Decision making of vending machine users
|
Update: October 2017.
Then at Netherlands Railways, The Netherlands Contact.
|
|
Abstract |
This article discusses two solutions to decision of vending machines: 1) the one button to press system (the pressing of one button only); 2) the several buttons compound decision (the pressing of several buttons).
|
The basis for the discussion is a field evaluation of a train ticket vending machine (TVM) that can sell 800 different types of tickets and can accept all kinds of payment. For this evaluation several hundred TVM users and ticket window users were observed and interviewed. Special attention was paid to the errors which were made.
|
Introduction |
From an ergonomics point of view it is very easy to design a bubble gum vending machine. There is only one product and there is only one coin to be inserted. The Figure at the right shows a one product one coin machine for vending platform tickets. |
However, society is becoming more and more complex and so are vending machines. Hence, when this article was published on paper in 1988, it was expected that new problems will emerge when people have to use these machines when screen interface technology is available. | The introduction of the Dutch Public Transport Card in 2005 made clear that there were more problems than anticipated in 1988
and after tomorrow when
Back to the past of 1988 for the lesson: userfriendly complex systems are possible using applied cognitive psychology. |
 |
Fig. 1 A one-coin product
vending machine for vending platform tickets
|
|
1. The ticket vending machine evaluated |
The ticket vending machine (TVM) sells tickets for 80 destinations and of several fare types (first/second class; single/return; whole day/evening only; return full/reduced price; etc). The machine accepts coinage and paper currency and in future it will also be possible to use credit cards. |
The steps to be taken by the purchaser are described in the written instructions in the upper left hand corner of the machine front. Numbered arrows indicate the controls associated with each step and these arrows start flashing when that step has to be taken. The first step a passenger has to take is to select between some 80 destinations. He or she then has to choose fare type. Only when these two steps have been taken is payment possible. The figure at the right shows the TVM which was evaluated. |  |
Fig. 2 The Autelca B100 train ticket vending machine |
1.1 Method | Users were observed and interviewed and the errors the users made were analyzed. Interviews and observations were carried out in the Central Stations of Amsterdam Rotterdam and Utrecht on three working days between 8.00 and 18.00 hours in August 1985. The TVMs had been in use for more than 9 months (since September 1984). Simple activities were observed without technical recording means. For example whether passengers had noticed that a TVM was out of order was observed by an investigator who stood near the TVM. For the observation of more complicated activities a hidden video camera was used (See the Figure at the right). | To collect data that could not be obtained by observation of overt behavior users were interviewed (n = 316). The passengers were interviewed immediately after they had bought a ticket. In this research project much effort was devoted to trying to detect errors made by users. There were several reasons for this. Firstly, it could be argued that errors made by users originate from errors made by designers and that designer errors can be analyzed by analyzing user errors. Secondly, the number of errors made provides very useful data when discussing the costs of ergonomics improvements. A distinction is made between several kinds of errors. |
 |
Fig. 3 The hidden camera was mounted on top of the luggage cart. The monitor, recorder and batteries were concealed in the luggage. |
| |
1.1.1 Errors noticed and corrected before paying
It was impossible to observe all the errors which the passengers corrected themselves. Very often there was no overt behavior to indicate that and error was made. However, it was possible to interpret from all kinds of additional information (head movements, hand movements, reaction times, hesitations) that a mistake had probably been made. |
1.1.2 Errors noticed and corrected after paying
Tickets that were returned by passengers for exchange were examined. For some of these the reason for their return was known but not all the returned tickets would be so analyzed. The reason for this was that ticket window personnel who received these tickets followed different procedures. One of them for example had been reselling the tickets. |
1.1.3 Errors not noticed by the users
After each interview the ticket bought by the passenger was checked. Table 1 gives the numbed of passengers involved in the observations and the interviews. The proportion of first-time users was about 10%.
|
| |
Table 1: Samples taken by the techniques used
|
|
| Study technique |
Sample size |
| Video observation with TVM users |
428 (A’dam 34%, R’dam 36%, Utrecht 31%) |
| Interviews with TVM users |
316 (A’dam 46%, R’dam 25%, Utrecht 28%) |
| Returned tickets: |
|
| reason for return known |
57 tickets |
| reason not known |
299 tickets
|
|
1.2 Results |
Several types of problem can occur in the interaction between intending passengers and a fare ticket vending machine. For instance we have observed that:
information presented by the system may not be noticed by the user;
information necessary to the transaction but not presented by the system, may not be available in user’s long term memory;
user’s tasks may not be clearly presented to them;
users may not perform all tasks in the correct order. |
These problems result in:
giving up the operation which happened for 13% of the first-time users
longer purchase time;
frustration;
wrongly ordered tickets.
|
| |
The purchasing errors that were made are in table 2. The figure of 1 .6% errors may seem very but the following points have to be considered:
It is supposed that the results are quite conservative because the criteria for selecting passengers were rather rigorous: all passengers used the vending machines voluntarily and passengers who were in a hurry (26%) did not answer the questionnaire and their tickets were not checked.
