Chapter 2 described many of the types of stakeholders that a project might have. As shown in Figure 6-1, a user class is a subset of the product’s users, which is a subset of the product’s customers, which is a subset of its stakeholders. An individual can belong to multiple user classes. For example, an application’s administrator might also interact with it as an ordinary user at times. A product’s users might differ—among other ways—in the following respects, and you can group users into a number of distinct user classes based on these sorts of differences:

Figure 6-1. A hierarchy of stakeholders, customers, users, and user classes.

It’s tempting to group users into classes based on their geographical location or the kind of company they work in. One company that creates software used in the banking industry initially considered distinguishing users based on whether they worked in a large commercial bank, a small commercial bank, a savings and loan institution, or a credit union. These distinctions really represent different market segments, though, not different user classes.

A better way to identify user classes is to think about the tasks that various users will perform with the system. All of those types of financial institutions will have tellers, employees who process loan applications, business bankers, and so forth. The individuals who perform such activities—whether they are job titles or simply roles—will have similar functional needs for the system across all of the financial institutions. Tellers all have to do more or less the same things, business bankers do more or less the same things, and so on. More logical user class names for a banking system therefore might include teller, loan officer, business banker, and branch manager. You might discover additional user classes by thinking of possible use cases, user stories, and process flows and who might perform them.

Certain user classes could be more important than others for a specific project. Favored user classes are those whose satisfaction is most closely aligned with achieving the project’s business objectives. When resolving conflicts between requirements from different user classes or making priority decisions, favored user classes receive preferential treatment. This doesn’t mean that the customers who are paying for the system (who might not be users at all) or those who have the most political clout should necessarily be favored. It’s a matter of alignment with the business objectives.

Disfavored user classes are groups who aren’t supposed to use the product for legal, security, or safety reasons ([ref081]). You might build in features to deliberately make it hard for disfavored users to do things they aren’t supposed to do. Examples include access security mechanisms, user privilege levels, antimalware features (for non-human users), and usage logging. Locking a user’s account after four unsuccessful login attempts protects against access by the disfavored user class of “user impersonators,” albeit at the risk of inconveniencing forgetful legitimate users. If my bank doesn’t recognize the computer I’m using, it sends me an email message with a one-time access code I have to enter before I can log on. This feature was implemented because of the disfavored user class of “people who might have stolen my banking information.”

You might elect to ignore still other user classes. Yes, they will use the product, but you don’t specifically build it to suit them. If there are any other groups of users that are neither favored, disfavored, nor ignored, they are of equal importance in defining the product’s requirements.

Each user class will have its own set of requirements for the tasks that members of the class must perform. There could be some overlap between the needs of different user classes. Tellers, business bankers, and loan officers all might have to check a bank customer’s account balance, for instance. Different user classes also could have different quality expectations, such as usability, that will drive user interface design choices. New or occasional users are concerned with how easy the system is to learn. Such users like menus, graphical user interfaces, uncluttered screen displays, wizards, and help screens. As users gain experience with the system, they become more interested in efficiency. They now value keyboard shortcuts, customization options, toolbars, and scripting facilities.

User classes need not be human beings. They could be software agents performing a service on behalf of a human user, such as bots. Software agents can scan networks for information about goods and services, assemble custom news feeds, process your incoming email, monitor physical systems and networks for problems or intrusions, or perform data mining. Internet agents that probe websites for vulnerabilities or to generate spam are a type of disfavored non-human user class. If you identify these sorts of disfavored user classes, you might specify certain requirements not to meet their needs but rather to thwart them. For instance, website tools such as CAPTCHA that validate whether a user is a human being attempt to block such disruptive access by “users” you want to keep out.

Remember, users are a subset of customers, which are a subset of stakeholders. You’ll need to consider a much broader range of potential sources of requirements than just direct and indirect user classes. For instance, even though the development team members aren’t end users of the system they’re building, you need their input on internal quality attributes such as efficiency, modifiability, portability, and reusability, as described in Chapter 14. One company found that every installation of their product was an expensive nightmare until they introduced an “installer” user class so they could focus on requirements such as the development of a customization architecture for their product. Look well beyond the obvious end users when you’re trying to identify stakeholders whose requirements input is necessary.

