_____                   _                  _____            _____       _ 
  |     |___ _____ ___ _ _| |_ ___ ___ ___   |  _  |___ ___   | __  |___ _| |
  |   --| . |     | . | | |  _| -_|  _|_ -|  |     |  _| -_|  | __ -| .'| . |
  |_____|___|_|_|_|  _|___|_| |___|_| |___|  |__|__|_| |___|  |_____|__,|___|
  a newsletter by |_| j. b. crawford                       home subscribe rss

>>> 2021-02-27 kitchen videos

It's important to take a break every once in a while, so let's distract ourselves from telephony for a bit and talk about another old favorite of mine, point of sale. The needs of the point of sale have produced a number of interesting computer systems, but restaurants have a particular set of constraints and requirements that have produced a universe of computer solutions dedicated to restaurants.

Unlike retail point of sale, which does seem to have a small corps of amateur historians, it's relatively difficult to find information on the history of restaurant systems. As a result, some of what I say here will be a bit speculative, based on assumptions made from things that I do know. But to start, what differentiates restaurant POS from other POS applications?

We previously discussed how POS equipment has evolved over years from mechanical cash registers that were limited to totaling purchase amounts to enterprise computer systems that automate back-office functions like inventory management and reporting. Similarly, restaurant POS systems have expanded over time to cover automation needs specifically to restaurants. A full-featured restaurant POS is expected to coordinate the kitchen with the front of house (FOH) while also automating the more bureaucratic parts of the FOH like assigning parties to tables, reserving tables, tip accounting, etc.

Restaurants vary widely in how much they lean on automation for these functions. Generally, larger and more "corporate" restaurants (e.g. chains), and especially quick-service and fast-food restaurants, are likely to lean more heavily on automation than more local, boutique operations. As you might imagine from that generalization, McDonalds is more or less the peak of conventional restaurant technology; given the history of that chain it might not be surprising that it was also an important innovator in the field.

It is widely reported that the first restaurant POS system was created in 1974 by Brobeck and Associates and was used by McDonalds. This is, of course, wrong. Not completely wrong, but wrong in a way that makes research rather frustrating. William Brobeck was a nuclear physicist by training, and through his companies Brobeck and Associates and Cyclotron Corp is best known for his work designing cyclotrons for various applications. He was also, I have found, a consummate tinkerer, and published designs and filed patents for various robotic devices. As best I can tell, Brobeck and Associates never had anything to do with any POS systems, but Brobeck also founded a company called Transactron to do exactly that.

In fact, once you know what to search for, you'll find that the Computer History Museum has a Transactron McDonalds POS device in their collection. I have found photos of the same device branded Transactron, and CHM lists it as manufactured by Transactron, but theirs actually bears the logo of "Courier Terminal Systems." There is very little information out there about Courier[1]. The patent covering the device is in the name of Transactron, but I speculate that at least early on Transactron may have partnered with Courier to actually manufacture the unit. Oddly, the Computer History Museum also lists their example as "circa 1972," which seems unlikely considering that the patent was filed 1974 and most other sources say that McDonalds began use of the system that year. But, well, they said "circa" after all.

The Transactron system is well described by the patent, and is surprisingly feature-rich for the time. Based around an Intel 8008, it consists of multiple terminals with a grid of buttons, most of which are labeled with menu items but some of which contain numeric keys or function keys (such as total). Orders can be edited as they are entered and can be stored for recall, both of which were features that surprised me for such an early example.

The terminals seem to have been "dumb" devices and the microcomputer logic was housed in a central location and connected to a printer. Each order entered at a terminal would be printed.

One thing I am very curious about is the extent of the relationship between Transactron and McDonalds. Was the system originally designed as a partnership, or did McDonalds purchase it more or less off the shelf? If it was originally designed as a partnership or on commission, that might explain why it is widely said that Brobeck and Associates designed the device while the patent is under the name of Transactron---perhaps they formed Transactron as a company to market the device, after Brobeck and Associates designed it. Oddly, online biographies of Brobeck tend to completely omit his involvement in something as impressive as the first automation of McDonalds, apparently obscure compared to his work on cyclotrons. It seems to be, overall, rather forgotten.

