Humanizing the Machine

Portibiti is dry cleaner conveyor system developed by Metalprogetti, a leading company based in Italy, that designs and manufactures automated systems for moving, sorting, and handling hanging and folded garments as well as a range of other light items. The innovation of mixing computer systems with mechanical conveyor systems was and still is leading the industry in capability. Unfortunately, as the product gained popularity from the time and cost savings its owners experienced, installations and features started growing quickly. The “early adaptors” of this technology were eager to get, use, and understand the machine and its capability. As the product became more widespread, and the decision to purchase a machine in the $60,000 to $450,000 range became more of a business decision, machines were placed in plants, stores, and operations where the manager and employees often did not want them, or did not understand how to use and benefit from them.

The group of engineers and designers at Metalprogetti are highly innovative and dedicated to the work they do. Their testing lab looks more like the showroom of a Ferrari dealership than the workshop of a machine manufacturing operation. They are well educated and some of the top engineers. As new machines are designed, they arrive in the testing lab, are operated as designed, and made to perform as designed. To add to the initial success, the “early adaptors” of the product were a group of well-educated business owners that understood the workings of the machine and how to make it perform as expected. In fact, many of the new owners would take a trip to Perugia, Italy to see their machine being built and tested. These owners had a personal interest in using the latest technology and making it work for them.

Metalprogetti gained lots of attention. It was the star of trade shows for the textile industry, and the testimony of the “early adaptors” was very beneficial to the company; however, as the product customer base grew, the accuracy many newer customers achieved fell dramatically. The research into these problems dealt with the human computer interface and not the design of the machine itself. Metalprogetti’s lab results, as well as those first customers, was more than a proof of concept, it verified the system worked well when used as designed. The following paragraphs discuss the causes of this decline and the steps taken to bring the desired accuracy to all the users.

The accuracy observed with the machine operated by skilled workers was everything the designers and developers wanted. Unless there was a mechanical failure, the machine operated as expected with near perfect results, and with the checks in the software system, the machine would give the operator enough information to find the problem if one did occur. Then, with the largest single day sale of machines from a trade show, Metalprogetti sold twenty-four systems to be built and installed in rapid succession. A new type of customer started purchasing and using the machine, and for the first time, the system was not performing as expected. It was imperative to find the issues and resolve them.

Testing in a laboratory differs from studying that same item in the real world. Many things shape what actually happens in the real world that can be taken out of the environment in a lab. Progress happens when some piece of research adds new knowledge to or corrects other knowledge [1]. Our field consisted of nineteen different locations throughout the United States. Of these locations, fifteen were seeing far too many failures to make this a viable product to use. Four locations were seeing the expected results, and the results consistent with those “early adaptors” that created much of the desire in Metalprogetti’s products.

To operate a Metalprogetti conveyor, a user scans a barcode attached to a garment. The conveyor records the item number and is ready to accept the item if it is in its database. Then continues to accept additional items until all pieces of the order are collected. In a different location, the conveyor deposits the complete order to the exit rail, giving space for more garments to be added. Ideally, when all garments have been loaded onto the conveyor, all orders are complete and delivered to the exit rail, and the conveyor is empty. Initially, short cryptic messages from a small display was the only communication from the machine to the user.

Our first observations in Orlando Florida were strictly from the point of view that the machine was not performing in the same manner. The owners described how the employees had been trained, and they were certain the machine was the culprit of the high failure rate. After two days of observation and interaction with the development and support staff in Italy, it was quite clear the machine was working as it was intended. This shifted our efforts to finding the problems with the use of the machine. Our first field observation was a location that had a failure rate of 20%. In real numbers, this was over 400 garments not being returned to their rightful owner. The majority of the problem at this first location was not the users of the machine, but the method of entering the items into the Point of Sale (POS) software was not correct. Therefore, the machine simply did not know about certain items and could not successfully sort them. The message was simply “NOT FOUND” and gave the user no context to continue. We observed some users did not understand the message displayed and would either load the item anyway or scan another barcode to get the machine to accept the prior item. The development team was unable to understand this use of the machine and offered almost no help, simply stating the employees were causing the problem. While this location was having the largest problem, our initial thoughts were to correct the problems we observed and move on to the next few locations having similar problems. We identified corrective actions for two major problems:

1. 1.

   Correct the process being used to enter the items into the system at the beginning, eliminating the “NOT FOUND” error that was encountered most often.

