Outliers in usability testing:
How to treat usability problems found for only
one test participant
      Asbjørn Følstad, SINTEF
      Effie Lai-Chong Law, University of Leicester
      Kasper Hornbæk, University of Copenhagen

      NordiCHI 2012
Content

1   Single-user problems
2   Yes, they are abundant
3   … but how to deal with them?
4   Current practices – straight from the horse's mouth
5   Recommendations




                                                     2
Single-user problems


                       The problems backed up
                       with data from only a
                       single participant in
                       usability test




                                                3
Single-user problems

                       Are they relevant?
                       May be infrequent usability
                       problems
                       - Point estimate: .25 (LaPlace)
                       - 95% conf. int.: .01-.58 (Adj. Wald)


                       Are they valid?
                       May be an artefact of the test
                       situation
                       "there is always a risk of being misled
                       by the spurious behavior of a single
                       person" (Nielsen, 2000 – useit.com)

                                                         4
Single-user problems are abundant

                                             Office system eval.
                Content
                                                15 participants                         Nielsen and
              management                                                                  Landauer
              system eval.                        77 of 145                                  (1993)
                                               problems single-
             17 participants
                                                user problems
            41 of 88 problems
Law and        single-user
Hvannberg       problems
(2004)

                                Law, E.L.-C. Hvannberg, E.T. Analysis of Combinatorial User Effect in
                                International Usability Tests. In Proc. CHI '04, ACM Press (2004), 9-16.

                                Nielsen, J., Landauer, T.K. A mathematical model of the finding of usability
                                problems. In Proc. CHI '93, ACM Press (1993), 206-213.                 5
Advice on how to deal with them is scarce
                               View "unique
  Short discussion of          problems as noise                          Kjeldskov, Skov
  single-user problems.        rather than real                           and Stage
  Recommend to report          usability problems"                        (2004)
  these as outliers            in study of instant
                               data analysis

                               Report single user
                               problems as real
                               problems in a stress-
                               test of problem                            Woolrych
                               predictions                                and Cockton
                                                                          (2001)

                          Kjeldskov J., Skov M. B., Stage J. Instant Data Analysis: Evaluating
                          Usability in a Day. In Proc. NordiCHI '04, ACM Press (2004), 233-240.

                          Woolrych, A., Cockton, G. Why and when five test users aren’t enough.
                          In Proc. IHM-HCI 2001, Cépadèus Éditions (2001), 105-108.      6
Asking the practitioners for current practices
Opportunity: Larger survey on
analysis practices in usability
evaluation

Included question on single-user
problems

89 usability practitioners answered
this particular question

Median 6 yrs. work experience

17 different countries

Usability tests with median of 8
user participants

                                                 7
Potential outcomes for single-user problems


                                 8 accept

Participants as divided as the
                                 4 classify as low priority
little advice provided in the
literature                       4 record as outlier

                                 6 reject

                                 (22 items total on this theme)




                                                                  8
Relevant conditions when making the call
                             18 Problem severity

                             9 Test participants' profile

A range of conditions        6 Sample size
reported as relevant. But
some maybe deserving to be
reported more often?         6 Artifact of the test situation?

                             5 Task importance

                             5 Other

                             (49 items total on this theme)

                                                              9
Resources and strategies when making the call
                             9 Discuss with experts or team
                             members
20 reported to rely on own   9 New/extended evaluations
professional knowledge and   8 Check against heuristics /
experience.
                             guidelines / principles
However, several potential
useful resources and         […]
strategies were reported
                             3 Specific process or policy
                             2 Confirmed hypotheses /
                             previous experiences
                             2 Debrief with users
                             (63 items total on this theme)
                                                              10
When considering the study findings …


  How did you handle
  single-user problems
  in your latest
  usability test?




                                        11
Recommendations

1   Procedure for handling single-user problems
2   Pay particular attention to sample size

3
    Check against knowledge resources –
    guidelines, heuristics, or previous evaluations
4
    Seek advice - from experts or team members
5   Be alert: Artefact of the test situation?



                                                      12
Thank you




            13

Single-user problems

  • 1.
    Outliers in usabilitytesting: How to treat usability problems found for only one test participant Asbjørn Følstad, SINTEF Effie Lai-Chong Law, University of Leicester Kasper Hornbæk, University of Copenhagen NordiCHI 2012
  • 2.
    Content 1 Single-user problems 2 Yes, they are abundant 3 … but how to deal with them? 4 Current practices – straight from the horse's mouth 5 Recommendations 2
  • 3.
    Single-user problems The problems backed up with data from only a single participant in usability test 3
  • 4.
    Single-user problems Are they relevant? May be infrequent usability problems - Point estimate: .25 (LaPlace) - 95% conf. int.: .01-.58 (Adj. Wald) Are they valid? May be an artefact of the test situation "there is always a risk of being misled by the spurious behavior of a single person" (Nielsen, 2000 – useit.com) 4
  • 5.
    Single-user problems areabundant Office system eval. Content 15 participants Nielsen and management Landauer system eval. 77 of 145 (1993) problems single- 17 participants user problems 41 of 88 problems Law and single-user Hvannberg problems (2004) Law, E.L.-C. Hvannberg, E.T. Analysis of Combinatorial User Effect in International Usability Tests. In Proc. CHI '04, ACM Press (2004), 9-16. Nielsen, J., Landauer, T.K. A mathematical model of the finding of usability problems. In Proc. CHI '93, ACM Press (1993), 206-213. 5
  • 6.
    Advice on howto deal with them is scarce View "unique Short discussion of problems as noise Kjeldskov, Skov single-user problems. rather than real and Stage Recommend to report usability problems" (2004) these as outliers in study of instant data analysis Report single user problems as real problems in a stress- test of problem Woolrych predictions and Cockton (2001) Kjeldskov J., Skov M. B., Stage J. Instant Data Analysis: Evaluating Usability in a Day. In Proc. NordiCHI '04, ACM Press (2004), 233-240. Woolrych, A., Cockton, G. Why and when five test users aren’t enough. In Proc. IHM-HCI 2001, Cépadèus Éditions (2001), 105-108. 6
  • 7.
    Asking the practitionersfor current practices Opportunity: Larger survey on analysis practices in usability evaluation Included question on single-user problems 89 usability practitioners answered this particular question Median 6 yrs. work experience 17 different countries Usability tests with median of 8 user participants 7
  • 8.
    Potential outcomes forsingle-user problems 8 accept Participants as divided as the 4 classify as low priority little advice provided in the literature 4 record as outlier 6 reject (22 items total on this theme) 8
  • 9.
    Relevant conditions whenmaking the call 18 Problem severity 9 Test participants' profile A range of conditions 6 Sample size reported as relevant. But some maybe deserving to be reported more often? 6 Artifact of the test situation? 5 Task importance 5 Other (49 items total on this theme) 9
  • 10.
    Resources and strategieswhen making the call 9 Discuss with experts or team members 20 reported to rely on own 9 New/extended evaluations professional knowledge and 8 Check against heuristics / experience. guidelines / principles However, several potential useful resources and […] strategies were reported 3 Specific process or policy 2 Confirmed hypotheses / previous experiences 2 Debrief with users (63 items total on this theme) 10
  • 11.
    When considering thestudy findings … How did you handle single-user problems in your latest usability test? 11
  • 12.
    Recommendations 1 Procedure for handling single-user problems 2 Pay particular attention to sample size 3 Check against knowledge resources – guidelines, heuristics, or previous evaluations 4 Seek advice - from experts or team members 5 Be alert: Artefact of the test situation? 12
  • 13.