Boxes and Arrows: Stories by Christian Ricci

Personalization is not Technology: Using Web Personalization to Promote your Business Goal

Christian Ricci — Mon, 12 Jan 2004 21:33:22 GMT

“Web personalization is a strategy, a marketing tool, and an art.”

Personalization, properly implemented, brings focus to your message and delivers an experience that is visitor-oriented, quick to inform, and relevant. Personalization, poorly implemented, complicates the user experience and orphans content. If you are a web strategist, designer, or content manager, you are undoubtedly familiar with the value of web personalization. For years, we've been working to ease the complexity associated with authoring, delivering, and consuming rich, dynamic content via browser-based applications. Introducing more content, and varying it routinely, may make your site fresh, but it can also have a negative impact on your overall message. When does freshness become noise? And how can personalization cut through the clutter?

What is web personalization?

Web personalization is a strategy, a marketing tool, and an art. Personalization requires implicitly or explicitly collecting visitor information and leveraging that knowledge in your content delivery framework to manipulate what information you present to your users and how you present it. Correctly executed, personalization of the visitor's experience makes his time on your site, or in your application, more productive and engaging. Personalization can also be valuable to you and your organization, because it drives desired business results such as increasing visitor response or promoting customer retention. Unfortunately, personalization for its own sake has the potential to increase the complexity of your site interface and drive inefficiency into your architecture. It might even compromise the effectiveness of your marketing message or, worse, impair the user's experience. Few businesses are willing to sacrifice their core message for the sake of a few trick web pages. Contrary to popular belief, personalization doesn't have to take the form of customized content portals, popularized in the mid-to-late 90s by snap.com and My Yahoo!. Nor does personalization require expensive applications or live-in consultants. Personalization can be as blatant or as understated as you want it to be. It's a tired old yarn, but if you hope to implement a web personalization strategy, the first and most important step is to develop and mature your business goals and requirements. It is important to detail what it is you hope to do and, from that knowledge, develop an understanding of how you get from an idea to implementation. You might be surprised to discover that it won't require most of next year's budget to achieve worthwhile results.

What makes personalization successful?

Too frequently, personalization initiatives die on the white board. It can seem a daunting task when development teams gather to consider technical and business requirements (such as changes to architecture, user profile storage and analysis, and content management). Analysis paralysis kills personalization projects early and often because teams overreach. So what's the key to successfully implementing personalization initiatives? Start small and pick achievable goals that integrate well into your existing presentation framework. Think of personalization as a way to enable your business plan. Over time, with successful implementations, it can become an enabling technology; a component of your overall marketing strategy, your communication message, even branding. However, in order to accomplish any level of personalization, whether it's for your internet, intranet, or extranet site, you need:

A high-level driver, owner, and/or sponsor
This should be someone in management, executive management, or at the C-level who has ownership of the “bottom-line” results.
Measurable business goals
Your personalization initiatives must be measured against practical and relevant business metrics.
Long-term commitment
This is an iterative process; some phases will be very successful, others will be less so.

Most importantly, keep the process simple. Stay focused on the business goals, tackle manageable projects, measure the success or failure of your changes, and learn from your mistakes.

What are your business requirements?

Think through this carefully. What are your business goals? How can you turn these business goals into personalization business requirements? By giving prudent forethought to maturing your intention and measuring your results, you can keep the process well focused. For example, if your goal is to increase sales revenue, you might use personalization to better transition anonymous internet visitors to sales leads. Or, if your goal is to decrease software support costs, you might use personalization to promote online support tools for an application or service that you know a specific user is interested in.

How are you going to do it?

Once the business requirements are well defined and understood, refine and elaborate upon them until you can develop use cases to support the end goal. I am using the software engineer's definition of “use case” here, focused on describing the precise behavior of the application, not necessarily the user interface. For example, if your goal is to collect more email addresses from job-seeking internet site users, your use case might explain how you intend to identify visitors as job-seekers, how you will prompt them for their email addresses, and how they will be rewarded for providing the information. (Remember, these are your customers. Don't force them to provide data. And when they do provide personal details, offer them tangible rewards for doing so.) User interface design, when implementing personalization initiatives, remains an important part of the design process. In fact, careful user interface design may be more important than ever. Don't allow your modified presentation framework to become a barrier to end users, compromising your message or intentions. Keep in mind:

This is a partnership
You are engaging in a partnership with your visitor, using what they share with you, explicitly and implicitly, to facilitate a more productive relationship. They need to trust you and you need to honor their wishes. These objectives may manifest themselves in the user interface.
The message is still key
When choosing to display or hide content from your site visitor based on a personalization initiative, you need to fully understand the ramifications of such an effort. Will this adaptation of the user interface render some content inaccessible, or orphaned? Will this adaptation of the user interface alter the presentation such that the overall integrity of your site is compromised?

