Content uploaded by James R. Lewis
Author content
All content in this area was uploaded by James R. Lewis on Nov 09, 2017
Content may be subject to copyright.
HUMAN FACTORS, 1987,29(4),461-464
Slot versus Insertion Magnetic Stripe
Readers: User Performance and Preference
JAMES R. LEWIS,1
IBM Corporation, Boca Raton, Florida
Four magnetic stripe readers were evaluated for percentage of good reads, throughput, and
user preference using a simulated time-and-attendance task. Two were slot readers, one with
a vertical slot and one with a horizontal slot. Two were insertion readers, one with the
insertion parallel to the horizontal plane and one with the insertion 20 deg below the hori-
zontal plane. Twelve participants were divided into four groups of three members each. Each
group approached each reader from both directions. The percentage of good reads was high
on all readers and was not significantly different. The differences in throughput and prefer-
ence were significant and favored the slot readers. It did not seem to' matter whether the slot
was vertical or horizontal. The angle of entry for the insertion readers produced no signifi-
cant differences.
INTRODUCTION
There are two basic styles of magnetic
stripe readers in current use: slot and inser-
tion.
A
slot reader is designed to allow the
user to pass a coded magnetic stripe (affixed
to a card or badge) in front of a magnetic
sensing head while guided by a slot. With an
insertion reader, the badge is inserted into
and then withdrawn from an opening a little
larger than the cross-sectional dimensions of
the badge.
Because the insertion reader controls more
degrees of freedom of the movement of the
badge, it is possible that insertion readers
have a higher percentage of good reads,
where a good read is defined as the correct
sensing of the information on the magnetic
stripe. On the other hand, this additional
constraint on movement may have adverse
I
Requests for reprints should be sent to James R. Lewis,
IBM Internal Zip: 3401, Human Factors Department, 1501
NW 51st Street, Boca Raton, FL 33432.
consequences for throughput, or the rate of
good reads per minute.
Another issue is that of horizontal versus
vertical orientation of the slot in a reader. It
was hypothesized that for a queue of users
(like that in a time-and-attendance task) the
throughput for a horizontal reader would be
superior to that for a vertical reader, since
using the horizontal slot seems to be more
compatible with the motion of a user walking
past the reader while clocking into work.
The angle of entry into an insertion reader
is usually parallel to the horizontal plane.
Because lowering the angle of entry into the
reader (i.e., tilting the point of entry below
the horizontal plane) may help to prevent
contaminants from falling into the opening,
the angle of insertion of the badge into the
insertion reader was studied. The advantage
of reducing contamination could be offset by
a disadvantage in user performance.
For some applications (e.g., banking tasks)
it may be more important to emphasize accu-
© 1987, The Human Factors Society, Inc. All rights reserved.
at HFES-Human Factors and Ergonomics Society on December 21, 2015hfs.sagepub.comDownloaded from
462-August 1987
racy, and for other applications (e.g., time-
and-attendance tasks) it may be more impor-
tant to emphasize throughput. The purpose
of this study was to investigate the trade-oUs
in user performance and consistencies in user
preference regarding the two slot and the two
insertion magnetic stripe readers for a simu-
lated time-and-attendance task.
METHOD
Subjects
Ten employees of a temporary-help agency
(six men and four women) and two IBM em-
ployees (both men) participated. The partici-
pants represented a range of age, handed-
ness, education, employment experience, and
badge-reader experience.
Materials and Apparatus
HUMAN
L----
7}
- -.---.---11.
1
J
~I ••. --- - ~~zontal slot
Vertical slot
FACTORS
/1
/
I
.
,
L'
r"/l
~ 1
j
c
1
Four magnetic stripe readers were con-
structed for use in this study (see Figure 1):
(1) a horizontal slot reader; (2) a vertical slot
reader; (3) an insertion reader with the inser-
tion level with the horizontal plane; and (4)
an insertion reader with the insertion angled
20 deg below the horizontal plane.
An LED display was attached to the
readers to monitor the number of good and
bad reads. Performance data were collected
at the reader and transferred to diskette
storage on an IBM Portable Personal Com-
puter.
The Mag Tek Model 2105-002 magnetic
stripe reader was used in the slot readers,
and the American Magnetic Model 100 mag-
netic stripe reader was used in the insertion
readers. All of the readers were presented
under the same conditions, with the mag-
netic heads at a height of 107 em.
Procedure
Instructions were read to the participants,
who were then randomly divided into four
groups, with three members per group. Each
Angled insertion
Figure 1.
The four badge readers.
participant was given a badge with his or her
participant number coded on the stripe so
that performance could be individually
tracked. The groups were assigned to readers
using a Latin square designed to control for
first-order transfer effects (Bradley, 1958). In
other words, each participant used each
reader, and each group of participants used
the readers in a different order so that each
condition and its immediately preceding
condition were counterbalanced.
