How To Survive A Death March

Death marches I have done:

- Five months 80+ hours/week on my first video game

- Nine months of similarly insane hours shipping a consumer PC platform

- Well over a year of crazy on the Apple Newton, including six back-to-back 100+ hour weeks

- A couple more for Microsoft, on various things, including a few of those century weeks (my wife called me "broken" on the last one)

In each case it took me 3+ months to recover, probably more.

Last year, faced with the "opportunity" to do another one, I resigned. By all accounts I made the right choice.

A death march is a sign that something is seriously no-kidding broken in planning and management. It may be necessary to actually do a DM to recover from the situation you've dug yourself into, but organizations should learn from this.

But instead of learning, companies come to depend on them. The expectation develops that the technical staff can be leaned on to rescue bad decisions. One of my ex-cow-orkers likened it to an addiction.

The price? Loss of good talent, bad morale that leads to mistrust, even poorer decisions, a sense of powerlessness at the worker-bee level, and lack of continuity in follow-on products because of turnover.

I will say that I've made fast friends and learned a hell of a lot about people during DMs. But I'd much rather work in a place that does a DM, learns from it and adapts.

So, I think I can give some much shorter and better advice for how to survive a death march, in the form of a flowchart:

  1. Do you excel at your job?

    - Yes: Ask around and get a job somewhere with management that isn't shit.

    - No: Change careers, goto 1.

Although I appreciate, and read with interest, articles such as "how to survive in a POW camp" or "How to survive in prison", I don't really see why you would need that advice if POW camps and prisons were voluntary and you could freely leave.

(The reason I say change careers if you are not good at your job: Good management will be able to tell you aren't good and they won't hire you, or if you do get in the door it won't last. Being bad at something means your entire experience of work is necessarily at the kind of place that hires bad people, which is not how you want to spend your short experience of being alive)

I'll vouch for everything in this article [1] and add: if doing this, sleep consistently every day.

If you're in the middle of something, just drop it and go to sleep. If you've been working on something all week, you're not going to forget something or lose your place or indeed, if done right [2], even lose your 'flow' by sleeping.

Getting 80-100 useful hours out of an 168 hour week can only come if you use a good chunk of the other 68+ hours sleeping - but even moreso than getting enough hours of sleep, getting them at roughly the same time every day can make a big difference.

[1]: Having learnt most of this the hard way.

[2]: As inadvisable as it might sound, working immediately before bed is the only way I've found that ensures this.

I agree with and can vouch for every one of these points - particularly when working is isolation.

Thing is, every death march I've been on has been a group effort where appearing to suffer for whichever executive dropped the ball at the onset is at least as important as actually producing. In these situations, appearing healthy much less taking breaks will be taken as a lack of effort.

The guide I could really use is one on how to manage "up" and limit the effects of or outright avoid working for the panderers and incompetents who create death marches in the first place.

People repeatedly assert, without ever citing anything, that 40 hours/week is some theoretical max on your productivity. Can anyone provide some actual citations on this?

Robin Hanson did some investigation and found little empirical support, and his data put 60 hours as the optimum (for construction, not knowledge work).

http://www.overcomingbias.com/2011/12/work-hour-skepticism.h...

http://www.overcomingbias.com/2011/12/construction-peak-60hr...

Rule 0: don't participate in death marches. Seriously. Death marches mean that management has no idea what it's doing and you would be better off quitting and getting another job. Then again, I am an old and bitter programmer. :/

Today I completed a difficult five days and the blasted thing works and I hope I never see this code again. Before you laugh because you've all pulled off 9 months without sleep or whatever manly chest beating, hear me out, because its all fresh in my mind (obviously, with successful completion being today).

1) WRT athlete, your brain is overworked. The most important thing leading to the quick(ish) resolution of the redesign was doing absolutely nothing on Friday. If I hadn't done that, it would probably be another week... or two... until I got this thing done. After the first ten hours on Wednesday and Thursday I was making, in retrospect, provable negative progress at the rate of about five hours of re-work required per hour of work.

2) Never deviate from procedure. When you're sleepy, cranky, panic, and tired, skipping unit tests or a sort of individual code review is an excellent way to make negative ten hours progress. If you know how to do it the right way, do it that way even if its not immediately the fastest. I've never wasted so much time as when I've tried to speed things up (this applies to general life, now that I think about it).