This basic design of a restaurant POS has proven quite durable and is still in use today in many smaller or less automated restaurants. Orders are entered front of house and, once committed, are printed in the form of a "kitchen ticket" by a printer in the kitchen. This is essentially just a light automation of the older practice of waiters hand-writing orders on a paper ticket which they deliver to the kitchen, which is once again still in use at many restaurants today.

The kitchen printer is a somewhat specialized animal. The direct thermal paper typically used for POS applications (e.g. for receipts) does not hold up well to exposure to grease and heat, both of which are present in abundance in the kitchen. For that reason, kitchen printers are typically actually impact printers on plain paper, and the most popular model from Epson makes use of a two-part tape that allows for printing in black or red---a relatively inexpensive enhancement for impact printers that can't be done with direct thermal printers. This leads to a sort of irony that kitchen tickets are often printed in two colors (e.g. red for substitutions), which is a bit fancier than the consumer-facing receipts. On the other hand, most major restaurants seem to be using a twenty-year-old printer with a dry tape and in poor adjustment that produces extremely light output. Printers always have their problems [2].

You can likely see an obvious enhancement to this concept of the FOH POS automatically 'sending' the ticket to the kitchen via a kitchen printer. What if, instead of a printed slip of paper, the kitchen made use of computers to view and manage orders as well? This is clearly an interesting idea, but the practical constraints of operating computers in the harsh kitchen environment made it impractical for many years.

As best I can tell, the first such system was patented in 1981. The term was not yet in use, but by the end of the '90s such a system would be called a "Kitchen Display System" or KDS (less frequently Kitchen Video System or KVS). This early patent described a system where a series of letters were displayed on a CRT corresponding to different items which had been ordered. It is clearly very primitive, but was presumably heavily limited by the microcomputer technology of the time.

This patent was originally assigned to OCR Marketing Associates, which I can find no information about. However, in 1988 it was assigned to the Stanley Hayman company, and one of the inventors, Richard Hayman, is listed on later Hayman company patents and just by the name may have been related to the founder [3]. As a result, I suspect OCR Marketing Associates may have been a subsidiary or was otherwise related to the Hayman company from the beginning.

In any case, much like Brobeck's 1974 work, Hayman et al's 1981 patent lays out the groundwork for the kitchen display systems that are still in use today. Busy restaurant kitchens often consist of multiple people at various stations which specialize in specific items or methods of preparation. As a result, a single order is often cooked by multiple people, and finally an individual called an expediter will collect the items to consolidate them into one dish. This basic process can be seen very clearly at most fast food restaurants, where line cooks arranged in an assembly-line fashion will slide sandwiches down a counter while a fry cook collects fried items and puts them on a tray; the expediter gathers both and bags them to finish a typical burger-and-fries combo.

This process can clearly be frustrated by the need to pass a paper ticket around, but multiple paper tickets create their own problems as it becomes difficult for the expediter to be sure what goes together. Instead, the Hayman patent describes a system in which each workstation has a computer display which shows only the items to be done at that station. The expediter's display does not show an order at all until the cooks have indicated that they completed the preparation of the items in the order, so they should be ready for the expediter to collect.

Actually the Hayman system is not quite that sophisticated, it simplifies the electronics by having many of the displays be exact mirrors with symbols indicating which station should pay attention to which items. The expediter, though, is provided with an individually controlled display so that they are not distracted or confused by the items not yet ready for them.

In the Hayman patent, a standard keyboard is apparently used for data entry in the kitchen. In the kitchen environment this must have been a rather high-maintenance piece of hardware. In part due to the greasy environment in a kitchen and in part because of the relatively higher complexity of connecting multiple keyboards to a single computer, the keyboard would be replaced in KDS applications by a "bump bar."