If business goals describe what you want, business requirements describe what you need to do, and use cases describe how you plan on doing it.

Who is your visitor?

From an understanding of your business requirements, develop a visitor profile definition and visitor segments. A visitor profile is a collection of attributes that you'll need to either maintain or derive in order to support personalization. Implicit profile attributes can be derived from browsing patterns, cookies, and other sources. Explicit profile attributes come from online questionnaires, registration forms, integrated CRM or sales force automation tools, and legacy or existing databases. In short, explicit profile attributes come from customer responses, while implicit profile attributes come from watching or interpreting customer behavior. A visitor segment is a collection of users with matching profiles. Certainly, a loose definition of target segments may develop as business requirements mature. After all, these are the people you strive to reach with your personalization initiatives. Visitor segments may be very broad or very confined in scope. However, once a visitor's attributes and the mechanics of maintaining and collecting visitor profile data are known, rules can be developed that formally define segments. Sample visitor segments might include registered site users who have not purchased any services, customers who have not purchased a service in more than 12 months or, simply, investors. How you collect and store this information is a sensitive and timely topic. In many parts of the world, and among some segments of the internet community, cookies are despised. Take this into account when determining what data you have access to and how you leverage it.

How do you measure success?

How will you measure the success or failure of your personalization business requirements once they become technical deliverables? It is important to measure success or failure in any personalization exercise. Failures need to be eliminated before they cause further trouble. Successes can be used to drive further financial, time, and personnel investment. As you determine your business goals, requirements, and use cases, keep in mind what sort of metrics you can collect before and after implementing any changes to your user interfaces. Also, try to determine how this data should change as a result of personalization.

Case Study: Improving the Effectiveness of an Internet Site for Human Resources

What is the business requirement?

To enable Human Resources to increase their pool of candidates, and improve their ability to leverage information about existing candidates using an existing internet site.

How are we going to meet the requirement (what are our use cases)?

The web delivery application will detect first-time website visitors browsing the job openings page. These visitors will be prompted for their email address and given an opportunity to register for job opening announcements by email.
The web delivery application will detect returning website visitors interested in job openings, offering them a chance to register for email announcements (see above) and a chance to win a new laptop computer. Visitors who register to win the laptop will provide their name, address, email address (if unknown to us), and phone number.
When a known visitor submits a resume for a job opening, additional profile information will be collected. Known attributes (name, address, email address) will be populated from the profile.
All visitor profile information collected will be stored in an internally accessible database and used by the HR department to promote job openings and career fairs that might be of interest to the candidate.

Who is the visitor/What is the visitor profile?

The following information can be collected and associated with the user in question:

Number of site visits
Name
Address
Email address
Phone number
Resume
Interest in job openings (implicitly derived—based on browsing patterns)

How will we measure success?

By reduced recruitment costs due to lessening the time it takes to fill job openings and eliminating recruiting expenses.
By an increased number of candidates hired via website.

Summary

Personalization may be tough to define and hard to measure, but it doesn't require a rocket scientist or piles of cash to accomplish. As with most business initiatives, developing that first business requirement and making the first commitment, right or wrong, is the hardest step. The software market is flooded with companies ready to sell you an off-the-shelf, shrink-wrapped personalization solution. Unfortunately, what buyers don't often realize until it's too late is that personalization isn't a plug-and-play solution. Know your goals and stay focused on long-term improvements by following these steps:

Define your business goals.
Convert your business goals into personalization business requirements.
Convert your business requirements into use cases.
Define the user profile and formally define the user segment(s).
Determine which metrics you will use to evaluate the initiative.
Implement.
Repeat.

Personalization requires analysis of your goals and the development of business requirements, use cases, and metrics. Once these are fully understood, you may find that your personalization strategy doesn't require substantial augmentation of your application environment. If you do find that the integration of a personalization tool is necessary, with this knowledge, you'll be able to better analyze and judge the offerings.

