Lazy thinking. Although deadlines should be avoided as much as possible and should be viewed realistically (shit happens), lots of things that depend upon the shipped product need to be planned well ahead of time, and can't easily be changed.
Such deadlines aren't there to annoy engineers, they are there because a lot of things need to be coordinated to happen at exactly the same time and place. Software is a means to an end, rarely a goal in itself.
Also, the "it ships when it ships" leads to an attitude in which nothing ships at all. Even if it's no hard deadline, people need a target in which they can frame their effort, especially if it's a joint effort. Nothing is more frustrating than really wanting to get something finished but your coworkers on who your work depends want to do something else. Without a fixed timeframe as a joint reference, it's very hard to stay in sync and make joint progress. I've seen it kill productivity more than unrealistic top-down deadlines.
Finally, there's no excuse for not making the effort to get better at estimating. I've seen well gelled teams play planning poker, give exactly the same estimates independently and hit their targets perfectly. It's not an exact science, but it's not dark magic or wild-ass guesswork either. It's a skill you can hone, especially as a team.
I'll go one step further. Not only are deadlines necessary, but they should be announced to customers and stakeholders.
Being able to say, "We are going to release version 2.0 on June 24th, 2013" does a lot of things. First, it signals confidence in your ability to manage your time and resources. Second, it allows your customers to hold you accountable for missed deadlines. Third, it prevents cases where you con yourself out of meeting the deadline with seemingly valid reasons.
I refuse to buy products from companies who do not publicize their development roadmap. I'll give games as an example. Back in the day I used to play World of Warcraft, and whenever we asked Blizzard when the next patch or expansion was, they would reply with a flippant and condescending "when it's ready." This was such a common response that players would often use it mockingly and append it with a "TM", because it was a trademark of how Blizzard communicated with its customers. It was one of the main reasons why I canceled my subscription.
These days I play Planetside 2, and SOE is a lot better with communication and they go so far as to publish a list of scheduled features and their estimated timeframes. And if schedules change and something gets postponed, they update the list with an explanation of why that happened.
As a customer, I want to know what features to expect and when to expect them. Do not try to hide it from me because that will piss me off. The end.
I also know many Blizzard customers who appreciate that about them because they know the quality won't be abysmal because they rushed to meet a deadline - deadlines and granular communication of roadmaps are less important to them than the quality of their IP.
I think they have deadlines internally, but prefer not to publicize them.
There seems to be no incentive in doing so. What's better for PR in case of missed deadline, imply that you will only release game when it's good enough, or make apologies? Do you get any reward for non-missed deadline?
I can think of several multi-billion-dollar customers that schedule their product releases to the day, quarters in advance, with untold advertising / validation / manufacturing commitments on the line. If your prototype isn't on their doorstep ON THE DAY that you promised, a competitor is selected to take your place. NO BULLSHITTING AROUND. MAKE A SCHEDULE, ACCOUNT FOR THE RISKS, AND EXECUTE.
I can't believe this blog post is even allowed to exist.
When you're presented with a deadline, the first question that you should ask is what type of deadline you have. Here are some common possibilities:
- A number to use to whip people to work harder. (Yes, despite much research demonstrating how counterproductive that is for developers, a lot of management teams still try it. In part because big audacious goals tend to work well in other fields, like marketing.)
- A hard date that must be met. You need something to announce at a conference. A new tax law is in place. And so on.
- A date for the rest of the company knows what to plan around.
How you should treat a deadline varies by the type. If it is one that exists to whip you, ignore it as you file your resume looking for somewhere better. If it exists because something needs to be done, you should work hard to make sure that there will be a minimum viable product available before that date, then improve it as much as it can be improved for that date. If it exists for planning purposes, then make sure it has sufficient padding, and develop as efficiently as you can. And so on.
Now that you know what kind of deadline you have, where did it come from? In most organizations, the schedule estimate arrives as the first date that nobody can disprove. That is why software usually takes about 2x as long to deliver (if it ever is delivered) as it was estimated. But research indicates that, for software developers, the more confidence that they have in the schedule, the more productive they are. Many have noticed that when developers work to a schedule estimate that they produced, developers are more productive than if it is a schedule made up by management. However if developers work to a schedule estimate produced by an expert estimator, productivity goes up again.
Most organizations do not have expert estimators. So involve developers in schedule estimation. Add in fudge factors for sick time, unplanned obstacles, etc. As a good approximation the ratio between past plans and past performance is a good ratio to apply. As the project goes on, if the schedule starts to slip, immediately assume that all later parts will slip by the same ratio that you're experiencing already. To be able to catch slips, you need unambiguous milestones that can't be fudged. Otherwise people will say things like, "Coding is 90% complete" having pushed off the hard 10% that will take as long as what is done.
And if you want more good advice, I highly recommend Steve McConnell's Software Estimation. (You could substitute most of his books for that one, and the recommendation would stand. But that's the one on estimating schedules.)
"A hard date that must be met. You need something to announce at a conference. A new tax law is in place. And so on." that's the only event I would label as a deadline, and these real deadlines don't happen often. The rest are artificial and could be goals rather than stressful deadlines...
The challenge is that a date for the rest of the company to plan around has a tendency to become a hard date once the rest of the company has planned around it. This frequently has real impacts on everyone else.
Also the frequency of real deadlines depends on what you do. My neighbor is a project manager at JPL. They have a lot of deadlines and all are hard - you're either ready for the rocket launch or you're not.