In the future about 190 000 tickets a day will be sold by these kinds of machines. In other words even at a 1.6% error rate every day about 3000 passengers will buy a wrong ticket.
The complexity of the system will increase and so probably will the percentage of errors; more destinations (from 80 now to 340 in the near future) more ticket types (now only 10) more kinds of payment (more types of banknotes, plastic money) will all be incorporated.
As we will argue, the errors that did occur were induced by the design. In a redesign based on an ergonomics analysis the number of errors can be minimised. |
In this paper we will confine ourselves to the passenger decision making problem only. We will not discuss, therefore, such issues as the routing problem - how the purchaser’s tasks flow through the machine layout.
The passenger had to make the following selections:
Choose one out of 80 destinations accounting for 18% of all errors. Destination errors are perceptual errors. These kinds of errors will be reduced by improving the lay-out of the list of destinations.
Choose full or reduced price accounting for 50% of all errors. These problems will be solved by presenting better information on the fare types.
Choose single return accounting for 18% of all errors. We expect that this number of errors can be minimised by using a more buttons to press system. This system will be discussed below.
Choose first or second class accounting for 9% of all errors.
|
| |
To exemplify our approach further details will be given on this last kind of error. Second class tickets could be bought by pressing one of the upper group of ticket type buttons. First class tickets could be obtained by pressing one of the lower group of buttons. (See Fig. at the right.)
Some passengers used the TVM without noticing that a class type selection had to be made. They bought a ticket of the wrong class and noticed this before paying after paying or not at all. These points will be discussed in more detail below. |  | Fig. 4. The ticket type buttons on the TVM, giving choices of single full fare single reduced fare return full fare return reduced fare or evening only return within the choice of 1st or 2nd class
|
| |
Table 2: Errors that went unnoticed during operation of the TVM made by passengers who were not in a hurry
|
| |
|
Proportion |
Estimated | |
of |
percentage |
|
returned |
over all |
|
tickets |
tickets |
|
9% |
0.147% |
| Errors with single/return | 13% |
0.213% |
|
| Errors with the price |
50% |
0.819% |
|
| Wrong destination |
18% |
0.295% |
|
| Wanted only to change a banknote |
4% |
0.065% |
|
| Wanted a ticket for the next day |
7% |
0.114% |
|
| Total |
100% | 1.6% |
|
|
n = 56 |
n = 122
|
*It should be stressed that the types of error as a percentage of all the tickets sold by the TVMS are very approximate estimates based only on two wrong tickets in a sample of 122 tickets which were checked by the interviewers. |
To top.
1.3 Recall of class type selection | When the passenger buys a ticket at the ticket window only first class passengers mention the class type.
Most are not used to paying attention to the class type. However, when buying a ticket at a TVM the passenger has to were made to passengers mention the class type. Several tests were made to see if the passengers could remember making this selection.
The first way to test if the passengers had paid attention to the class type selection was by asking if they could remember the difference in colour coding. The first class buttons were coded green. The second class buttons were white as were all the other buttons. The users were asked if they could remember if green buttons were present on the TVM. Only 14% could remember that there were green buttons and that these buttons were for first class tickets. No was the reply of 41% and 45% answered Don’t known. |
The second way to test if the passengers had paid attention to the class type selection was by presenting a color picture of the TVM. On that picture the texts first class and second class were removed. The (question was: Do you remember the difference between these two rows? Of 111 TVM users who were not in a hurry 59% could not recall this difference.
