights to edit and update content," says Hank Barnes, vice president of strategy at Eprise, Framingham, MA. "There are specific areas you will let people change, and areas that you won't."

Eprise's Participant Server WCM platform handles permissions by controlling viewing and editing at multiple levels: by individual content elements, by blocks of content and by pages assembled from elements and blocks. To address creation and editing privileges, Participant Server provides its own ID and password authentication, and it interfaces with existing security schemes such as LDAP servers and NT domain authentication.

Templates are also part of the WCM approval workflow. In effect, templates enforce an approved format, ensuring consistency of page design or even the wording of common information, such as headers and footers in press releases. Templates also allow authors to see the results of their work in WYSIWYG format and help get content to the Web more quickly.

Also important to the approval process (as well as for general content creation and collaboration), WCM systems include check out/check in controls to ensure that content can't be edited by multiple users concurrently. This version control also ensures that when modifications to content are made, the original content is still available for comparative approval. Auditing can track who made which changes when, and rollback capabilities let users return to a previous version, if desired.

When it comes to version control, more is definitely better. "It's not difficult to track a version, but most systems don't track attribute values," says Whitney Tidmarsh, vice president of product marketing at Documentum, Pleasanton, CA. "Let's say version one is marked 'draft,' and it has certain permissions. It only exists in its original format in an isolated part of the repository. Contrast that with version six that has gone through multiple review cycles. It has now been linked into the site, it's been rendered in HTML and PDF, it has links to other content and it has been translated into multiple languages." Documentum 4i WCM Edition tracks all these values, says Tidmarsh.

Match Approval to Content

When an author is satisfied with his or her document, it's off to the next approval stage. But what that next stage might be should depend upon the type of content and its intended destination. While an author may be able to publish directly to a corporate intranet with little or no review, an entirely different workflow is likely required for press releases, financial results postings and product information.

The ability to subject content to different levels of approval is a crucial feature. "Rather than mandate a single process for all content, management can evaluate the intended use and potential liability associated with specific content and let that drive the depth of the approval process," says Barnes of Eprise.

All WCM systems provide approval routing via proprietary workflow systems, and most use a browser-based interface to provide users with a familiar look and feel. At a minimum, workflow systems post tasks to queuing systems, requiring users to seek out work or approvals pending for them. Some systems, in contrast, actively push tasks to the next person in the workflow.

For example, TeamSite 5.0 from Interwoven, Sunnyvale, CA, notifies users of pending tasks via email. "When the user submits [a content change], it will automatically generate an email with an embedded link," says Tim Hampson, vice president of enterprise marketing at Interwoven. "That link will pull up a tab that would allow you to approve or modify [the content] and, in turn, send it back to the author."

Set the Stage for Delivery

Of course, simply because content is approved doesn't mean it should immediately be published. To maintain accurate and timely site information, most WCM systems allow content to be staged for later deployment set by "effective dates." Removing content with preset "expiration dates" from the site is also important. Some systems manage staging directly, while others use separate staging servers to hold content pending migration to a live server.

To this point, all the content approval and staging mechanisms are dealing with content that has been directly created by business users. But what about pages that are dynamically generated from pieces of information stored in databases or Web content stores?

"A dynamically generated page is generally composed of a template, of content that can come from a variety of sources and of rules that decide which piece or pieces of content to display," says Imad Mouline, director of engineering at BroadVision, Redwood City, CA. "A content management system needs to be able to manage all of the above."

In short, dynamic pages differ from static pages because they are assembled according to prebuilt page generation definitions rather than human review and approval workflows. The danger is that while the WCM system handles assembly of dynamic pages, correct data entry and edits of external data are outside of its scope. Even with robust workflow and life cycle controls, things can still go wrong any time human beings - and their human-designed systems - are involved. The first priority when disaster strikes is to stop the bleeding; the second is to find out what happened in order to prevent future recurrence.

"The problem businesses face is how to quickly correct mistakes when they occur," says Hampson. Interwoven's OpenDeploy provides a rollback system to quickly pull Web content and return it to a previous state. "If there is a problem, you would be able to roll back in a matter of minutes," Hampson says. Documentum also incorporates a "byte-level differencing" self-repair feature that can compare deployed content with the last "good" version of content.

Audit to Eliminate Error

After errors are fixed, it's time to find out what broke down in the creation, approval or deployment process by reviewing the audit trail. Just as the level of approval for content is variable, so, too, is the level of audit provided by most WCM systems. In general, however, audit systems capture several key components.

"The audit subsystem must capture which user performed what action on what object, and when," Mouline explains. "Additionally, the audit system must be able to keep track of what that object looked like in relation to other objects in the system at specific points in time. This is important because the original piece of content does not necessarily tell you what a page looked like."

BroadVision's One-To-One Publishing system takes snapshots of not only pieces of content but of a hierarchy of content at each update point. "This allows us to go back and [ask], 'What was the last known good snapshot taken?'" explains Mike McGahey, technical director for BroadVision global services.

Why pay so much attention to approval, staging, auditing and rollback? A quick review of headlines over the last two years highlights that Web content errors aren't limited to problems with the creation process. There have been outright errors. Earlier this year, Argos, a U.K. retailer, discovered that it was selling color TVs online for about $5. In another instance, a few lucky travelers managed to buy round-trip tickets to Paris for less than $30 on the United Airlines site.

There have also been problems related to the premature delivery of staged content. For example, the Associated Press incorrectly informed the world in 1999 that Bob Hope had died, and Sun Microsystems released its earnings a wee bit too soon in October of last year.

When you dig deeply, the root causes for most errors are not technological, say independent experts. "It's primarily due to human beings forgetting to do something," says Mark Gilbert, research director at Gartner, Stamford, CT. "There may be something in the database that's wrong, something doesn't get coded right. When you track it all back, the software only does what we tell it to do."

Can WCM systems prevent headline-making mistakes? Perhaps. The approval workflows of WCM systems allow reviews of author-generated content at both the business and technical level and control of the release of staged content based on dates provided by authors or administrators. Unfortunately, WCM systems have no way to know if a given release date was correct or if pricing or product information used in dynamically generated pages was accurately entered.

In the end, therefore, people - not technology - make the biggest difference in getting content right. By putting the ultimate approval and delivery of content directly in the hands of those who know it best, Web content management systems help businesses improve their odds of preventing content errors. As always, good tools are only complements to good business practices.

Michael P. Voelker is a freelance writer based in St.Cloud, MN.