a user-centered design approach to personalization

5
44 August 2000/Vol. 43, No. 8 COMMUNICATIONS OF THE ACM

Upload: john

Post on 19-Dec-2016

215 views

Category:

Documents


1 download

TRANSCRIPT

Page 1: A user-centered design approach to personalization

44 August 2000/Vol. 43, No. 8 COMMUNICATIONS OF THE ACM

Page 2: A user-centered design approach to personalization

JEA

N-F

RA

NC

OIS

PO

DEV

IN

COMMUNICATIONS OF THE ACM August 2000/Vol. 43, No. 8 45

As with any other disorganized assortment oftools—especially fascinating new tools—designersare often drawn into the trap of trying to find usesfor the tools, and deploying the coolest new features,forgetting their primary focus should be on provid-ing value to the end user. So we often hear peopleask, “Is this recommendation engine (or technologyX) worth its fancy price, or is it just the fad of the

moment? Would personalization feature Y providereal value to my customers? How important is it rel-ative to other capabilities that I could implement?How should I use it, that is, when and where in myoverall design does it fit?”

This article presents a framework for analyzingthe role of personalization in the design of any userexperience. In a nutshell, we advocate using a six-

A USER-CENTEREDDESIGN APPROACH to

Personalization

Joseph Kramer, Sunil Noronha, and John Vergo

Personalization is a toolbox of technologies and application features used in the

design of an end-user experience. Features classified as “personalization” are

wide-ranging, from simple display of the end-user’s name on a Web page, to

complex catalog navigation and product customization based on deep models of users’

needs and behaviors. Similarly, personalization technologies range from commonplace

use of databases, cookies, and dynamic page generation, to esoteric pattern matching and

machine-learning algorithms, rule-based inferencing, and data mining.

The key to successful design is grounding the choice of features and tools upon value to the end user.

The Human Element

Page 3: A user-centered design approach to personalization

step user-centered design (UCD) process we devel-oped over the course of designing object-orientedinterfaces for e-commerce applications. This partic-ipatory design process generates each design throughexplicit elicitation and modeling of users’ goals,beliefs, and behaviors, and ensures the resulting per-sonalized system delivers genuine value to the user.

The first step in the UCD process is determiningthe target user segments of the experience beingdesigned, for example, the “most-valued customers.”This step is often guided by the market researchprocesses that define the total pool of potential cus-tomers, segment the customers based on their wantsand needs, and evaluate each market segment’sattractiveness [2]. However, it has significant techni-cal implications for the personalization capabilitiesthat will have to be supported.

An important implication pertains to automatedidentification of the target user segments: How dowe detect if a given user belongs to a target segment?

For example, marketing may have concluded thatone of the most profitable customer types is the“technically savvy small-business owner who cur-rently owns a home PC.” Even if we knew how todesign the perfect experience for this type of user, wewould not be able to deliver the experience unless wecould identify that a given user fits this profile. Inother words, we would need to design ways to assessthe technical knowledge of the user, their businesssize, and so on. This process forms a set of key per-sonalization requirements. Market research is typi-cally conducted via large, rather intrusive,questionnaires; administering a similar quiz to a useron a Web site is generally impractical. Further, thesediagnostic questions may not be easily ported to dif-ferent media.

An alternative way to deal with this problem is toexchange the technical design challenge for a marketanalysis challenge: Given a set of user profile elementsthat we know we can feasibly acquire, base the usersegmentation analysis solely on this set. At the end ofthis step, all the stakeholders, including the businessand technical teams, should be in agreement as to

who the target users will be, such that their identifi-cation by the system can be practicably automated.

We use standard task analysis methods to learnthe impetus for users’ actions (“triggers”), methodsof completing the tasks (“processes, use cases”), andthe ultimate intention of the user (“goal”) [1]. Goal-decomposition graphs can be used to map high-levelgoals to specific subgoals and to the tasks thatachieve those subgoals. Further detail on the taskflows is captured through activity diagrams. Thisstep must be done with enough detail to clearlyunderstand the different triggers, processes, andgoals employed by different subtypes of users withinthe target user group, since a given trigger can resultin different goals for different users, and a given goalmay be accomplished through different tasks.

The branch points in the goal trees and activityflows are key control points for personalizing theuser experience. For example, the task of “buying anew computer,” for an expert may represent the sub-

goal “Let me build my ideal machine,” which can bedesigned as a detailed configuration. However, anovice may express a subgoal as “Help me under-stand which one of your standard offerings makesthe most sense for me” which can be better designedthrough needs-based recommendations.