Identify and characterize the different user classes for your product early in the project so you can elicit requirements from representatives of each important class. A useful technique for this is a collaboration pattern developed by Ellen Gottesdiener called “expand then contract” ([ref093]). Start by asking the project sponsor who he expects to use the system. Then brainstorm as many user classes as you can think of. Don’t get nervous if there are dozens at this stage; you’ll condense and categorize them later. It’s important not to overlook a user class, which can lead to problems later when someone complains that the delivered solution doesn’t meet her needs. Next, look for groups with similar needs that you can either combine or treat as a major user class with several subclasses. Try to pare the list down to about 15 or fewer distinct user classes.

One company that developed a specialized product for about 65 corporate customers initially regarded each company as a distinct user with unique needs. Grouping their customers into just six user classes greatly simplified their requirements challenges. [ref082] offer much advice about casting a wide net to identify potential users, pruning the user list, and seeking specific users to participate in the project.

Various analysis models can help you identify user classes. The external entities shown outside your system on a context diagram (see Chapter 5) are candidates for user classes. A corporate organization chart can also help you discover potential users and other stakeholders ([ref013]). Figure 6-2 illustrates a portion of the organization chart for Contoso Pharmaceuticals. Nearly all of the potential users for the system are likely to be found somewhere in this chart. While performing stakeholder and user analysis, study the organization chart to look for:

Figure 6-2. A portion of the organization chart for Contoso Pharmaceuticals.

Organization chart analysis reduces the likelihood that you will overlook an important class of users within that organization. It shows you where to seek potential representatives for specific user classes, as well as helping determine who the key requirements decision makers might be. You might find multiple user classes with diverse needs within a single department. Conversely, recognizing the same user class in multiple departments can simplify requirements elicitation. Studying the organization chart helps you judge how many user representatives you’ll need to work with to feel confident that you thoroughly understand the broad user community’s needs. Also try to understand what type of information the users from each department might supply based on their role in the organization and their department’s perspective on the project.

Document the user classes and their characteristics, responsibilities, and physical locations in the software requirements specification (SRS) or in a requirements plan for your project. Check that information against any information you might already have about stakeholder profiles in the vision and scope document to avoid conflicts and duplication. Include all pertinent information you have about each user class, such as its relative or absolute size and which classes are favored. This will help the team prioritize change requests and conduct impact assessments later on. Estimates of the volume and type of system transactions help the testers develop a usage profile for the system so that they can plan their verification activities. The project manager and business analyst of the Chemical Tracking System discussed in earlier chapters identified the user classes and characteristics shown in Table 6-1.

Consider building a catalog of user classes that recur across multiple applications. Defining user classes at the enterprise level lets you reuse those user class descriptions in future projects. The next system you build might serve the needs of some new user classes, but it probably will also be used by user classes from your earlier systems. If you do include the user-class descriptions in the project’s SRS, you can incorporate entries from the reusable user-class catalog by reference and just write descriptions of any new groups that are specific to that application.

When developing commercial software, it can be difficult to find product champions from outside your company. Companies that develop commercial products sometimes rely on internal subject matter experts or outside consultants to serve as surrogates for actual users, who might be unknown or difficult to engage. If you have a close working relationship with some major corporate customers, they might welcome the opportunity to participate in requirements elicitation. You might give external product champions economic incentives for their participation. Consider offering them discounts on the product or paying for the time they spend working with you on requirements. You still face the challenge of how to avoid hearing only the champions’ requirements and overlooking the needs of other stakeholders. If you have a diverse customer base, first identify core requirements that are common to all customers. Then define additional requirements that are specific to individual corporate customers, market segments, or user classes.

Another alternative is to hire a suitable product champion who has the right background. One company that developed a retail point-of-sale and back-office system for a particular industry hired three store managers to serve as full-time product champions. As another example, my longtime family doctor, Art, left his medical practice to become the voice-of-the-physician at a medical software company. Art’s new employer believed that it was worth the expense to hire a doctor to help the company build software that other doctors would accept. A third company hired several former employees from one of their major customers. These people provided valuable domain expertise as well as insight into the politics of the customer organization. To illustrate an alternative engagement model, one company had several corporate customers that used their invoicing systems extensively. Rather than bringing in product champions from the customers, the developing company sent BAs to the customer sites. Customers willingly dedicated some of their staff time to helping the BAs get the right requirements for the new invoicing system.

