Letter to Matt

April 30, 2015


I appreciate your speaking with me about the issues we touched on after the last all-hands meeting. I appreciate your meeting with me now.  I have always been impressed with Amazon’s open door policy. Any time I have needed to speak with a manager for any reason, they have always taken the time to address any issue.

As I mentioned to you, I often find it easier to collect my thoughts on paper, rather than trying to speak off-the-cuff, so I have attempted to create a small, but relatively comprehensive document expanding on our conversation. I want to begin with a disclaimer–the following is only based on my perceptions and my opinions. I have worked at Amazon for less than a year, and I have only worked in two departments. I lack a lot of experience and a lot of knowledge of the inner workings of the facility, management practices, and procedures of the company as a whole. For instance, you spoke to me about the scale of the workforce which always has to be kept in mind when considering implementing any new policies. What works for a smaller, more intimate work environment, might never be practical with the growing scale of SDF8 and the wider company.

I am concerned about any policies, especially policies designed to increase efficiency and productivity, which actually might have a reverse effect. We spoke briefly about this. The facility team is already addressing this problem in a very big way with the new shift to team production goals, to be evaluated monthly for team members. I believe that this has an immediate positive effect on employee morale. I say this because it aligns with my own theories of a productive work environment, but also because of the positive talk on the floor. People feel that they are under less daily pressure to meet goals which are sometimes out of their control.

We have mentioned employee morale and production goals. I believe, and I know the company understands, that employee morale and production are interrelated across a wide spectrum of variable factors. It is important to take a multifaceted, nuanced approach to find a practical balance which will allow us to maintain a level of employee satisfaction and meet production goals. Production goals can be measured by the seconds and minutes as well as by months, years, and decades.

If our goal is maximum production efficiency, that is, to do the most work for the least cost in the least amount of time, then why do we care about employee morale? I’m going to take the position that we do not care at all about employee morale, except where it can help us attain our goal of maximum production efficiency, measured in the short and long term.

A worker’s discontent is, in my opinion, a barrier to maximum productivity in the short and long run. What happens when a worker is unhappy? The employer can, in a sense, force the employee to make production goals with threats of disciplinary action or withholding of advancement opportunities, but these actions will further alienate the worker. She will lose motivation, miss days, possibly have health issues, emotional problems, and she will eventually quit or be fired. When she quits or gets fired, she has to be replaced by a new, untrained employee who might turn out to be a very skilled and efficient team member, or not.

SDF8 has a fairly high turn-over rate. We spoke about this briefly. It is the nature of this kind of work, and as you mentioned, the turn-over rate might seem higher than it actually is–people move to other shifts, other departments. Many new employees are not taking the place of a lost employee, they are only filling a new position created by production growth. This is a good thing.

I am concerned, though, that the labor pool for new or replacement employees may be shrinking due to Amazon’s poor reputation as an employer in the community. I know the company does a considerable amount of research on this topic, so my perception might have zero importance. Nevertheless, I feel we need a solid, experienced workforce to maintain production efficiency and long term viability. A shrinking pool of potential employees means the company has to choose from a group of applicants with fewer and fewer qualifications. Not that associate jobs at SDF8 require a highly skilled work force, but we absolutely benefit from having a workforce with a general level of desired competency. An always inexperienced workforce requires micromanagement, decreasing workflow–even the most mundane judgement calls require management approval. For instance, missing-out an item in AFE, one of the most fundamental actions a picker or packer must engage in daily, now requires gatekeeper or process-guide approval. We constantly have to teach even the simplest of concepts (how to place a box on a conveyer). Experienced employees obtain consistently higher rates with fewer problems. It is always preferable to favor experienced workers. A constant supply of new, untrained workers from a diminished workforce pool because of a bad reputation in the community = low productivity.

How do we increase worker contentment, thereby lowering attrition and raising Amazon’s reputation in the community as an employer? Some big ways and many small ways. One very obvious way, which is probably a non-starter, is to give everyone a raise. Start everyone at twelve an hour and keep the standard rate increases intact as they stand. This would bump up morale, reduce turn-over, and slightly increase the pool of qualified potential applicants in the community. I believe the company already has a good benefits package. I don’t think I’ve ever heard anyone complain about any part of it. In fact, many people talk about how good the health benefits are, etc. One other thing I might suggest is to allow stock to vest after one year instead of two. This would offer employees more figurative and literal ownership in the company, increasing morale and worker engagement. Of course, the above suggestions would increase production cost, and would have to be weighed against the perceived increases in stability and efficiency.