3) Explain the whole thing as if to a noob or a coworker, over and over, until you find the bug. So this case class inserts into that db table using ... uh.. that code that doesn't exist. Yeah. (Edited to emphasize "the whole thing" means to find and fix major system design issues... the compiler will do a fine job of finding missing semicolons, but entirely missing flowchart boxes can only be found via systems analysis, and this is probably not the ideal time for it)

I can mathematically prove that working 13+ hours a day results in about as much net forward progress as working 4 hours per day. I would probably have been done on Friday if I had just stopped early on Wednesday and Thursday, but instead it took till Sunday afternoon. People who claim to work 90+ hours are probably doing the mental equivalent of ditch digging or are only getting net 10+ hours per week forward progress.

Back to easy, much more productive, shorter hours next week...

It's not necessarily a very good idea (although death marches, used judiciously, do work)

No they don't.

Or that is: they sometimes work for your boss (or your customer), in the short run. But they don't work for you.

And when push comes to shove, that's who you need to look for, right?

There's no reason, other than a feeling that you're expiclity under the client (or employer's) thumb -- e.g. a short-term financial emergency; or you're feeling very early in your career, and that you can live without a shining reference from them, after you've made your way through the tunnel of pain they've laid out for you; or you just didn't vet the other party properly, and weren't aware of their manipulate (and perhaps sociopathic tendencies) -- to put up with death marches, ever.

All you have to do (absent one of the above-mentioned emergencies) is say, "I'm sorry about the external pressures you're facing, but asking me to put in these hours this doesn't seem to be in line with the parameters we set up at the outset for this project/job. It's also just not going to be good for your working relationship, in the longer haul. So either you're going to need to find additional resources to help you, or I'm going to have to pass."

It's not easy. But it's doable. And trust me, you'll respect yourself a lot more in the morning.

For further guidance and inspiration on this topic, please see:

   https://en.wikipedia.org/wiki/50_Ways_to_Leave_Your_Lover

"The answer is easy if you take it logical-ly..."

Good advice. I work with groups (LSAT and SAT students) who almost inevitably end up in death marches in the weeks leading up to their exams. They need to be told this.

I would add: delegate those things in your life that are work, but aren't contributing to the objective of the death march. For instance, I work from home, and certain administrative tasks are a thorn in my side. I now have someone cleaning my place, and scanning documents, two of my least favourite jobs.

This has increased the hours I can work in a week without feeling worn out. This stuff can also be scaled up in a death march. I once spent a month writing, and produced 50% more than my normal output.

But I felt more relaxed than normal. How? I was staying somewhere where the hosts cooked, cleaned, and ran errands for me. I had zero responsibilties apart from work. I think I worked 60 hours a week for a month, but I cut out about 20 hours a week of tedious chores, and so I didn't notice the difference.

My advice would be: starting dead march anytime sooner then last week and half is irrational. I do not believe authors estimate of three weeks 80 hours, that is too long too much. It may feel good, but it is not a good idea. First, you need to keep clean head and think eg.) * prioritize and cut everything non-essential, * being organized is not waste of time, but do not overdo it, * prioritize again and cut everything not super essential, * sleep enough hours, * eat regularly and real food, * take short break after every longer subtask (!short - measure it!), * be conscious: if you find yourself staring at the screen or procrastinating, then it is good time for short break, * use the above breaks for short but intensive physical activity, * socialize less then usually, but find some time to socialize anyway.

Important: * have good testers and let them check for regressions as often as possible. Really, they are more valuable then new programmers on team.

If you need to achieve something exceptional, you need to think clear above all. The idea "wait we do not have to do it at all" saves more time then all clever hacking in the world. And if you do not sleep and do not eat, you will not be clever anyway.

As for last point, most people need to talk. If they do not get to talk out of work, they tend to get more talkative in work. That means a lot of wasted time. The last thing you want is to unconsciously use team meetings to satisfy your communication needs and waste everybody's time.

It is ok to bundle things together eg. run to grocery store for food (3 in 1: physical activity and break and food).

Pretty sensible advice for dealing with long hours over an extended period, but I was confused by the use of the term "death march" which in terms of software development is known as an anti-pattern:

http://en.wikipedia.org/wiki/Death_march_(project_management...

"But you can work 80 hours a week for the next three weeks and double what you get done"

That seems rather optimistic, since your productivity and focus will probably drop sharply after about 40-50 hours. So you're probably lucky if you're getting 1.25 or 1.5 times more done by working twice as many hours, and sometimes your total productivity will even decline.

For example, after 50 hours, a developer might be coding more bugs that they'll have to spend time fixing the next day. If it takes longer to fix the bug than it did to write that code, the productivity of the time spent writing the code will have been negative - you would have been better off going home and getting some sleep.

Currently doing a death march. Not sure how to deal with it once it's done.