Anytime the product champion is a former or simulated user, watch out for disconnects between the champion’s perceptions and the current needs of real users. Some domains change rapidly, whereas others are more stable. Regardless, if people aren’t operating in the role anymore, they simply might have forgotten the intricacies of the daily job. The essential question is whether the product champion, no matter what her background or current job, can accurately represent the needs of today’s real users.

To help the product champions succeed, document what you expect your champions to do. These written expectations can help you build a case for specific individuals to fill this critical role. Table 6-2 identifies some activities that product champions might perform ([ref240]). Not every champion will do all of these; use this table as a starting point to negotiate each champion’s responsibilities.

One person can rarely describe the needs for all users of an application. The Chemical Tracking System had four major user classes, so it needed four product champions selected from the internal user community at Contoso Pharmaceuticals. Figure 6-4 illustrates how the project manager set up a team of BAs and product champions to elicit the right requirements from the right sources. These champions were not assigned full time, but each one spent several hours per week working on the project. Three BAs worked with the four product champions to elicit, analyze, and document their requirements. (One BA worked with two product champions because the Buyer and the Health and Safety Department user classes were small and had few requirements.) One of the BAs assembled all the input into a unified SRS.

Figure 6-4. Product champion model for the Chemical Tracking System.

We didn’t expect a single person to provide all the diverse requirements for the hundreds of chemists at Contoso. Don, the product champion for the Chemist user class, assembled a backup team of five chemists from other parts of the company. They represented subclasses within the broad Chemist user class. This hierarchical approach engaged additional users in requirements development while avoiding the expense of massive workshops or dozens of individual interviews. Don always strove for consensus. However, he willingly made the necessary decisions when agreement wasn’t achieved so the project could move ahead. No backup team was necessary when the user class was small enough or cohesive enough that one individual truly could represent the group’s needs.^[2]

Expect to encounter resistance when you propose the idea of having product champions on your projects. “The users are too busy.” “Management wants to make the decisions.” “They’ll slow us down.” “We can’t afford it.” “They’ll run amok and scope will explode.” “I don’t know what I’m supposed to do as a product champion.” Some users won’t want to cooperate on a project that will make them change how they work or might even threaten their jobs. Managers are sometimes reluctant to delegate authority for requirements to ordinary users.

Separating business requirements from user requirements alleviates some of these discomforts. As an actual user, the product champion makes decisions at the user requirements level within the scope boundaries imposed by the business requirements. The management sponsor retains the authority to make decisions that affect the product vision, project scope, business-related priorities, schedule, or budget. Documenting and negotiating each product champion’s role and responsibilities give candidate champions a comfort level about what they’re being asked to do. Remind management that a product champion is a key contributor who can help the project achieve its business objectives.

If you encounter resistance, point out that insufficient user involvement is a leading cause of software project failure. Remind the protesters of problems they’ve experienced on previous projects that trace back to inadequate user input. Every organization has horror stories of new systems that didn’t satisfy user needs or failed to meet unstated usability or performance expectations. You can’t afford to rebuild or discard systems that don’t measure up because no one understood the requirements. Product champions provide one way to get that all-important customer input in a timely way, not at the end of the project when customers are disappointed and developers are tired.

The product champion model has succeeded in many environments. It works only when the product champions understand and sign up for their responsibilities, have the authority to make decisions at the user requirements level, and have time available to do the job. Watch out for the following potential problems:

Beware of users who purport to speak for a user class to which they do not belong. Rarely, an individual might actively try to block the BA from working with the ideal contacts for some reason. On the Chemical Tracking System, the product champion for the chemical stockroom staff—herself a former chemist—initially insisted on providing what she thought were the needs of the chemist user class. Unfortunately, her input about current chemist needs wasn’t accurate. It was difficult to convince her that this wasn’t her job, but the BA didn’t let her intimidate him. The project manager lined up a separate product champion for the chemists, who did a great job of collecting, evaluating, and relaying that community’s requirements.