I was a direct hire in July of 2014.  I was assigned to AFE donut nights. It began to seem to me, from fairly early on, that managers and PAs, implementing company policies with varying degrees consistency, were alienating some of their best workers with constant hassling about rates, break times, talking, and other every-day interactions, creating a needlessly high-stress work environment. AFE is probably the highest stress production area in the facility, so the AFE environment can heighten some of these problems. Since moving to pick and day shift, I have felt more relaxed, and I believe the pick/stow/ICQA environment is more relaxed as a whole. But let me talk about the problem:

We talked in the last all-hands meeting about the need for better training for associates, and I agree with this and will offer a few suggestions momentarily. First I want to point out that some of our managers and PAs are also, in my opinion, poorly trained and/or inexperienced. I realize that all managers and PAs are going to have a unique management style, but some may not be following company policies regarding respect for and open dialogue with all employees. When this manifests into a disharmonious work environment, it can become a barrier to production. I’m not here to name names, only to suggest better adherence to the new team-goals methodology by embracing it with consistently respectful management styles. I’ll give a few examples of what I mean by poor management styles or hassling good employees:

–When I was in AFE, another employee and I were greeting each other at our pack stations at the start of shift. My fellow employee had been here over a year, is a consistent performer, and has a very quiet demeanor. A PA walked by and said, “Less talk and more packing, please.” I am a solid performer as well, and we looked at each other like we couldn’t believe what just happened. My friend, who, again, is a very solid employee said, “They treat us like children.”

–A manager in AFE said frequently at standup, and over the speaker throughout the shift that if we are late from breaks, “There will be write-ups.”

–One PA who gave the standup address in pick called us “boys and girls.” (In my studies, I have been trained to consider the implications of names and labels, so I immediately did a double-take at this, but I really don’t know if anyone else cared or not. At any rate, I can’t imagine that this event conforms to company policy.)

–Another manager in pick said to us regarding being late from break, “I’m not going to send someone out to ask you about barriers, I’m just going to go find you in the mod and we’ll take it from there.”

These are just a few examples. Managers can get burned out too. They can become apathetic, patronizing, threatening, or act out in different ways. It is important that they have the proper training to avoid these pitfalls and can maintain an atmosphere conducive to positive workflow. If I am going out on the floor with a resentment about something my manager did or said, I’m not in full production mode.

I have two managers on my shift for pick. Obviously, it wouldn’t be hard for you to figure out who they are, but that’s not my intention. Anyway, their management styles are like night and day. One is constantly complimenting us, encouraging us, even asking us politely to do our best in a certain area that needs attention that day. I rarely talk to him except for mutual positive feedback and understanding of my job. The other manager is often threatening, patronizing, rarely complimentary.

My point is that managers and PAs need to be well trained, thoughtful individuals. Pan Pan is a great example. He is fairly new, but very well trusted by the employees in AFE. I know he has an MBA, and when I talk to him I think to myself, The guy paid attention in class. He shows real concern and respect for all employees. When I spoke to him about how dismissive and abrasive one of our AFE PAs was, he said to me that I’m right. The PA is really not good with people at all, but the person is really good at the job of keeping AFE running. I accepted that, and the PA and I had a respectful working relationship after that. I believe it would be ideal if all of our area managers had business degrees or any kind of degree. PAs should receive standardized training to understand that they are facilitators, not watchdogs. Sometimes the newest, most inexperienced managers can be the most overbearing and can become an actual barrier to production. I heard countless times from the very top performers in AFE, “I wish (some PA) would just let me do my job.” Of course, I do not want to sound overly critical. We have a great group of Managers and PAs who are great at keeping the wheels of SDF8 turning day in and day out and do so while treating everyone with great respect. I am only emphasizing the need for maintaining standardized training and methods in order remove as many barriers to production efficiency as possible.

