Give up working overtime, but it's not worth the candle

In fact, as early as 75, most of the industry has given up to work overtime to catch up. and countless industry experiences and research facts have proved that if you want to get the job done, working overtime is the most expensive thing to do.

Origin

In 2004, the family of an international video game company, in the name of Ea_spouse, posted an article on a Web site and described how their spouses had a bad impact on their physical health and family life because of their high strength and long overtime. A stone stirred the waves, there is about the development of electronic game industry People's quality of life has aroused the enthusiastic attention and discussion. Ea_spouse received thousands of responses, and it was quickly followed by major media reports. Over the internet, thousands of people have been involved in a spontaneous, large-scale discussion involving forced overtime, work efficiency, laziness, unions, lawsuits and complaints against many companies.

I have now done more than 20 years of software project development and management work. Every project I've done in the past year has made me more convinced that overtime--"rush"--is a costly, unproductive and destructive way of working. "The longer the work lasts, the lower the productivity" is common sense. Over time, however, I have noticed that the speed at which productivity declines due to excessive extra overtime is more than most software industry managers perceive. As the investigation went deeper, I was surprised to find that I was not the first person to recognize this: For nearly a century, engineers in traditional industries have reached a consensus on what I have observed.

History

Abbe, a pioneer in industrial efficiency research in 1908, released his findings that the daily output would increase from nine hours to eight hours a day. And he wasn't the first person to notice it. As early as 1893, William Mather in Salford Steel Company (Salford Iron Works) has adopted eight hours a day work system. In 1909, Sidney J. Chapman published the Work Practice (Hours of Labour). He describes the changes in the number of hours worked per day and the productivity of workers. Its conclusions are elaborated below. Henry Ford, which adopted a 40-hour workweek in 1926, had at least 12 years of experimentation to convince him that the 10-hour daily work would be adjusted to 8 hours, and six days of work per week would actually increase the total output and reduce production costs. Ford has touted the social benefits of shortening working hours and stressed that it is good for everyone to spend more time. But the central point remains that reducing shift time can lead to more output.

I found that many organizations (such as business, universities, industrial associations and military units, etc.) carried out the same basic point of view, that for most people, "work eight hours a day, work five days a week" is to ensure high output and low physical and mental consumption between the best balance point. Similar studies have been done hundreds of times in the 30 and 50 of 19th century, and the benefits of a 40-hour work system in the 60 have been indisputably accepted by the American business community. In 1962, the Business Association (Chamber of Commerce) even published a small brochure to compliment the productivity gains from reduced working hours.

But, somehow, Silicon Valley has left it behind. Ea_spouse wrote:

"Now the mandatory working hours from nine to 10 o'clock in the morning, and seven days a week, based on the good performance of the staff, the Saturday hours of work occasionally advance to the evening 6:30." On average, this is equivalent to working 85 hours a week. ”

In fact, there are six days in the week from nine in the morning to 10 o'clock in the evening, and one day from nine to 6:30 in the morning to work 87.5 hours a week (6x13+9.5=87.5). But after such a long period of work, who will be so detailed?

In this respect, the Electronic Arts sector is no different from other hi-tech companies. Look for people who want to improve their productivity and keep them sane together: the assumptions that managers make in terms of working hours, output, efficiency, and cost of production, and how a century of industry research has proved that these assumptions are completely wrong.

What does the manager want?

What exactly do they want when management sends their employees to the "Death trip"? Do we really believe that the CEO of EA wants to see employees running for 7x24 hours in the office with their butt in their pants?

Managers want to get the biggest output from their employees, with the least amount of cost involved in producing the best products. They do not want to have to pay for the extra human resources to complete the product, thereby increasing the cost. On the surface, to achieve these two goals, "Rush" seems to be the most obvious and reasonable way.