Goal and task analysis goes a long way towardshaking out the personalization features and toolsthat are (not) important for a given project becauseeach task can be traced up the goal tree to determinevalue to the end user. We emphasize a holistic view;task analysis produces one big, structured pile ofrequirements. Requirements that comprise tailoringof the task based on some user information are clas-sified under the rubric of personalization, but areevaluated no differently from the others. It may turnout that an anticipated feature (for example, a per-sonalized shopping assistant), does not show up as animportant feature in the pile, perhaps because usersaren’t interested in recommendations for the givenproduct space or it conflicts with other, more impor-tant, requirements (for example, most users may pre-fer to step through a complete product hierarchy,

46 August 2000/Vol. 43, No. 8 COMMUNICATIONS OF THE ACM

DESIGNERS ARE OFTEN DRAWN INTO THE TRAP OF TRYING TO

FIND USES FOR THE TOOLS, AND DEPLOYING THE COOLEST NEW

FEATURES, FORGETTING THEIR PRIMARY FOCUS SHOULDBE ON PROVIDING VALUE TO THE END USER.

Page 4: A user-centered design approach to personalization

reassured they are not being excluded from someparts of the product catalog). Conversely, it may turnout that an ordinary-looking requirement such ascomparing two PCs in an online catalog turns into achallenging new personalization requirement becausethe user wishes to compare against a PC he or she hasat home. Either way, much of the confusion regard-ing the value of different personalization features iscleared away by the end of this step.

Exploring the Blue SkyFollowing the task analysis, it is useful to conjecturewith your target users what would be the ultimatedesirable set of triggers, processes, and goals. Thisstep has been called “blue sky” as it attempts toderive the triggers, goals, and the desired intermedi-ate processes without worrying about any real-worldconstraints [2].

The primary purpose of a blue-sky exercise is toprovoke the creation of novel (and more valuable)task flows. Often, the tasks identified in the task-analysis phase are derived from users’ current prac-tices, which are overly constrained by assumptionsabout what can be implemented. Thus, the task of“shopping for groceries” may lead to a Web sitedesign that mimics a physical grocery store, which isfar from optimal. It takes blue-sky thinking to gen-erate an alternative task flow in which “not shoppingat all” is considered an implementation of this task,that is, having the grocery store predict one’s routinegrocery needs, and manage one’s refrigerator. As inthis example, and most others (for example, supply-ing the perfect clothes, assembling the perfect PC),blue-sky task flows require considerable knowledgeof the user and are likely to provoke exploitation ofnew personalization technologies.

To design for the user, we need to base our designson the user’s mental models of the domain (lan-guage, concepts, beliefs, among others). Therefore,the fourth step in our UCD process is the elicitationand documentation of the user’s object model. Thisstep is achieved by extracting noun phrases and verbsfrom the use cases and defining them with yourusers. Subsequently, the users iterate on the concep-tual relationships that form their mental model ofthe domain. The results of this analysis are capturedin artifacts such as cognitive maps; our preferredtechnique is object diagrams.

Ideally, everyone would speak the same language,and there would be a single model of the applica-tion’s content (object model). However, the fact isthat different users use different object definitionsand have different gaps in their understanding (forexample, different levels of knowledge about the

attributes of objects and their relationships). Design-ers must deal with the differing object definitions byrefining the terminology used during the experience,or by educating the user. For example, we foundsome users define workstations as networkedmachines, while others define them as high-perfor-mance PCs. The preferred way to deal with this is tocome up with more specific product categories,which capture the different capabilities and shades ofmeaning (“networked PC”), ensuring these productnames match most users’ object definitions. Theother gap that can exist is disparity in understandingthat results from population differences (for exam-ple, technically knowledgeable vs. eager novices).This can be mitigated by delivering different levelsand modes of education to different users (personal-ization of help facilities, information detail, mode ofpresentation of information).

On a completely different note, the objectmodel is key to mass customization, whichis the most powerful form of personaliza-

tion. Mass customization is based on assembly ofcomponents, and judicious componentization is keyto managing manufacturing costs. The componentsand assemblies are reflected in the object model asartifacts the user selects and puts together. If there isa gap between what a user expects as a configurablecomponent and what is actually supplied, there is amajor ease-of-use cost (and subsequent economicloss). During the design of the componentizationand the assembly process, designers must elicit themental constructs (a.k.a. object model) most mean-ingful to the user (for example, base product configurations defined via distinctive value proposi-tion statements) and map them to manufacturablesystems.

While the previous steps have focused primarilyon creating models of different facets of the user(goals, tasks flows, mental models), the fifth step ofthe process combines all of them into an end-to-endanalysis of each potential user experience. We stepthrough every flow in each task, determining theinformation (object properties and relationships)that must be accessed to complete each action, anddocumenting it via interaction diagrams. Visualtreatment rules are then applied to design the“views” (renderings of the accessible information ofeach object) the user ultimately sees.