As I have been working on this document, I have noted several instances just in the last couple of days  illustrating the above point. I don’t want to beat a dead horse, but I want to help you understand your workforce. For two days in a row during standup, a PA, following instructions, has given new warnings for time-off-task overruns, explaining that if associates go over a half hour, they will get training, or a warning. Over an hour will get some other kind of more severe warning. Some associates were disturbed by this, many don’t exactly know what it means. Someone asked me–I told him I’ve never really understood time-off-task. The first day of these two days, our manager sent us a message on our scanners reminding us that breaks are fifteen minutes, and that our last scan should be at 3:45 and our first scan should be at 4:00. When I went into the break-room and sat down, an associate said to me, “Why don’t they just make breaks two minutes–we can walk in (the break room), look around, and walk out.” I have to admit, I thought this was pretty funny, but it illustrates the kind of attitude a worker can develop from being repeatedly nagged by management. The same day, the same manager sent another message at 5:30 that said something like, Great job today, team, but we need to really push to meet our goal for today. I was already pushing. I always try to finish with a strong fourth quarter. I think most employees try to do their best every day. It’s my personal opinion, that blanket imperatives like the above have either no effect, or the opposite effect of increased production. I know from experience, most people are already hustling. So many times in AFE, a PA or a manager would come on the loudspeaker and say we need to step it up, and everyone around me would already be packing their asses off, and would just say, Really? The effect is a constant cry–whatever you are doing here, it’s never enough;  however diligently you are trying to do your job, you are under daily scrutiny. I may be very wrong, and Amazon’s statistics might prove that I know absolutely nothing about what I’m talking about, but it’s my personal opinion, gathered through my own observations and through the things employees have said to me many times that this kind of atmosphere reduces long and short-term production efficiency.

These are the themes that come up over and over again. People have said to me many times, they treat us like children, and, no matter what we do here, it seems like it’s never enough. Now, you made a point to me, that these might only be perceptions–that in reality, the employees are appreciated, written warnings and firings are far fewer than they might seem. It is the atmosphere that I am talking about. It’s what lowers morale, makes people feel beat up, tired and lethargic, and increases turn-over rates.

As I noted above, many associates do not really understand time-off-task. It’s a vague area even for those who do. It’s vague, because managers say what the requirements are, and enforce another set of requirements that they adhere to. Managers who say breaks are fifteen minutes pick to pick are making a fallacious statement, over and over again, because this unreasonable instruction can never be enforced. Only some managers and PAs say this. I don’t know exactly where and when it started–I think sometime around peak. I asked one of my managers about it, and he said to come after break, and he would show me the break log. I did. I could see the data in spreadsheet form. There were many people who had still not made their first pick. I asked him what these people were doing–were they in the mod, but just hadn’t gotten to their first pick location yet? He said that yes, the majority were. He also explained to me what time off task is, and how it is monitored throughout the shift. His showing me visually and teaching me the significance of  time-off-task  helped me to understand the importance of minimizing it to meet production goals. This helped me to internalize the importance of consistent workflow, so that I can continue to self-regulate my productivity.  It might be desirable to present a visual guide, show employees the break log, the time off task log, be open about what our desires are as a unit, rather than making group threats. Of course, deal with individuals who abuse the system when required.

There is always a contingency of associates who are not self-regulating, who do abuse breaks, who spend too much time-off-task, who do not meet reasonable rate expectations. Believe me, I understand there is a problem. I see people chatting, and then, after I have been across the floor several times and made ten picks, I walk by and they are still there, this time chatting to a different associate.  We should not try to correct these individuals through group meetings or group imperatives. Stand-up meetings should be used for information, education, and positive feedback only. We should be very careful never to vary from this requisite. Many employees who fall within the above contingency and require correction are Integrity employees. Amazon managers and PAs should not have to spend their time and resources monitoring or coaching Integrity employees. This automatically reduces the need for group warnings during stand-up or during shifts by reducing the total number of employees a manager or PA has to track. Of the remaining Amazon associates, those who require correction should be approached individually when they show a pattern of negligent behavior.

Training for new-hires and transfers is important as well. As it stands, training in these areas is somewhat hit or miss, depending on the time of year, department, shift, whatever. If there are standard training manuals for pack or pick or stow (and I suspect there are), I have never seen them. Trainers are sometimes from the learning department, sometimes they are ambassadors, and sometimes they are just someone a PA grabbed to train people for that day. This is all fine–it works well enough using the resources on hand on any given day. I have heard many people say, however, that they don’t feel they were properly trained. And I have seen some trainers who are pretty apathetic.  In AFE, there are plenty of people around to ask if a packer has a question. In the pick-mod, one is often on one’s own. When I was trained for pick, I took notes in a little notebook on the basics like how to reboot, how to log in and out, how to miss-out an item, how to skip and why to skip an item, etc. I would have loved to have had a cheat sheet, a short manual of some kind to refer to. I have been picking for about a month now, and I still wish I had one. Often, managers and PAs talk about procedures during stand-up, but they mention the procedure very quickly, probably only confirming the procedure to those who are already familiar with it, but not teaching the procedure to anyone who isn’t. For instance, when I was trained, my trainer told me if an item is unscanable, to reject it. Then, I think, one of our PAs at stand-up told us to take the item to a problem solve station and call for a new label. I’m still not sure what the correct procedure is for various label problems–I’d have to ask. A cheat-sheet would solve that problem. Cheat-sheets and those little laminates get lost though. Another option would be a help page on the scanner. It could just list procedural options, or it could be presented in an If, then format. A short, printable process manual for each department, I think, would be ideal. A new associate could be given a copy on her first day of training to take home and look over at her will. Copies could be placed in work areas for employee reference if needed. This could take a lot of the guessing out of the pick process, improve quality, and reduce the need for problem solve calls. A short training manual could also contain information to help the associate understand what happens–how the product gets to her, where it goes when she is done–other information that pickers often hear about but that remains vague like tote etiquette and the nature of  pick paths. I believe few pickers really understand these terms. I suspect the more an associate understands about the process and her part in it, the more she will have a sense of ownership and agency in her work.


