Re-Architecting PeopleSoft from the Top Down

Previously, Chiara Fox, senior information architect at PeopleSoft, presented a case study about the remarkable redesign that she and her team accomplished last year. Adaptive Path was fortunate to work with Chiara to develop the architecture. The project was a compelling one: PeopleSoft hoped to consolidate three highly trafficked–and highly redundant–sites into one site that would be served dynamically to a range of user types from a single content management system. It’s the sort of project that many companies talk about and few ever accomplish.

The challenge we faced was this: By what process can a team create a system that is both user-centered and content-centered?Adaptive Path’s role in the project was to lead the information architecture development. Because of the complexity of the problem, success for this project would hinge upon a sound methodology.

PeopleSoft had massive amounts of content stored in hundreds of databases, much of which was duplicative, and they wanted it unified, culled, and put into a single content management system. This was certainly a project that demanded bottom-up information architecture.

Despite the clear necessity for bottom-up IA, alone it would be insufficient to produce a successful architecture. Bottom-up techniques are content-centered, which is to say that they aren’t necessarily user-centered. To be successful, the finished system had to support user needs and business requirements, and also had to accurately represent the content assets that would reside in—and be made accessible through—the architecture. The challenge we faced was this:

  • By what process can a team create a system that is both user-centered and content-centered?

And, to bring the question to its finest point:

  • By what process can a team create a system that simultaneously reflects content patterns, supports user needs, and delivers on important business objectives?

In most development processes, it seems that these design objectives are mutually exclusive–that satisfying business objectives must necessarily impede the user’s ability to accomplish tasks, or that designing up from the content precludes designing down from user’s goals.

For the PeopleSoft re-architecture, we resolved this classic conflict by using a top-down architecture that integrated ethnographic user research, content analysis, and business requirements.

User-centered top-down methodology
For the PeopleSoft redesign, the top-down process had three components. We began by understanding user needs through a series of interviews that resulted in a mental model diagram. We used key deliverables from the bottom-up architecture work, specifically the content map, to begin establishing an organizational structure for the site. Finally, we used business goals, derived through stakeholder interviews and a process called “goal alignment,” to set build priorities. The result was a top-down architecture and navigation structure that was strongly user-centered, supported the content schemes emerging from the bottom-up activities, and could measurably achieve business goals.

User interviews, task analysis, and the mental model
The three PeopleSoft sites spanned a range of user types: four types of prospective customers, current customers, and five types of business partners. The users of the sites included people at all levels of the organization, from C-level executives to technical staff who implement the software. Despite their diversity, we believed that a common thread of activity tied them to the PeopleSoft websites–the cycle of selecting, buying, installing, and upgrading enterprise software.

The strong workflow common among PeopleSoft’s audiences allowed us to use a qualitative research process developed by Indi Young (a partner at Adaptive Path) to create task-based mental models. The methodology draws heavily upon techniques used in ethnographic research, contextual inquiry, and traditional task analysis. The result is a research-based visualization of the user’s mental model.

The first step was to define user types and conduct the interviews. Indi, along with Peter Merholz interviewed 19 people: six potential customers, seven current customers, and six partners. The hour-long interviews were designed to uncover the goals and tasks that each person had encountered when researching, buying, and maintaining enterprise software. During each interview we asked participants to tell us their story–to describe in detail what steps they took to accomplish their goals.

The discussion guide was used as prompts in a conversation, rather than a verbatim script (as you would, for instance, during a usability test). Interviewers encouraged participants to follow relevant tangents and probed for details by asking questions like “how did you…” and “what steps did you take to….”

Once Peter and Indi completed the interviews, they thoroughly dissected the transcripts. Every comment that included a “task” was pulled out and placed in a “task table.” When multiple comments mentioned the same task, those comments were placed together in the table. They grouped similar tasks, and gave the groups names. This analysis process removes the comments from their context in order to reveal patterns of activity across multiple users. By analyzing the interviews for potential customers, current customers, and partners separately, Peter and Indi were able to create three sets of task tables that they would turn into separate mental models.

To get from task table to mental model is a simple matter of grouping related tasks, then grouping the groups. Each task is a little white box, and related tasks are stacked and grouped inside a gray box. The stacks are arranged along a horizon line, and related groups of tasks are separated from one another by vertical rules.

The result looks like this:

In this way, we were able to create a fairly accurate “map” of which activities customers and partners needed to engage in when purchasing or maintaining PeopleSoft products.