Finally, the third way to test the attention paid to the class type selections was to ask passengers if there were first class tickets available. Of the 55 second class users who were not in a hurry 16% did not remember that first class tickets were available and 2% answered not available. | Although recall of information concerning a decision taken one or two minutes ago is not equivalent to having noticed and used that information these figures suggest that many passengers may not have made a conscious class type selection.
|
1.4 Wrong class type tickets bought
| For this investigation, 122 tickets of passengers not in a hurry were checked. Two passengers had bought the wrong ticket (1.6%). As the numbers are far too low to be able to draw reliable conclusions about the number of wrong class type tickets bought in particular, an additional analysis was made of 356 tickets which for some reason were tickets the reason returned by passengers. For 57 of these tickets the reason for their return was known, and 9% changed their ticket because they bought a ticket of the wrong class type. |
Furthermore, of the 299 tickets which were returned and for which the reason for return was not known the percentage of first class tickets was 13%. Since the TVMs in practice sell only 2% first class tickets it is likely that 11% were returned because of a class type error. From these figures it is estimated that at least 0.147% of all passengers who are not in a hurry buy a ticket of the wrong class type.
|
1.4.1 Changing class before paying | It was observed also that passengers changed the class of their ticket before paying. Many of them did so because they wanted to compare prices which could easily be observed. These passengers pressed several buttons and then looked at the price each time. | The latter behavior was quite obvious because the price is presented somewhat to one side and rather large head movements have to be made to look at the price. If one disregards the price comparers this leaves four passengers (1%) who changed class probably because they noticed their class mistake. The performance of these passengers is described in detail:
|
| |
Passenger 1
A young man not looking like a first class passenger walked to the TVM. It could clearly be seen that his eyes were directed to the fare type buttons. When he reached the TVM he immediately pressed first class single full price After 3 s he pressed his destination and looked at the price. His eyes went to the fare type buttons his hand was in the first class area. 1 s later his hand went to the second class area and he immediately pressed a second class button. |
Passenger 2
A young man pressed a destination at the bottom of the destination field. 2 s later first class single reduced price was pressed (first class buttons are also at the bottom). 0,5 s later he looked at the price and seemed astonished. Looking at the destination and fare type again he pressed Press for mistake restarted and quickly bought a second class single reduced price ticket.
|
| |
Passenger 3
A middle-aged man studied the banknote slot for 3 s. After that he immediately pressed a destination. 5 s later second class single reduced price was pressed. Next he looked at the price. Then he went to the second class area with his eyes and his hand. After 3 s he went to the first class area and 1 s later he pressed first class single reduced price. He looked at the price and about 4 s later he pressed Press for mistake. Apparently at this moment he understood the operation of the TVM and intended to restart from the price and about 4 s later he pressed Press for mistake. Apparently at his moment he understood the operation of the TVM and intended to restart from beginning. In this second effort he bought a first class single reduced price ticket. |
Passenger 4
A well dressed gentleman immediately pressed a destination at the bottom of the list. 3 s later he pressed first class single reduced price. Then there were problems with the banknote acceptor. The TVM reset itself. The passenger again ordered a first class ticket. Again there were problems with the banknote and the passenger decided to try another TVM. However, he now ordered a second class ticket. Again he had problems with the banknote. The passenger pressed first class. The banknote was not accepted and the passenger left without a ticket.
|
| |
Passengers 1 and 2 obviously made class type mistakes. One cannot be sure for passengers 3 and 4 but they probably did as well. If so 1% of the passengers made a class type error but corrected it before paying.
|
2. Discussion | When one considers the perceptual and the mental activities which have to be performed using the TVM then the errors described above are not surprising.
First, we will discuss several ways to design ticket vending machines which aid decision making. When using a one button to press solution the user presses only one button to indicate a combination of several selections. A one-button system can consist of a list in which for every element all selections are presented on or near to an associated button. On the TVM evaluated this solution is used for the selection of the fare type (first/second class; single/return; whole day/evening only return; full/reduced price). (See the figure at the right.) | 
|
| |
Another one button to press system is the matrix type system (See Fig. at the right.). Here also for every combination of elements there is only one button. However, the information which belongs to this button is positioned elsewhere - for instance at the start of a list or to the left margin and at the top of the matrix. In Fig. 6 the decision single/return is taken at the top and the decision concerning class is taken to the left. On the TVM evaluated a one button to press is applied to the class type decision. |
Fig. 5 A one button to press” list system for a coffee vending machine
|
| |
When using a more buttons to press solution the user presses several buttons. For every decision they make a selection between the buttons of one particular group. The figure at the right shows a redesign of the TVM using this principle for selection of fare type.
From a psychological point of view the one button to press system has many disadvantages which will be of more importance when the system is complex (many decisions and much information). |  |
Fig. 7 A more buttons to press system on a redesign of the evaluated machine
|
| |
Perceptual disadvantages of one button system
For users there are two perceptual disadvantages of a one button to press system: (a) there is no need to go through the whole list and (b) the information cannot always be presented in the proximity of the controls.