Assuming that output is measured by piecework, managers who have not read the above report may infer that if a person produces 16 products within eight hours, he should be able to produce 18 products within nine hours, and possibly 10 in 20 hours. We can use the following simple formula to express this view:

O = y * t

Where o is the total output, X is the output within the base time Y (measured in hours), and T is the actual number of hours worked. Under this assumption, increasing the time t is the simplest way to increase the output o. In individual cases, this assumption is valid, for example, to extend the working hours only for a short period of time so that the deadlines can be delivered. However, long-term tests and studies in other industries show that the limit duration of overtime sprints is shorter than most people think, and that when the limit is reached, the punch will be in trouble.

Hour productivity is important

A more realistic way to look at the "Worker's output" is to consider whether the hourly rate of productivity changes with the length of working hours. There are two main sources of this change: mental and physical fatigue in the last few hours of the day, and a build-up of mental and physical fatigue as the working days that have extended work hours continue to increase.

The following equation expresses this more complicated situation:

O = P (T1, T2, T3, ..., TN)

Where, o represents the total output, P () represents the change in hourly productivity over time (T1-TN). In this equation, P () is a function, not a constant. P () varies according to the worker, because some people are more productive than others. P () also varies over time because the person is not a machine and the work done in the first 14 hours is not exactly equal to the work done in the first 1 hours. In addition, P () also varies with the worker's recent status, such as the morning after the Midnight Oil and the morning after a good night's sleep.

The following diagram is shown in the "Working hours" published by Sidney J. Chapman in 1909:

Where the curve p represents "a fixed number of labor (according to the number of hours per workday) output marginal value of long-term change." The ox axis represents the working time of the day, and the Oy axis represents the value of the output. On the n point of the ox axis, that is, if you work n hours a day, its gross output is the area of the graphical Onda (see http://www.worklessparty.org/timework/chapman.htm for details). It can be found that the height of the curve p represents the worker's productivity (the output per unit time per day for a given number of hours of work).

Smart readers have noticed that at point B, working more hours does not create more value. And after point B, every one hours of work, output is negative. How could that be?

Chapman's work graph assumes that the working day of a given number of hours will last a certain amount of time. Therefore, it will be the daily fatigue status and long-term fatigue cumulative unity within a model. First, the decrease in the hourly output reflects the effect of fatigue on the quality and quantity of work at the end of the close day. But then, daily fatigue is compounded by cumulative fatigue. In other words, the productivity degradation resulting from fatigue during the first day of overtime can have a superposition effect on the hours of production later in the day.

Even within a day, when exhausted employees are no longer able to work, output will stagnate. If an employee who has become numb has made a disastrous mistake, destroying the work that has already been done and the capital invested, the output will become negative.

In the case of factories, the productivity of a worker declines over time. A worker may produce 10 jobs per hour at the start of a shift, while close to work may produce only 6 per hour, and in the meantime a few hours may reach a peak of 12. In the future, the work will be slower and will make more mistakes. This deceleration and error eventually makes productivity 0, which takes a long time to produce an artifact, and at the end of the day there is always something wrong. Managers of the assembly line have long discovered that when this level of fatigue is reached, there will be large failures and more costly losses, such as costly machine damage, inventory destruction, or serious injury to workers.

As a mental worker, after a full break, programmers produce more high-quality code and fewer bugs. At the beginning of the first hour, the programmer will gradually enter the state, the next few hours is the best state. After that, we feel tired and our hourly productivity declines; we take a long time to fix a simple bug or add a simple feature. And if it gets done in the first few hours, it could be a matter of minutes. A bit worse-it seems that many companies in the electronics entertainment industry are working in this extreme situation most of the time-an overworked it technician may delete valuable files, it will take extra work to restore the backup; he may also have a traffic accident on his way home and will not be able to work for months.

This is the first lesson: In a business day, productivity changes with time H. In the first 4-6 hours, the production efficiency is the highest. Over time, productivity will fall to 0, or even become negative.

Where is the balance point?

If productivity is going to fall over time within a business day, and long hours can lead to low productivity, how can we find a way to achieve maximum output, and where is the balance?