Nuts and Bolts:

Enough of my bickering about management strategies. Below are some ideas for practical process improvement. SDF8 is already a technological marvel of modern efficiency. Any improvements are made incrementally. We can measure workflow by the second. I know from my experience packing and picking that seconds saved or lost throughout the shift can greatly increase or decrease production rates. Amazon, obviously, is always trying to save seconds.

AFE is already pretty amazing. I only have a few short comments. Guns–what happened to all the guns? I know AFE managers are asking the same thing. When I started there last August, there were plenty of guns. Sometimes during peak, they started getting hard to find. Now, when I go there as shared labor, sometimes I can walk down five walls without being able to find one. I’ve seen employees using wired guns. I’ve dropped my share of guns in the trash and had to fish them out. We should just buy some. Amazon is always excellent about supplying its workforce with all needed materials. Why are we balking at buying a few guns for AFE?

Traffic is a barrier in AFE that I don’t have a solution for. Workers are constantly forced to navigate around each other. The design of AFE necessitates this, but it slows everything down and increases stress in an already stressful production area. It should be minimized if possible. My only suggestion, a very minor one, is to keep water spiders on the back side, not the pack side. Some like to do their job on the pack side, which often interferes with packers. Mostly, it seems, they do this so they can chat.

One more thing–ergonomics. The pack table is low for almost anyone but the very shortest of associates. It places the packer in a slightly hunched-over position for long periods of time. I’m a bit on the tall side. My solution, someone showed me, is to place a stack of flat boxes on my work area like a raised desk. But people of all sizes have remarked about the uncomfortable position. In future iterations, the table could be designed higher, or adjustable.

We could obviously make improvements in the pick department. When I started picking, I was immediately thrilled by how intuitively the pick mods are laid out. (I have only picked in the East Mod.) Within the first day or two, I was finding my way around with very little effort. I could find a pick location, almost without looking at the bin labels. So the layout is very good. I would make one recommendation regarding aisle numbers. In the East Mod, bin numbers always run ascending from north to south. Aisle numbers always run ascending from east to west. The signs at the end of each aisle list the aisle number, and under the aisle number is the range of bins located in that aisle with an arrow pointing towards those bins. On the north end of each aisle, the bin numbers are listed from least to greatest, or left to right, with an arrow pointing south. On the south end of each aisle, the bin numbers are listed backwards, from greatest to least, with an arrow pointing north. The reason for this is, following that arrow north, pickers will be entering the aisle with bins descending in number. This is fine in theory, but unnecessary and distracting. Western learners are always accustomed to seeing a number range written from least to greatest. When entering an aisle from the south, a picker will have no trouble associating the numbers written least to greatest, following the arrow north in a descending path. The picker is already well oriented because of the intuitively laid-out grid. Numbers written backwards require a re-orienting. The picker must register in a miniscule moment each time she enters an aisle from the south, These numbers are written backwards to accommodate for my direction. Then, returning to a north to south path, the picker has to re-orient again, These numbers are written in standard form, from left to right. This is really a minor inconvenience. Some might not agree with me, but I see it as an unwanted precedent.

There are problems unique to the type of picking we do at SDF8. Amazon has a distinctive system for inventorying product units designed to maximize inventory space. It continues to serve our production system very well. Clothing, however (and I understand that management is already considering this carefully), may be a unique product that may require a slightly different approach. Because of the malleable nature of clothing, it tends to cover up or hide other items, and it can create difficulties for pickers. The usual way of stowing and picking solid items, where most are easily visible and can be picked with minimal shuffling around in the bin, is often impossible with clothing. A picker, very often, can scan a location, pull open the bin drawer to find the item sitting right there on top, grab it, throw it in the tote, and go. Other times, the picker is found sitting on the floor, sifting through mounds of similar looking plastic bags, items scattered around her, until she finally finds the right item.