Christian Ricci is a consultant, application developer, web designer, and project manager with over 11 years of experience in software design and development, network and server administration, and software project management and engineering. As a Senior Solutions Architect for Saillant Consulting Group, Chris has led portal, content, and document management projects for Qualcomm, Intermountain Health Care, J.D. Edwards, EAS, and the Denver Post.

Site: http://www.chiamonkey.com/
Resume: http://www.chiamonkey.com/mealticket

Developing and Creatively Leveraging Hierarchical Metadata and Taxonomy

Christian Ricci — Sun, 23 May 2004 04:27:58 GMT

When confronted with projects requiring content, document or knowledge management, and presentation, more likely than not, the information architect will be expected to lead or contribute to development of the content classification requirements. And we don't classify our content without reason. "In content metadata and hierarchies, you will often find a goldmine of implicit and explicit data that you can leverage to creatively contextualize content." As site creators, it would be time-consuming, expensive, and contentious to develop and maintain the necessary infrastructure and processes to manage unorganized content. As site users, it would be maddening to try to sift through the links in Yahoo's directory trees if contributors and reviewers hadn't organized them ahead of time. By organizing content for presentation according to its metadata, we can "contextualize" for potential users. In content metadata and hierarchies, however, you will often find a goldmine of implicit and explicit data that you can leverage to creatively contextualize content. Following a brief introduction on taxonomy and metadata (what I call content classification requirements), this article will focus on finding and utilizing such relationships in hierarchies. Content classification requirements: taxonomy and metadata "Taxonomy" is a terribly overused term these days. Bob Boiko, in the book Content Management Bible, goes so far as to call it "trendy." Specifically, taxonomy is a hierarchical structure for the classification or organization of data. Historically used by biologists to classify plants or animals according to a set of natural relationships, in content management and information architecture, we tend to leverage taxonomies as a tool for organizing content (For additional information, see Christina Wodtke's interview with Samantha Bailey elsewhere on Boxes and Arrows). Metadata (data about data) describes an asset and provides us with a meaningful set of attributes that we can use to further classify or consume content. While much metadata is flat or one-dimensional in nature (e.g., size or weight), some of it is hierarchical (e.g., taxonomies), making the definition and distinction between metadata and taxonomy vague and fuzzy. Collectively, I tend to manage taxonomy and metadata needs as simply content classification requirements; taxonomy as a means of organizing content and metadata as method of further describing it. If your task, for example, is to classify independent musicians for a website and database, you may choose a taxonomy that organizes the artists by musical style and then create metadata to describe the members of the band, the year of inception, record label, discography, and geography. Figure 1: Sample taxonomy for classifying music and musicians (Click to enlarge).

Geography is particularly interesting as it has the potential to be hierarchical metadata itself. Let's say your database expands to include musicians from around the world. Your hierarchy of geography could include country, state or providence, region and/or city. That's a whole new hierarchy, perhaps even a taxonomy. Geography, by itself, has become an effective and alternative means of organizing your content. Figure 2: Sample geography taxonomy How to develop content classification requirements When developing a content classification strategy, it's important that you know your needs, your application, and the technical limitations of your software infrastructure and content producers. It's quite possible that your content management and delivery needs will tax the capabilities and APIs of your content management tools. It's also quite possible that you will severely overtax the abilities and/or patience of your content producers (see sidebar). If there's development to be done or sacrifices to be made, keep these constraints in mind as you design the application. There's a very good chance that you will need to revisit and revise your content classification strategy periodically. Life changes, business changes. It is more important that you plan for and design a mechanism and set of policies for easy adaptation, than waste countless hours fine-tuning your taxonomy here and now. Therefore, here are my guidelines for development of your content classification requirements:

Address governance as early as possible in the design cycle. How will your content classification get revised and extended? How frequently? What person or team will be responsible for the maintenance of your metadata? How will you settle disputes?
Identify the scope. Are you organizing content for creation? Consumption? Both? A specific department? Some combination of departments? Intranet? Extranet?