A bump bar is an input device, usually using membrane keys for ease of cleaning, that usually consists of a row of numbers corresponding to positions on the display where order tickets are shown, and several action buttons, the most important of which is "bump." Selecting a ticket and pressing "bump" indicates that the ticket is complete at that station. Typically, once all stations preparing components of an order bump the ticket, the ticket will appear at the expediting station.

An important parallel innovation in point of sale was the increasing popularity of the microcomputer. Most point of sale systems into the '90s were based on a mainframe architecture in which individual devices (cash registers, kitchen display stations, etc) acted as terminals to a midcomputer or minicomputer. By the '80s, though, it became possible (albeit complex) to use a system of networked microcomputers to build a similar system.

The relation of the microcomputer to to restaurant POS is rather interesting due to an interesting central character, Gene Mosher. Mosher operated delis in New York City and was apparently also a bit of a dweeb. He reports that, in 1978, he started writing software for his early-production Apple II to manage POS at his delis. In 1986, he upgraded to an Atari ST with a touchscreen. He marketed his touchscreen restaurant POS system (believed to be the first) under the name ViewTouch, and delightfully, a descendant of ViewTouch is open-source today, still maintained by Mosher and now targeting devices like tablets and the Raspberry Pi.

In fact, the development of practical touchscreens was nearly as significant to POS technology as networked microcomputers. Virtually all restaurant POS today is touchscreen at the FOH, and increasingly many retail systems are as well. Mosher's influence on touchscreen POS is clear. The faux marble GIF background is decidedly dated, but otherwise ViewTouch looks nearly identical to most touchscreen restaurant POS products today. Mosher seems justifiably a bit peeved at the extent to which major players like NCR seem to have copied his work.

That brings us more or less to the modern day of traditional, enterprise restaurant POS solutions. For large companies, there are two dominant players: Oracle Micros, and NCR Aloha. Micros, today part of Larry Ellison's wrathful empire, actually acquired the previously mentioned Hayman company in 1999 but seems to have licensed patents from them even earlier, as Micros was an early player in the KDS market with a terminal-and-minicomputer solution. By the time of the acquisition, Hayman was actually described primarily as a value-added reseller (VAR) of Micros, suggesting that they had lost their technology edge. The acquisition of Hayman was part of a larger trend (still going on today) of legacy technology vendors shifting towards direct-to-consumer sales by acquiring their former distributors and VARs.

Aloha was established to market the (believed) first Windows-based restaurant POS system in 1992, and so is firmly rooted in networked microcomputer solutions. They were later acquired by Radiant Systems, and you will occasionally still see Aloha systems showing the Radiant logo. More recently, though, Radiant was acquired by POS giant NCR, leaving the current branding.

Despite the separate histories of Micros and Aloha, both are, today, microcomputer-based. Aloha remains Windows hosted, while Micros is available for both Windows and Oracle Linux (it was presumably UNIX based for a time). Micros is often (but not necessarily) run directly on Oracle-manufactured hardware, but Aloha is most commonly seen on purpose-built restaurant POS devices from Elo Touchsystems. The two are physically fairly similar, and if you will picture with me, in your mind's eye, the device that a waiter taps on, you are seeing either a Micros Workstation or an Elo Touchsystems product.

Because of the difficult kitchen environment, touchscreens are not common for KDS. Most modern KDS still use a physical bump bar, which are available from some vendors with lightning connectors, because everything has to be these days.

In the modern world most of these established restaurant solutions are starting to look pretty legacy, due to competition from several dozen upstarts with a basic knowledge of PHP and a love of iPads. A quick google for "restaurant pos" will find you a dozen iOS-based, cloud-backed restaurant POS solutions with varying degrees of feature-completeness compared to legacy solutions.