The labeling issue is probably the greatest barrier in the current system of clothing picking. Labeling for clothing is, by nature, inconsistent. Because of clothing marketing techniques, there are no standardized colors, sizes can be hard to recognize, brand-names can be confusing. Each time a picker scans a bin, she has to then try to decipher what her gun is telling her, then try to match that information with the often non-standardized label of the product she needs. This process is complicated by many factors–many look-alike items, missing size information, missing color information, misleading color names, hidden or buried items. One picker called our pick-bins mystery boxes. We already have the LPN label in partial use. The LPN label can be very helpful.  When it is easily visible, it can make product identification simple. But sometimes, for instance, there can be many units of the correct item in a bin, all with unique LPNs. In this case, a picker can see she has the correct item in her hand, but then she has to find the one with the correct LPN. A standardized labeling system with large print, listing brand, item, color, size would be ideal. Implementing such a system would obviously be very difficult, but it is something to consider.

Another option to increase pick efficiency (perhaps at the loss of stow efficiency) is to decrease the number of items in a bin. I believe the standard unique item limit for stowing is eight, although I’m not sure. From a picker’s standpoint, sometimes it seems like there are a comically ridiculous number of clothing items in a bin–that’s when you see pickers sitting on the floor. After a picker picks an item, she has to then re-stow all the items that have been removed or displaced. Items end up being stowed over and over again. This is a lot of extra work.  Smaller bins are often easier to pick from, simply because they contain fewer product units, although the same problems of product identification can occur in these as well. Large, single-product floor bins are very easy to pick from–scan, grab, go. These large floor bins, however, can only work for a small percentage of items in our system.

It is the nature of our inventory/pick system that pickers are often required to change directions, areas, floors several times during a shift. It’s part of what we do. It does come at a cost to efficiency. Our system is a balance between maximizing inventory space and product unit movement efficiency. Switching gears takes time and energy.

Making sure that pick carts are available quickly when a picker begins work or moves to a different floor is important to minimize time lost. Our pick carts are great. They are very light and roll easily. Most of our pick carts are virtually right-hand or left hand models. It would be ideal if all pick carts could be used universally by a right-handed or left-handed employee. I have found a way around this problem by using the water holder in the right-hand position of some carts for a gun holder. They work fine for this as long as the water holder still has the strap. Pickers must rescan the tote every fourth pick. This is a quality assurance measure, but it breaks picker’s work rhythm. Maintaining steady work rhythm is a good way to maintain efficiency. To rescan the tote, a picker must reach around in a somewhat cumbersome position to scan the bar-code on the side. Adding a bar code to the side facing the picker could reduce the break in rhythm, although I’m not sure if this is practical.

Our wireless scanners are not the most sensitive. When I picked at GSI, the scanners used there could scan a bar code from about six feet away. Our scanners seem identical, except they scan from about eight inches away. This might be by design, in order to prevent mis-scans.

Bathroom breaks between scheduled breaks make up a lot of time-off-task. Bringing pick paths closer to bathrooms or bathrooms closer to pick paths might help to reduce this. On a side note regarding bathrooms–with the importance of washing hands in an environment like SDF8, I believe it would be a good idea to remove the flow limiters on the bathroom faucets. They make it tedious and time consuming to wash one’s hands. I hear people comment on this quite often.

On a final note, there is a natural threshold for human work capacity where, no matter how efficient we make the work environment, we are going to see smaller and smaller returns for human production.  People get tired. People don’t always work consistently. They have strong days and weak ones. They have strong quarter periods and weak quarter periods. One manager told me that he can usually see production rates falling on his computer screen at the fourth quarter. As I have said, I usually try to have a strong fourth quarter, but it might be in my mind–I could be lagging as well. And sometimes I’m just tired, and I’m not going to go fast, period. Can we find ways to compensate for these cycles to reverse these trends?

It’s not all gloom and doom. There is a lot of great comradery among our workforce and between our managers and associates. I don’t like to see anyone unhappy at his or her job, so I want to always try to create an environment where we can all flourish and work effectively. I believe these are the same goals we all share. If I can be of service in any way, please let me know.


John Burgett