Unfortunately, it is very difficult to quantify the output of the mental workers. I would also like to give a simple formula that calculates the number of hours a person needs to work to achieve the maximum output by substituting a few numbers. But I can't, because even with such a formula, it's not possible to agree on which basic data to find and substitute. Common software metrics, such as code line or function point method, either simply collect numbers, do not convince people of their value, or it is difficult to define and collect data. Useful metrics, such as the number of bugs and the number of bugs to fix, are less reliable and may be unfairly used for annual reviews, and may be used by smart programmers to calculate year-end appraisals or performance bonuses. The architect's output can easily be measured by some data, such as the number of models or frame compositions, but it is also difficult to measure with other data (such as subjective quality, perception, and model complexity). It's easy to measure the output of a tester with the unique number of bugs found, but it's a little tricky to measure with code coverage, which is even harder to measure with the total percentage of defects found.

In short, most companies seem to fall into a "least common denominator" metric for team output. Either the factory quantity and the sales volume of the game, or not. While these are indeed the metrics that most shareholders care about, the measurement of productivity is totally useless, especially the daily or hourly productivity.

Lesson two: For mental workers, productivity is difficult to quantify.

So we have to make an analogy with other industries: The following comes from work pager Cato of Marvell, an article in the "Mental physics of the IT industry," which responds to Ea_spouse:

"It was more than a century ago that Dr. Ernst Abbe the work hours and outputs of the Zeiss optics factory (Zeiss Optical works) in Jena, Germany. Dr. Abbe, a director of the plant, reduced working hours from 9 hours a day to 8 hours and detailed the daily output of each worker before and after the adjustment, and the results were consistent with other studies in 19th century, in which the appropriate reduction in working hours did increase the overall output. ”

Tom Walker points out:

"The increase in output is not proportional to the length of working hours" as if it were a compulsory course for every generation. The 10-hour Work Act was passed by the British Parliament in 1848, resulting in an increase in total daily output per person. By the 1890s, employers had been experimenting with a 8-hour work system and were discovering that the total output of each worker was still rising. Frederick, founder of "Scientific Management theory", points out that reducing working hours can significantly increase personal output.

In the 20 's, Henry Ford worked on the schedule for years, eventually introducing a 40-hour workweek in 1926 and a six-day salary. Why would Ford do that? Because his experiments show that its plants produce more than six days of output in five days. And at every step of the shift in work (1840s,1890s and 1920s), there are some sectors that insist that shorter working hours will reduce output and damage the economy.

Lesson Three: After a century of research, 8 hours of work per week, Five days a day, will produce the biggest output in the long run. What is the reason for us to think that our industry can not abide by this rule?

What about short-term output?

If the 40-hour work week is the best way to schedule the most output, can we expect short-term results for "extended working hours in the short term"? In short, from a few days to a few months, how much extra output you can get by working longer hours depends on how long a workday is working.

Obviously, if a worker produces an artifact every hour under the eight-hour system, the number of artifacts produced by the 16-hour work will be between eight and 16. This is hidden in the "rush" behind the less obvious nature. In addition, workers ' productivity depends on their status.

Working 60 hours a week compared to working 40 hours a week can lead to lower productivity. In the beginning, this extra 20 hours will make up for the decline in productivity and increase the total output. But studies have shown that the productivity of construction workers will soon begin to decline when the work is changed to 60 hours a week. This decline can soon be felt, within a week, will be obvious, and will continue to decline. Within two months, the cumulative loss of productivity will fall to the same level as the output of 40 hours of work a week. The study cited in the same report showed that the total output of working eight hours a day was higher than 16% or 20% per cent of work nine hours a day.

So, yes, "rush" can boost output in the short term. But in the case of working 60 hours a week, this "short term" must not exceed eight weeks. Because from this point on, the cost will be more serious than the benefits. Will not only lose the results of the rush, but also make the staff feel tired, irritable, emotional difficult to control. When the recovery is 40 hours a week, it will take some time for their output to return to their original level.