Admittedly, it's important that you not rule out the eventual integration of this work with other initiatives and parts of your business but consider the value of bounding your project. Scope creep is a serious risk as you begin cataloging and classifying your content. Instead of solving all the organization's problems right now, invest time and energy in defining how your classification requirements can and will change over time.
How does the scope of this application relate to your business organization. Are you striving for the development of an enterprise taxonomy? Enterprise taxonomies are a popular concept right now, especially in organizations initiating development of 3rd and 4th generation intranets. An enterprise taxonomy is typically perceived as a single, monolithic, corporation-wide, structure for the classification of all things related to your business. Unfortunately, development of an enterprise taxonomy requires the careful coordination, and cooperation, of departments within your organization. Will you be able to coordinate the efforts, language and needs of your IT and sales departments? If so, kudos to you. If not, welcome to the real world. Disparate divisions within an organization will often use different terms when referring to the same thing. This means, more often than not, the scope of your content classification requirements will be constrained to some subset of the organization or to a particular business process.
Catalog your content. The exercise of cataloging your content can be very informative. A lot of guidance, limitations, and input can be achieved during this process. More often than not, your classification requirements will vary by content class or type.
Focus on developing good criteria for the definition and extension of your metadata and taxonomy. Good rules will enable those tasked with management to respond quickly and with reason to requests for adaptation.

Organizing content for management and delivery So now we can build a strategy for classifying the content we hope to manage and we have attributes. Now, we need a method for presenting our content. Strategies for getting content from a management infrastructure to the delivery framework (internet, intranet, extranet, client-server application, XML-web services, etc.) vary. Here are two common techniques: 1. Universal Hierarchy A single hierarchy could be used to store and deliver content. When content contributors utilize the content management system, they add, remove, and manage content in a structure that closely resembles the navigation and hierarchy of the delivery framework (your website or application). The navigation structure is your taxonomy. Figure 3: Organizing content for delivery by Universal Hierarchy (Click to enlarge)

This method is conceptually simple and makes it quite easy to dynamically build your navigation from knowledge of this hierarchy. However, this model does have drawbacks:

Every time you reorganize the website, the organization of content in your management application shifts. Admittedly, this isn't much of a drawback if you're managing content for one moderately sized site or if your team of contributors is small.
It is difficult to reuse content in this structure. If you hope to reuse assets throughout your website, where are they organized in this structure?
In an environment with many contributors and diverse security requirements, organizing content (in the management application) in another way, say by contributor or by department, may be more intuitive.

2. Content Mapping A more robust, albeit more complex, method of managing content is to maintain structures and metadata in the content management application that is independent of the delivery system's organization (navigation). Figure 4: Organizing content for delivery by Content Mapping (Click to enlarge)

Content is organized, at the source, as may be required by your security, workflow, or organizational needs. Perhaps your data lives in a content management system or database where different organizational mechanisms exist. Unfortunately, oftentimes the navigation for your consuming application (the presentation framework) is managed by some other means. By some rule or algorithm, leveraging your content classification data, material gets "mapped" to the presentation framework. See the example below for an application of this model. This model is rich with possibilities:

There may be more than one way to organize content (think: content reuse). Given the same set of content, same set of classification criteria, but multiple algorithms, we can now build a delivery framework that allows for many methods of organization.
You no longer need to reorganize your content management application to change the delivery application. Just the algorithms (mappings) change.

However:

If you hope to build your navigation dynamically, often you'll need to build a tool or alternate hierarchy. You may not find much value in the content's taxonomy.
Content, in your management environment, may be orphaned in your presentation framework if there are no rules mapping to an accessible part of the site.
Parts of the site may only be sparsely populated. It may not be readily obvious that you are creating gaps (with little or no content) in your site.

While powerful, this technique can be difficult to administer without having a fairly comprehensive understanding of the site design and algorithms for "mapping." Creative contextualization of content using hierarchical metadata or taxonomy Assuming there are hierarchical structures within your content classification system, there is a very good chance that valuable information exists in the hierarchy. By taking advantage of relationships within your hierarchical metadata structures, richer algorithms may be developed for your content delivery framework. Lets identify some of these relationships and how you can leverage them:

Ancestors In the example below, North America is an ancestor of America, which is an ancestor of West (and Washington - WA). Ancestors can be valuable because content classified under an ancestor node may have relevance (albeit with less specificity) to child nodes. If you are looking at content classified under America, it's quite possible that there is relevance in the information classified under North America.
Descendents In the example below, Canada is a descendent of North America. Descendents have value because content classified under a child or descendent node may have greater specificity. This content provides consumers of your delivery framework (users of your website or application) with a means of realizing granularity in the data. A website visitor reading about your corporation's news and events relevant to North America may appreciate that there is new specific to Canada or California.