Agreed, we need to get advertising booked months in advance, we need to source materials 16 weeks in advance, book production two months, etc. If you aren't dealing in serious marketing and tangible products then you can release whenever you like!
Deadlines are not artificial. If you screw around long enough, various real world things happen, such as your competition leaving you in the dust, your product becoming irrelevant, customers not caring any more whether the problem is fixed because they've gone to a competitor. There are engineering realities, such as it won't be fixed until it's fixed. But there are business realities, too.
A formative moment for me was when I was the software manager for a company, and we had two clients with emergencies going on at the same time. The reality was, we couldn't really fix both problems at the same time. I went in to the president's office, and told her that we couldn't fix both problems right now, and I needed her to decide which client was going to suffer. She yelled at me, told me to get the hell out of her office, and figure out a solution that doesn't force either client to be the loser. I fumed and left, she obviously didn't understand the reality. But once I finished pouting, I realized there was a way I could implement a temporary fix for one real quick, and then work on the other...
There are business realities, and there are engineering realities. A business needs to manage both to deliver something awesome, or get overtaken by somebody else that will.
The one thing I would give as a benefit to setting a deadline is that it can help prevent feature creep for a specific version. The amount of times that I have had a manager come to me just prior to a version being ready to ship and ask me to squeeze a new feature into this release is staggering. This is where being able to push back with "not if we are going to hit the release date." really helps.
Now ideally you are able to agree on what to put in each release and stick to it, but a lot of work places are not ideal and feature creep can be a huge issue which can easily be addressed by implementing a release schedule and sticking to it (for the most part).
This is more applicable to areas such as mobile where it is unrealistic to ship releases every week.
Yep. I thought the football analogy was pretty good though. Many times i've worked for people that were only good at communicating the (always urgent) deadlines but not the overall objective of the product we were making, which was in no way very motivating or engaging. I think this is probably one reason for the low job retention we experience in this field.
"But is this even a real problem? No, not really."
If you have that luxury then that's great. For you. But if you're doing client work you likely don't have the luxury of never having to estimate. And even if you're not doing client work then it would likely be a good idea to at least give it a passing consideration.
Say you have a client who wants [x] and it needs to be done by a certain date (Yes! This does happen!), then it is helpful for you to know whether or not you might be able to deliver.
Say you have a client who wants [x] for [money]. Then it is helpful for you to know whether or not you might be able to produce it in a time frame which makes sense economically.
Say you have two ways of solving a certain problem. Then it is helpful for you to know if one solution would take a week and the other a year.
Estimation can be hard, sometimes impossible. But I'm not sure pretending it's not a problem is a viable solution for many.
My boss is pretty cool about this. He asks for estimates on tasks for resource management purposes, but he assumes the estimates are wrong out of the gate and we adjust as the project progresses. So, I agree that focusing on an arbitrary deadline is foolish, but drawing a line in the sand can help keep people on task.
So I believe you are wrong on thinking it is unnecessary stress, based upon some research done by Dan Ariely on students getting their work done with and without deadlines. You have to read his book Predictably Irrational for the full story, but here's a synopsis: http://danariely.com/2010/08/30/back-to-school-2/
Long story short is that without deadlines people tend to push off work to the latest possible moment, at which point they tend not to be able to do as much as they thought they could in that short time space, so it actually causes more stresses on them than it would have to have had a series of deadlines along the way.
College students actually. And I'd agree somewhat, although I'd say you could easily extrapolate it to newly hired developers fresh out of college, and that it may be extensible beyond that. Either way, research points against your conclusion and you have no factual basis for it, so it'd be good to find some support.
Edit: Here's some more research which backs up why your conclusion may not be correct: http://www.ncbi.nlm.nih.gov/pmc/articles/PMC3374478/ Granted it is probably too early in the research of this area to say your conclusions are wrong, just that the research is leaning that way
That research lines up pretty well with my own experience in college. It haunted me until my 3rd year when I was so bogged down with advanced projects that I just couldn't keep procrastinating because I was only barely getting the minimum done I needed to in the last hours I had before the projects were due.
Professionally, I've learned my lesson and while I've had some moments of procrastination (especially for tasks that I have little to no motivation for), I tend to work more steadily than I did in those first few years of college and I rarely run up against a deadline (soft or hard).
However, I've worked with plenty of software developers who wait until the last perceived moment and work crazy hours to try to meet the deadline. I'm not sure what they would do if they didn't have that deadline. I suppose they would either get fired for lack of productivity or just be productive enough when nudged to keep their jobs.
I always give a range with my estimates. The range is due to risks like current unknowns, probability of bugs due to interaction with other components, additional work that will possibly need to be done, documentation complexity, technical debt in related pieces etc.
Sometimes those can be rather extreme so my estimates have been "one month, plus/minus three months". If four months is acceptable then everyone is happy and as progress is made the range narrows. If time is tight then the discussion turns to reducing the range which puts everyone on the same page.
Another thing to consider, if you're doing client work and you charge per hour; the deadline is also when the agreed budget runs out. If you haven't finished by the agreed deadline you may have to get approval for extra spending.
You can't really have an open deadline, and just charge whatever when you eventually finish.
To further this conversation I think it's important to discuss continuous deployment or even shipping velocity that is both robust, lean, and efficient.
Consider this: If software, at a given state, is ready for consumption how long does it take until it is actually consumed and why is it that long? Could it be shorter? Could it be longer? Does it ever get consumed?
The soccer analogy fails because there is nothing else to coordinate besides scoring a goal. When you work for a company there are so many other areas waiting for that feature or release you just can't say something like that, I'm afraid.