What happens if you work 87.5 hours a week? Despite the lack of conclusive data, I estimate that efficiency will fall to 50% in one months, even if the extra 47.5 hours of work per week (twice times the "normal" working hours) are likely to have a fairly high output at the initial stage.

Lesson Four: When working 60 hours a week, productivity declines due to prolonged work can offset the output of months of overtime work.

Sleep factors

Another factor to consider when assessing the usefulness of "rush" is: how long can a worker stay productive if they don't get enough sleep?

Colonel Gregory Belenky is director of the Neuro Psychiatric Department of Walter Reed Army Institute. He worked for the Pentagon on how to maximize the efficiency and vigilance of soldiers in combat conditions. In his 1997 paper, "Continuous combat sleep, sleep deprivation and human performance," he said:

"In-lab studies have shown that the sanity function drops by 25% per 24 hours of continuous sobriety," he said. While the individuals who were deprived of sleep were able to maintain the accuracy of cognitive activity, the rate of reaction decreased markedly. ”

In our study, the team from the 82nd Airborne Fire-direction Center received a test that simulated the ongoing combat status for 36 hours. At the beginning of the test, when we asked for a mock fire at a hospital, the team was able to view the azimuth map, assess the target status, and refuse to fire the request. To the end of the simulation test, they ignored the target and did not hesitate to fire.

On the 15th day of the simulation exercise, the four-hour-per-night battery was only one-third of the seven-hour battery per night.

Lesson Five: Every 24 hours of continuous work, cognitive ability will decrease by 25%. People who burn the midnight Oil have a serious cumulative effect.

Cognitive retardation and Error rate

One of the most important factors in the reduction of productivity caused by "rush" is the increase in the number of errors. While most bugs are easy to fix, there are some mistakes that will deplete the gains from "Rush". The longer the "Rush" time, the greater the opportunity for the person concerned to get into trouble.

Programmers, architects, and testers get paid, not because they have well-developed muscles or what powers can move heavy weights from one point to another, but because of their brains. The longer you work, or the lack of adequate sleep (like sleeping only 1-2 hours a night), it can significantly reduce their efficiency in using the brain.

Colonel Belenky pointed out that, even if the soldiers sleep less than one hours a day, the consequence is "reduced ... The ability to maintain a sober and rational mental work. Reduce individual and team efficiency ".

Knowledge workers should feel lucky that they don't have to worry about "friendly fire" problems.

In the article "sustained sleep deprivation can have serious adverse consequences", it is mentioned:

Researchers at the University of Pennsylvania found that subjects who slept only 4-6 hours a day for 14 consecutive days showed a significant deficiency in cognitive performance, equivalent to three consecutive days without sleep. However, the subjects said they felt a little sleepy and did not realize how bad their situation was. ”

In the Los Angeles Times in January 2005, Karen Kaplan an article called Sleepy Medical Intern as a road killer:

"The study showed that drivers who did not sleep within 21 hours had a status equivalent to 0.08 of the amount of alcohol in the blood, and 0.08 were the legal limits on whether the blood alcohol content of a non-profit driver was exceeded in the US. ”

Ironically, most software companies dismiss a person who drinks at work hours without hesitating to hand over the most important item of the year to a lack of sleep (the equivalent of a legal driver's alcohol level). In fact, they are asking these people to work in "illegal and drunken conditions", as a condition for employees to continue to hire.

The risks are real--the mistakes that are caused can really lead to disaster. In the Dr. William Dement wrote the Promise of sleep:

"The night of March 24, 1989 is cold and serene, and the air is crystal clear." Exxon's tanker left the city of Valdez, Alaska, to Prince William's Bay. In such clear weather conditions, the tanker put down the pipeline as planned, but did not recover in time. A huge tanker ran aground and millions of of gallons of oil spilled into the Gulf. ...... In the final report, the National Transportation Safety Board (NTSB) found that lack of sleep was a direct cause of the incident. ...... The person directly responsible for the most serious oil spill in American history was the third, who slept only 6 hours in the first 48 hours of the incident and was severely deprived of sleep. ”

Exxon Valdez Prince William Gulf oil spill

"It is problematic to make launch decisions at critical moments," The Rogers Board of Inquiry (Rogers Commission) said in a final report on the US space Shuttle Challenger crash. The "Human Factors Analysis" section mentions that sleep deprivation "has a significant impact on this".