Figure 5: Nested sets within a geography taxonomy (Click to enlarge)

Nested Sets Nested sets are the union of data classified under a particular node and all of its descendents. The example, above, shows three sets. One of those sets is North America and all the nodes within it. Content classified under nodes within a single set may have relevance because they are related by something inherent in the structure (they're all part of the same ancestor). If you hope to convey, in a single "view" or page, contact information for all of your offices in North America, then you want contact information classified under North America and all its descendent nodes (i.e., within the North America nested set). If the user wants to see a summary list of contact information within a particular region, they navigate to a node of greater specificity (a descendent).
Peers Peer nodes exist within a nested set of the hierarchy. Peers have equal depth within a particular nested set of the hierarchy. Borrowing from the example hierarchy, below, if a user has navigated to content classified under Educational Toys (at depth 2 of the Toys nested set), you may find it valuable to provide them visibility into Plush Toys. These topics share context (Toys), but by providing visibility into peer nodes, you convey the breadth of your toy inventory.

Figure 6: Set and depth in a product taxonomy (Click to enlarge)

So what's the point? If you're building algorithms that map content, from your database or content management system to a delivery framework (website or application), it may be both necessary and beneficial to consider content by these relationships. For example, let's say the assets in your content management application are organized by type (press release, white paper, image, etc.) and corporate division; it's an easy and logical choice that allows you to manage security in the content management application at the department level, but it doesn't accurately reflect the organization of your internet site. Additional Guidance

Avoid burdening your content contributors with the maintenance of unnecessary metadata; define only as much metadata as you require for the effective presentation of your assets.
Whenever possible, leverage the information you already have. Identify, and aim to leverage, implicit relationships in the content classification criteria you already collect.
If you’re building your own content classification software, make it extensible – it shouldn’t be necessary for you to revisit the application whenever classification needs vary.

It's your task to map content from the source (content management application) to the presentation system (internet delivery framework). If several parts of your organization are capable of producing press releases, and these press releases are stored in department-specific areas of the content management application, how do you get them on the site? Example "algorithms" for populating the press release and investor news sections of your internet site may read like this: "All press release type assets, in a "ready" state (leveraging metadata), except those in the investor relations nested set are queried, sorted in descending order by publish date and rendered in the Press Releases section." Additionally, "All press release type assets, in a "ready" state, in the investor relations nested set are queried, sorted in descending order by publish date and rendered under Investor Relations > News." By leveraging the more complex relationships available in hierarchies, we facilitate the presentation (contextualization) of content without impacting it's organization. Too frequently, contextualization algorithms neglect implicitly derived information available in hierarchies. Narrowly scoped queries are used to relate content to its place in the delivery framework. It's up to the information architect to find and realize the value of these relationships when developing a content delivery framework. The opportunity for improved usability and greater content visibility using these relationships is tremendous and oftentimes supports the mission of an IA. Technical issues Unfortunately, knowing about these relationships is not enough. Pre-assembled content management and delivery frameworks may not provide the APIs or schema necessary to effectively leverage this knowledge. It's important to understand what you want, and then to find and understand the limitations of your tools. Maintaining and consuming hierarchical data in tabular relational database management systems (RDBMS) systems can be difficult, but with creative database administration, the relationships can be assembled in temporary tables and/or discovered by stored procedures. This is valuable stuff. If it improves content delivery, consider the cost of developing a method for accessing these relationships if it doesn't already exist. Summary When developing your content classification strategies (for both content management and presentation), make a point of evaluating your hierarchical structures for relationships and valuable data hidden within them. Relationships between nodes of a hierarchy are complex, but often mirror the perception of your users. By leveraging these relationships, you can broaden the scope of your queries. Broaden the scope of data presented to the users, and drive users toward tangential content.

For More Information:

Boiko, Bob. Content Management Bible. John Wiley & Sons, 2001.
Aifia Tools http://aifia.org/tools/
Series on controlled vocabularies and faceted classification by Karl Fast, Fred Leise and Mike Steckel

Christian Ricci is a consultant, application developer, web designer, and project manager with over 11 years of experience in software design and development, network and server administration, and software project management and engineering. As a Senior Solutions Architect for Saillant Consulting Group, Chris has lead portal, content, and document management projects for Qualcomm, Intermountain Health Care, J.D. Edwards, EAS, and the Denver Post. Site: http://www.chiamonkey.com/ Resume: http://www.chiamonkey.com/mealticket