Content matching
While Peter and Indi were interviewing users and assembling the mental models, Marcus Haid, our intern, and I were busy doing an inventory of the content on the three PeopleSoft sites. The exhaustive inventory resulted in a spreadsheet with thousands of lines that listed every HTML page and summarized every database on the three PeopleSoft sites. Although this level of detail was an asset to the bottom-up architecture activities that Chiara was leading, for the top-down architecture, it was like trying to understand a newspaper photo by looking at the pattern of dots.

We needed to zoom out, to see an accurate picture of what content was available on the sites, in order to understand how the content aligned with user tasks. This content/task alignment would form the basis for our high-level navigation decisions, so it was essential to get it right. To provide an appropriate view of the content, I created a “content map,” which summarized the site content in the same way that the mental model summarized the 170 pages of verbatim task tables. Using a combination of subject and document type, I drew general conclusions about the kinds of content assets available. In the end I was able to represent the complete content picture in about 50 objects.

The content map used the same “little-white-box” format that we had used to develop the mental model, which enabled us to merge the two diagrams. In collaborative working sessions that included PeopleSoft and Adaptive Path, we matched every content asset to tasks on each of the three mental models we had developed (one each for prospective customers, current customers, and partners).

Deriving high-level architecture and navigation
This user task/content comparison validated a few of our assumptions: First, users have a strong task orientation when working with a company like PeopleSoft. Second, PeopleSoft currently had (or was planning to add) ample content assets to support most of their users’ tasks and goals. Based on the strength of these findings, we agreed that a task-based navigation system would work well for PeopleSoft.

By comparing the three mental models (and the corresponding content) to one another, we could also see that there was substantial overlap in both the tasks to be accomplished and the content available. This validated our belief that best solution was not three separate sites, but rather a single site that would dynamically change when current customers and partners logged in. We knew that the dynamic changes would involve expanding and collapsing navigation to provide (or restrict) access to proprietary content, but it took several collaborative working sessions to define exactly how those would work.

To derive the navigation, we started by skimming the top off the mental model: We lined up the names of the mental spaces for each of the three audiences, eliminated redundancies and identified which parts would be available to current customers and partners only. This was our first draft of a top-level navigation. Next we filled in the local navigation using the task group names from the mental models, again removing duplication and identifying areas that would be private. Finally, we added the content assets within the local navigation, according to where we had placed them on the mental model diagram.

The end result of the derivation was a first-draft architecture diagram and navigation scheme. It showed us a single website with dynamic navigation that would expand and collapse to serve the specific needs of each user type. The architecture and navigation were refined and adjusted, but stayed true to the task-based organization that was developed in the mental model.

Business objectives and the final architecture
At this point in the development process, we had successfully developed an architecture scheme that satisfied user needs and supported the content patterns that were discovered during bottom-up architecture. But we knew that the architecture we had planned was still too blue-sky to be achievable in a single launch. We had to establish build priorities, and we wanted to use business objectives to drive the prioritization. This meant that the senior management of PeopleSoft had to define an achievable set of objectives through a process that they call “goal alignment.”

We started the goal-alignment process early in the project by interviewing 24 stakeholders, including senior executives. These interviews provided us with a long list of objectives. The next challenge was to gain broad support for a small set of objectives that could realistically be satisfied by the new architecture. Interview findings were analyzed and points of convergence and divergence across the organization were identified. A working session with senior PeopleSoft decision-makers was facilitated in order to achieve goal alignment and together they developed a prioritized list of measurable, achievable business objectives.

These objectives, combined with input from the build team about difficulty and available resources, made the implementation phases for our architecture plan very clear. With that in place we were able to quickly finish a set of architecture diagrams that described the site as it would be finally implemented.


Collaborative development
Possibly the most innovative aspect of the PeopleSoft redesign was not the techniques we used, but how we chose to employ them. At every step, vendor and client worked collaboratively as a team, revising and editing documentation in real time. This enabled us to advance the thinking and mature the design decisions before locking into a final solution. The top-down architecture was finished in August, and the site launched in December, after 10 months of work. By all measures, the launch was a success. PeopleSoft finished on schedule and under budget. Their customers and partners had a consistent experience across all of the sites. And in January, PeopleSoft’s website provided the company with a record number of leads.

Janice Fraser is a partner in Adaptive Path, a user experience design firm. She is on the faculty of San Francisco State University’s Multimedia Studies Program.

Posted in Case Studies, Process and Methods | 3 Comments »

3 Comments

  • Adele Framer

    August 2, 2007 at 6:44 pm

    As of today, sadly, images are broken. The article really needs them.

  • VJ Beauchamp

    September 20, 2007 at 5:40 pm

    Would it be possible to enable the images?

  • Kara Madden

    May 7, 2008 at 12:38 am

    Any way to restore the images to this article?

Sorry, comments are closed.