The New Interface Advocate

Vending machines provide an excellent and disappointingly universal study in overly complex interfaces. This brief post reflects on how the current designs are flawed, speculates on the forces behind the bad designs, and proposes a new design that overcomes current problems though an ultra-minimal interface.

Thanks to what appears to be historical inertia, vending machines demand the user work through varying degrees of abstraction:  the worst demand selection be indicated by typing a semi-arbitrary letter/number ID, the best use large buttons with pictures of available products.  The flaw in the former case is obvious, and in the latter case, the picture abstraction rather than direct display of the product precipitates infuriating incorrect product purchases if the vendor makes any loading errors at all.  Granted, using the main body for advertising precludes a direct product display, but in an age of pervasive, intrusive advertising, the gain over a large top banner is undoubtedly wiped out with ill-will garnered by the purchase errors that design nearly guarantees.  The first design aspect, then, is to display the actual products for purchase and select them directly.  Obviously, for security reasons, they must be enclosed and therefore not directly touched, but as any window-shopper or zoo-goer knows, there is another indication convention for things close by but behind glass—touching the glass itself directly in front of the item to indicate.  This gesture is exceedingly well-suited to selection; small children readily adopt it, it is trivial to remember, and it is exceptionally fast.

This approach, by exposing the underlying stock, confers additional improvements.  As in existing glass display fronted machines, exhaustion of particular products is apparent to users without any “sold out” indicators, such as the easily overlooked small LED on soda machine buttons (If the user touches in front of an empty item, the system should ignore it.  This is an intuitive, simple means to reject what is necessarily erroneous input).  Another paralleled feature is that though pricing might be incorrectly set, the user will not be surprised by the dispensed item because of stocking mistakes.

Some concerns for the touch glass in front of item selection method should be addressed now.  First, there is the obvious problem of dirtying from the touches.  I believe a careful combination of texturing, coating, light tinting, and/or using the right material can mitigate the problem, not to mention that increasing awareness of sanitary conditions already seems to be driving more frequent cleaning of public surfaces.  Second, the technology to register simple, non-moving, single point touch should be reasonably inexpensive in mass production.  The resolution of the sensor can be very coarse and still effective, which further drives down costs.

An easy means to indicate selection is dimming the lights on every other product and brightening the one for the selection.  The exact amounts should be governed by an ambient light sensor (also very cheap hardware) so the same machine can be used indoors and out.  As a bonus, this sensor can scale back the general state light in dimmer areas, which saves power.

A motion sensor (again, cheap technology) can reset the selection if a user walks away so incomplete purchases don’t affect the subsequent user.

Another flaw in current designs is the rarity of error-correction or prevention mechanisms.  On some number/letter entry systems, if the user were to type “a12″ as an “a” followed by “1″ and then “2″ (in place of the “correct” sequence “a” then the “12″ button), they would be given “a1″ immediately.  Sure, earlier days of vending probably could expect careful examination of such novel machines prior to use, but ubiquity has changed this norm.  Why not make the act of paying indicate selection finalization?  Payment must be given in any system, and placing it after selection mirrors the retail sequence where selection precedes payment.  No longer should finicky cash and coin acceptors justify demanding the that customer prove financial capability before even permitting item selection.  This re-ordering maximizes the time available to notice and correct errors.  In fact, coupled with the touch interface, the act of correction—(touch) “Oops, I don’t want that.” (touch elsewhere)—is so natural, that explaining it is only necessary to assure customers that they aren’t dealing with an arbitrarily hobbled interface.

Now that processing fee schedules are available for low-margin and low transaction value patterns, a vendor should accept credit-card payment.  This eases the burden for consumers who live in an increasingly cashless society.  To head off the annoyance of guessing reader orientation, the magnetic sensor in the credit card reader should be paired so swiping either orientation works.  Emerging features like the RFID-powered “PayPass” and “ExpressPay” also offer payment task improvements.

The entire transaction then becomes as little as two motions.  One, touch the glass in front of a desired item, and two, pay with a card- or RFID-swipe.  Correcting an error adds only one motion, and using cash remains the slowest, most complex method of payment, though it is as easy as the best other machines that accept cash.  Below are two pictures that illustrate one possible implementation of this overall design.  The first is in the default state, the second shows when an item is selected. Note that these two states alone provide all the dynamic information needed. (Click on the images for larger versions.)