Because of the relatively high cost of kitchen-quality hardware, and hardware in general [4], it's common for these solutions to still use a traditional kitchen printer or lack kitchen integration at all. Similarly, the credit card situation might be fiddly and dependent on Bluetooth and battery charging, and all of the other limitations I have previously discussed as coming from the increasing commodification of computing.

Or, for new readers, I will attempt to quickly summarize this theory of mine in the context of what we have just discussed: The restaurant POS solutions of the '80s, '90s, and even '00s were designed nearly full-stack by dedicated engineers for a specific purpose. As a result, they are very well suited for that purpose, but they are also expensive and often only available though irritating sets of VARs and service contracts due to the long shadow of IBM's strategy of computers as leased rather than owned. "Modern" restaurant POS solutions are rapidly designed on top of consumer hardware and consumer operating systems, which makes them less expensive and (at least perceptually) more friendly, but at the same time, tends to make them less suited for purpose.

In this way, the evolution of computing as a universal commodity has made computer systems more widely available but also subtly and pervasively worse.

Consider an example: touchscreens are specifically uncommon in the kitchen for legacy systems because kitchen staff often have dirty hands and/or are wearing gloves. These problems can be ameliorated by either using a physical bump bar (most common due to low cost, ease of cleaning, etc), or by the use of resistive or acoustic touchscreens. For similar reasons, dedicated touchscreen POS devices typically use resistive or acoustic touchscreens because FOH staff as well sometimes wear gloves, and are also fond of doing things like using the edge of a credit card to tap buttons.

iPads, though, are the development platform of today, and are not available with bump bars or non-capacitive touchscreens. Although in some cases specialized hardware features are available as accessories, the majority of users of such "off-the-shelf" hardware systems are attracted to them as low-cost, all-in-one solutions, and so instead of purchasing specialized accessories they just settle for the limitations presented by a consumer tablet in a special-purpose commercial application. Consider, for example, how rare it is for Square-based stores to be able to produce a printed receipt, something that was long considered an absolutely core feature of a POS device.

For relatively high-cost specialized systems, integration is usually viewed as a core part of the offering. For this reason, POS solutions, KDS, hospitality systems, etc produced by different major vendors can typically be integrated with each other. Producers of modern off-the-shelf systems tend to view integration as an unnecessary expense, though, which leads to the comical situation of restaurants with four or more iPads (someone who had heard this rant from me once sent me a photo of a restaurant with six), each of which runs one app, several usually being for different delivery services. Someone at the front desk has to copy incoming delivery orders from one iPad to another, the paragon of computing efficiency.

None of these problems are necessary an intrinsic limitation of the modern tech industry but they do seem to be, to at least some large degree, an intrinsic result of the values and motivations of tech entrepreneurs and investors, which usually value time-to-market over fitness-for-purpose.

You see, the development of the modern restaurant POS has taken the better part of fifty years. A minimum viable version of a restaurant POS can be produced in a week, and someone will probably give you $25mm to sell it to small restaurants at a cost too low to refuse. iPad not included.

I have left out several parts of this theory such as the powerful component of consumer-as-employee expectations, and left out the entire part of this story which is Oracle repeatedly and aggressively scaring off customers, but I've already put a lot of words here. Continuing this line of inquiry, though, in the future I will be talking about hospitality management systems: POS for hotels.

[1] The CHM also has an old Courier brochure listing a Phoenix address and advertising a terminal they apparently manufactured, so they seem to have been a small company with an in-house manufacturing capability.

[2] That's not entirely true, the Epson TM series direct thermal printers are flawless and I love them, which is why I keep buying used ones and finding stupid uses for them around my house. I think I've mentioned this before but, like, it's basically my main hobby.

[3] I found an obituary that would seem to confirm that Richard Hayman was Stanley Hayman's son, but I'm not entirely sure it's the right Haymans.

[4] Yes, "they can't afford hardware" is a weird thing to say about businesses that view the iPad as an embedded platform, but somehow it seems like in the mind of many small businesses the Apple store hits the wallet a lot softer than the same bill from a POS vendor.