2. 2.

   Correct the process employees used when a “NOT FOUND” message did appear so the problem would not spread to other garments and compound the issue.

Our initial findings were relayed back to the owners and the Metalprogetti development team that more comprehensive training and better processes were needed at this location. We were still concerned that the number of locations that were experiencing problems was growing at about the same rate as installations. Something was very different about the first thirty machines and the results those owners were seeing versus the second thirty. Later we would return to this location to find we solved only part of the problem.

Our next location, Jacksonville Florida, showed an entirely different aspect of the problem. The items were entered into the Point of Sale (POS) software correctly and consequently, the Metalprogetti conveyor recognized the item. The rate of failure at this location was just under 10%, which represented nearly 100 garments going to the wrong order. As we observed these employees, they appeared confident and knowledgeable about their job, yet they often would read the display and take what seemed to be random actions based on the same message. As we discussed the issues with the employees, we quickly learned they spoke different languages and usually just pretended they understood. With more observation and more pressing questions about the actions being taken, we found out this particular location had thirteen different languages spoken by the employees, with only a few translators in the group. The simplest of messages were not understood and the prevailing culture was not to ask questions and show you didn’t understand English. To resolve the first level of issues found at this location, as series of post-it notes and a chart of common messages with translations was located at each loading station to help the users understand the messages appearing on the screen. We also noticed many users did not understand the difference between machine status messages versus messages about the item to be sorted.

To solve this issue of understanding the two message types, we installed a yellow light and a small buzzer to inform the user that there was a machine issue and not a problem with the garment or items being sorted. This allowed the user to find the problem with the machine if they knew how, or find another employee that could help them. Just knowing it wasn’t something they did or should have done allowed them the knowledge of what to do next.

We then visited two locations with new conveyors, one in Las Vegas and one in Nashville, where both operations were achieving the results they expected. At both locations, we added the yellow light and buzzer and asked them to report back in a week about what they thought of the addition. The responses were identical: it helped to quickly realize when there was a machine problem so they could resolve it.

In both Chicago and Boston, we ran into a new problem. The operators of the machines could speak English, but it took a while for us to realize that a few of them were illiterate. The operators who were illiterate were not willing to disclose this information freely, but a few of their confidants informed us of the issues without letting anyone know they had told us. We learned this is common with many illiterate people [2]. In 2014, 47% of Chicago’s residents cannot read [3] and many of the lower skilled and lower educated people were the types of employees that ended up working on Metalprogetti conveyors. These locations achieved 95% accuracy overall. This was due to some operators being able to read and understand the messages and other operators making their best guess at what they should do when a message appeared that they did not understand.

During the short term, we continued working with the nineteen installations, to improve training and processes, but we knew the long term solution was to design the machine to communicate better with the users to achieve the desired outcome across the board.

New displays were developed to further communicate with the operator so they were aware of the next step. See Fig. 1. The text-based display is still present, but depending on the system two to four lights and up to two audio signals can now be added to communicate with the operator. In most cases, the lights added are a green light telling the operator it is now ready to accept the item to be sorted and a red light alerting the operator that an issue has occurred with their previous action. The audio signals allowed another form of communication.

New user interface with additional lights for enhanced communication.

Full size image

In one location, during a particularly bad day of employee absences, the ten-year-old daughter of the owner was allowed to load the conveyor. Her instructions were to continue only if she saw a green light. She was successful during her four-hour shift on the conveyor with the lights and buzzers installed.

The audio signals we added were two piezo buzzers attached to the display: a soft chime for when the scan was accepted, notifying the operator to continue with loading the machine, and a slightly harsher buzzer to alert a problem with the either the last action or the status of the conveyor.

The conveyor is now shipped with many more lights than those early days, including lights to display the status of compressed air and certain electrical inputs. This simple form of communication allows almost any level of employee to use the machine as effectively as the early adaptors and almost as well as the development team itself. During the installation of the machine, customers are offered the additional communication of the chime and buzzer.

The Metalprogetti conveyor was originally designed and initially deployed in the most desirable atmosphere for a sophisticated machine. Developers and initial users were not only impressed with the accuracy and ability of the machine but excited about using the new technology. When the success of the machine became known in the industry, many machines were installed in a much less desirable atmosphere for the same level of success. With a better understanding of the new conditions in which the machine was being used, better forms of Human Computer Interface elements were added. Metalprogetti now installs over 700 systems per year.