Groups were taken into the lab two at a
time and were given brief instructions on
how to hold the badge for the reader they
were to use. No specific instructions were
at HFES-Human Factors and Ergonomics Society on December 21, 2015hfs.sagepub.comDownloaded from
SLOT VERSUS INSERTION READERS
given regarding speed versus accuracy. Each
group used a reader for four trials before
using another reader. The duration of a trial
was five minutes on the first reader and three
minutes on the remaining readers. The
members of a group walked in a circle in
front of the reader to simulate a queue of
people waiting to use the reader. On the first
and third trials, the participants approached
the reader in counterclockwise movement,
and on the second and fourth trials they ap-
proached the reader in a clockwise move-
ment. The groups were given a rest period be-
tween each trial while the data collected
during that trial were transferred from the
readers to a portable personal computer.
These data were used to compute good-
read percentages and throughput rates. The
badge readers could detect and count good
and bad reads, and an observer was stationed
at each reader to keep track of null reads. A
read was bad if the readet" could detect the
magnetic stripe but could not read the infor-
mation. A null read occurred if the badge
reader could not detect the magnetic stripe
and therefore could not record the attempt.
In practice, the usual consequence of a bad or
null read is for the user to try again. In this
experiment, the participant did not try again.
After using all four readers, the partici-
pants were asked to indicate their preference
for the readers by ranking them from 1 to 4,
TABLE I
August 1987-463
with 1 for the most preferred and 4 for the
least preferred.
RESULTS
The averages for the three dependent vari-
ables are shown in Table 1. The throughput
averages are based on the performance of the
four groups, since individuals within a group
were not independent of one another. An
analysis of variance showed the differences to
be statistically significant,
F(3,9)
=
26.7,
P
<
0.0001. The percentages of good reads are
based on all 12 participants; the differences
are not statistically significant,
F(3,33)
=
2.4,
p
>
0.05. The preference data were also based
on all 12 participants, and were found to be
significantly different using a Friedman test,
S
=
510.5,
p
<
0.001. A Friedman test was
used because the preference data were ranks.
Table 2 shows the results of multiple-com-
parisons tests conducted on the significant
dependent variables of throughput and pref-
erence. Throughput was analyzed with Bon-
ferroni
t
tests, and preference was analyzed
using a test based on the Friedman statistic
(Hollander and Wolfe, 1973).
From these results, one would conclude
that there is no evidence for a throughput or
preference difference between the two slot
readers or between the two insertion readers,
but there is strong evidence that both the slot
Means (and Standard Deviations) for Three Dependent Measures
Reader
Horizontal slot
Vertical slot
Average data from slot
Level insertion
Angled insertion
Average data from insertion
Throughput
(Good Reads/Min)
27.50 (6.60)
26.13 (6.95)
26.81 (6.74)
18.38 (4.91)
18.56 (5.27)
18.47 (5.06)
Percentage
Good Reads
97.5 (3.6)
98.2 (2.4)
97.8 (2.8)
98.0 (2.2)
99.7 (0.7)
98.8 (1.29)
Preference
1.67 (0.65)
1.46 (0.58)
1.56 (0.16)
3.38 (0.57)
3.50 (0.67)
3.44 (0.16)
Note: For preference. 1
c
most preferred and 4
c
least preferred.
at HFES-Human Factors and Ergonomics Society on December 21, 2015hfs.sagepub.comDownloaded from
464-August 1987
TABLE 2
HUMAN FACTORS
DISCUSSION
Multiple Comparisons for Throughput and Prefer-
ence
Notes: HS
=
Horizontal Slot
VS """Vertical Slot
LI
=
Level Insertion
AI
=
Angled Insertion
'p
<
0.01
•• p
<
0.001
A larger throughput is better, but a smaller preference is better.
readers had a higher throughput than, and
were preferred over, either insertion reader.
The difference between the throughput
averages for the slot and insertion readers is
8.34, whereas the difference (statistically in-
significant) between the percentage of good
reads is 1.01. Comparing insertion with slot
readers, it seems that, with no significant dif-
ference in the percentage of good reads, there
is a gain of 45% in the throughput rate.
Throughput
(Good Reads/Min)
HS
>
VS
HS> U"
HS
>
AI'·
VS> U·'
VS
>
AI"
U
<
AI
Preference
HS
>
VS
HS
<
U'
HS
<
AI"
VS
<
U'
VS
<
AI"
L!
<
AI
For this simulated time-and-attendance
task, the slot readers were superior to the
insertion readers. The differences in
throughput and preference were statistically
significant, and favored the slot readers. Al-
though the differences in the percentage of
good reads appeared to favor the insertion
readers, these differences were not statisti-
cally significant. There was a large gain in
throughput for a nonsignificant difference in
the percentage of good reads when com-
paring slot with insertion readers.
It did not seem to matter whether the slot
of the slot reader was vertical or horizontal.
Nor did it seem to matter whether the
opening for the insertion reader was level or
20 deg below the horizontal plane.
REFERENCES
Bradley,
J.
V. (1958). Complete counterbalancing of imme-
diate sequential effects in a Latin square design.
Journal of the American Statistical Association, 53,
525-528.
Hollander. M .. and Wolfe, D. A. (1973). Nonparametric sta-
tistical methods. New York: Wiley.
at HFES-Human Factors and Ergonomics Society on December 21, 2015hfs.sagepub.comDownloaded from