As concluded
users of a one button system may read instructions and probably they will not go through the whole list of legends. One reason for not reading a whole list can be that the passenger is in a hurry; another reason can be that they do not expect a particular product to be in the list. The latter reason could be concluded from the results of this research. Although an evening return is about 40% cheaper than a day return and although an evening return has already been in use for 20 years of the passengers who bought a return ticket between 18.00 and 19.00 hours 30% pressed the more expensive day return button instead of the cheaper evening return button. This latter button is positioned one or two buttons below the former one.
A more buttons to press system has no list and consequently the user cannot make a mistake due to not going through the whole list. |
In a complicated system where a one button to press interface is used it is impossible to put all the necessary information on or very near the buttons. The visual angle between button and legend is increased and consequently the possibility of not noticing the legend increases as well
In a more buttons to press system there is less information for the number of buttons. Consequently, it is not such a problem to place information on or adjacent to the buttons.
|
| |
Cognitive disadvantages of one button system
For users, the cognitive disadvantages of a one button to press system are that there is a load on short term memory; that they have to apply a strategy which is mostly employed by experienced decision makers and that they have to understand the system.
A consequence of the frequent non-proximity of legends and buttons noted above is that when users do seek and note the relevant information, there is a requirement to store this in short term memory before selecting the relevant button. It can be argued that a prospective departing passenger should not be required to do this. In a more buttons to press system information can be presented with the buttons. There is no load on short term and consequently lower probability that the information will be forgotten. |
Psychologists often distinguish between two strategies in perception and decision making. When using a sequential strategy people take decisions one after the other. This strategy would perhaps be mostly employed by inexperienced decision makers for instance children who read words or passengers who use a TVM. When using a parallel or a simultaneous strategy people immediately and automatically identify an object or make a compound decision. This strategy is mostly employed by experienced decision makers - for instance adults reading words and staff selling tickets 8 h a day. As passengers are often stressed it could be argued that a human machine interface which forces a sequential strategy is preferable. This choice is supported by
who investigated the speed and accuracy of people using a variety of presentation formats for numerical tables. She concluded that taking four binary decisions one after the other (tree-structure) is faster and less error prone than taking two binary decisions at the same time (matrix-structure).
When one considers the decision-making strategies that people use it is better to force departing passengers to take their decisions in a sequence. A more buttons to press system is preferable.
Finally, the understanding of the operation should be mentioned.
From several investigations it can be concluded that 60% of the population did not understand this way of presenting information. This could be put forward as an argument in favour of the more buttons to press system.
|
3. Conclusion | It is argued that both empirical data and theoretical analysis clearly indicate that a more buttons to press system is preferable for users who will not become experts and who are stressed. For the users of the TVM evaluated, this conclusion comes too late.
The manufacturer is not able to change the machine. For technicians and ergonomists involved in designing these kinds of machines a lesson can be learned. | The designers of the TVM have chosen the wrong alternative because the psychological disadvantages mentioned above (non-proximity of legends and controls, load on short term memory, parallel decision strategy and need to understand the system) are less obvious than one practical advantage of the one button to press-system - that is the less work load for the fingers because only one button has to be pressed.
|
Acknowledgement | The author is indebted to colleagues in the department of Ergonomic Psychology at the University of Utrecht and to colleagues in the Ergonomic Advisory Group of Netherlands Railways for discussion, field observations and the construction of research instruments.
|
References |
Bartram, D. , (1984).The presentation of information about bus servicespag. 299-320. Easterby & Zwaga, Information Design
Cole, B.L., & Hughes, P.K., (1984). A Field Trial of Attention and Search Conspicuity Human Factors. Vol. 26, no 3 pag. 299-313.
Easterby, R., & Zwaga, H., (1984).Information Design The design and evaluation of signs and printed material. Chichester etc.: Wiley and Sons Ltd.
Hoonhout, J., (1985).De Autelca plaatskaartenautomaat een veldonderzoek in Utrecht-centraal (The Autelca ticket vending machine. Utrecht: Psychologisch Laboratorium Rijksuniversiteit Utrecht.
Oborne, D.J., (1986).Contemporary Ergonomics Proc. of the Erg. Soc. Ann. Conf. Durham England, 8-11 Apr. London, & Philadelphia: Taylor, & Francis. Proc. of the Erg. Soc. Ann. Conf. Durham England, 8-11 Apr.
Wright, P., (1977).Decision Making as a Factor in the Ease of Using Numerical Tables Ergonomics. Vol. 20, no 1 pag. 91-96.
|
|