Since different user types have different informa-tion needs, this step results in identification of mul-tiple views, which are used as the basis forpersonalization. Also, the interaction sequences helppinpoint the kind of persistent state information

COMMUNICATIONS OF THE ACM August 2000/Vol. 43, No. 8 47

Page 5: A user-centered design approach to personalization

(the context, for example, the status of the user’stransactions, tasks, and so on) that must be main-tained by the system as the user moves through theexperience. Some of these interaction sequences maybe long-running (spanning weeks, months, orlonger), for example, relating current browsing forPC accessories to last year’s purchase of a desktopmachine, and pose significant challenges to design-ing the interaction. Modeling and managing thestate of the long-term relationship with the user isthe Holy Grail of the one-to-one enterprise [4].

As always, we iterate through several designs,evolving from a low-fidelity prototype to a fully func-tional system, improving through testing and partici-patory design. It is worth noting that in testingsystems incorporating personalization, special atten-tion must be given to privacy concerns. Privacy andtrust are issues that should be dealt with right fromthe beginning of the UCD process, but it has beenour experience that despite one’s best attempts to doso, there are always surprises that come out duringtesting. Innocuous questions about when the userwishes to have a PC available for delivery (motivatedby the designers’ knowledge that users are keenlyinterested in the availability of products) backfirewhen users interpret it as an insidious marketing ployto bombard them with marketing information untilthe date they need the machine. Questions about theuser’s budget, intended to filter the product space torecommend only PCs that fall within the usersdesired price range, trigger a suspicion that this willaffect the recommended products’ prices in otherways. Beyond rewording questions correctly, animportant design technique that must be superim-posed on the use of any personalization feature is toensure that users clearly understand why the questionis being asked, and how it directly fits in with theirgoals. This means the consequences of the user’sactions must be transparent to the user; the modelunderlying the feature should be simple enough toengender trust. During testing it is necessary toexplicitly probe and measure how well users under-stand the benefits, and that the total value is adequateto justify the inclusion of the feature in the system.

ConclusionCurrent methods of software system design do notadequately capture the needs and values of endusers. We have presented a UCD methodology that,at its core, maintains a strong focus on bringingvalue to the end user. From our perspective, person-alization takes a back seat to end-user value. We havegiven examples of how the use of personalizationmight result in decreased value to the user. Adesigner who sets out to personalize an existingapplication, or to build a new personalized applica-tion, is poised to make the classic error of puttingtechnology before the needs of the end users. Per-sonalization is currently best viewed as an evolvingset of tools. As craft-persons, our job is to under-stand the capabilities, limitations, and end-user costsof the tools we brandish. We must make good deci-sions as to when to reach into our personalizationtoolbox. When we do, testing, measuring, and iter-ating on our designs, with an unwavering focus ondelivering value to the end user, will help insure thesuccess of the applications we build.

References1. Collins, D. Designing Object-Oriented User Interfaces. Benjamin–Cum-

mins, Redwood City, Calif., 1995.2. Dayton, T., McFarland, A., and Kramer, J. Bridging user needs to

object-oriented GUI prototype via task object design. In L.E. Wood(Ed.). User Interface Design. (1998) CRC Press, Boca Raton, Fla., 15–56.

3. Kotler, P., and Armstrong, G. Principles of Marketing. Prentice–Hall,Upper Saddle River, N.J., 1999.

4. Peppers, D. and Rogers, M. Enterprise One to One: Tools for Competingin the Age of Interactivity. Currency Doubleday, New York, N.Y. 1997.

Joseph Kramer ([email protected]) is a senior applicationdevelopment analyst; Sunil Noronha ([email protected] isa research staff member; and John Vergo ([email protected]) isa research staff member at IBM T.J. Watson Research Center,Hawthorne, NY.

Permission to make digital or hard copies of all or part of this work for personal orclassroom use is granted without fee provided that copies are not made or distributedfor profit or commercial advantage and that copies bear this notice and the full cita-tion on the first page. To copy otherwise, to republish, to post on servers or to redis-tribute to lists, requires prior specific permission and/or a fee.

©2000 ACM 0002-0782/00/0800 $5.00

c

48 August 2000/Vol. 43, No. 8 COMMUNICATIONS OF THE ACM

PERSONALIZATION IS CURRENTLY BEST VIEWED AS AN EVOLVING SET OF TOOLS. AS CRAFT-PERSONS, OUR

JOB IS TO UNDERSTAND THE CAPABILITIES, LIMITATIONS, AND

END-USER COSTS OF THE TOOLS WE BRANDISH.