If the lack of sleep can lead to combat failure, endanger the patient, stranded tanker, detonated spacecraft; think about what it would bring for a 15 million dollar game project.

The sixth lesson: the error rate will increase with continuous working time, especially when there is insufficient sleep time. In the end, failure will come to the door, leading to disaster. Can you really afford the risk when the time is tight and the budget is big?

What does that mean?

means lower productivity. Workers can maintain productivity by maintaining a total of about 40 hours of working time per week in Five. Work longer and productivity begins to fall. At some point in time between four days and two months, the benefits from overtime work are offset by a drop in the hourly rate of production. In some extreme cases (when workers cannot guarantee at least 7-8 hours of sleep per night, within a day or so), efficiency will plummet.

Many of the above research comes from the industrial environment, and some might argue that these conclusions do not apply to programmers, architects and testers who use more brainpower because they are different from ordinary physical workers. In fact, indeed, it is different, Colonel Belenky clearly stated:

"Compared with complex mental activities, it can be said that simple mental activity, physical work and endurance are largely unaffected by lack of sleep." ”

Mental workers who need to complete complex tasks are less affected by sleep deprivation than manual workers, and productivity declines faster. In knowledge workers, productivity losses due to overwork are quicker than the average soldier, because our work is more affected by mental fatigue.

Ea_spouse to tell us that her husband's team is far less productive than the best. They have been working more than 60 hours a week for a few months before their bosses let them do the super "rush" of 87.5 hours a week.

In the 20th century, for most of the time, place and industry, managers who worked for their employees were considered unfit for their jobs. This is not just because they are threatening good employment relationships, but also because of their misguided management of the company's productivity and assets as dangerous.

More than 100 years of industry research has proved beyond doubt that the mistakes that employees make as a result of exhaustion delay planning, damage equipment, increase costs, reduce product quality and ultimately threaten the organization's survival. This is a threat to the project and to its managers, employers, everyone, and itself.

In any case, it is economically infeasible to use "rush" as a long-term strategy. Delayed work does not increase output unless it is a short-term activity. In addition, "Rush" can not make the product launch faster, will only cause the product delay release. "Rush" can't improve product quality, only make it worse. "Rush" increases the chance of causing significant errors, such as delivering software that erases the customer's hard drive data, or removing the code tree, or spilling the cola to a server that has not been backed up recently, or even causing a fire. Indeed, I have witnessed the first three consequences in the last sleepy days of "rush". The fourth consequence will happen sooner or later, probably just a matter of time.

The managers decided to rush the work because they wanted to tell their boss "I've done what I can". They are rushing because they are evaluating "straw" on chairs rather than "brains" that can develop games. They rush because they don't seriously think about what they want to do, or who don't think about the job. They rush because they only know how important it is to show that they are doing their job well, rather than actually doing the job well. They rush because they are asked to do so when they are programmers, testers, "Assistant producers" or "deputy producers".

But that's not the only way. In fact, many of the literature has shown again and again that overtime is the worst way to work. This is also the root cause of the way in which many industries have abandoned their work 75 years ago. Managers, shareholders and employees are convinced that using a time-tested 8-hour-a-day, 5-day-a-week-management practice will benefit from quicker and more provincial delivery of better products without wasting organizational human resources and public prestige.

Evan Robinson19 into the gaming industry in TSR as a programmer. 22 years old, he has been playing computer games for EA as an independent programmer. For the next more than 20 years, he has worked as a programmer, senior engineer, technical director, engineering supervisor, process consultant, and technical specialist in several well-known enterprises in the industry. Evan's publication, as well as his early GDC speech, established his position as one of the best software engineering practices in the industry and the most valuable developer management. He lives in Vancouver at present.