Smart Software Development on Impossible Schedules

Andrew Stellman writes "Jennifer Greene and I have an article in the new issue of Dr. Dobb's Journal called "Quick-Kill Project Management: How to do smart software development even when facing impossible schedules." We got a lot of great feedback from our last article on open source development, but there were a bunch of people who wanted to know whether it was really feasible to do planning and reviews on a tight schedule. That's a fair question, and this article is meant to be an answer to it. We pulled out bare-bones project management practices that will help you protect your project from the most common and serious problems, and gave instructions and advice for implementing them that should work in a real-life, high-pressure programming situation."

Comments

Tried and true

[Anonymous Coward]

Just make your developers pick up the slack by working lots of overtime.

Re:Tried and true

[Lonewolf666]

Only that it does not really work. Overworked programmers are tired programmers who make more mistakes. Long ago, research has shown that the best productivity is reached with a 40 hour week.

For details, see http://www.igda.org/articles/erobinson_crunch.php

Techniques don't make up for a bad schedule!

[gasmonso]

In the real world, no amount of tips and tricks is going to make up for a bad schedule. We end up doing code reviews while are software is in verification and validation. Is that smart? No. But when you need to get features done, that takes priority. As for documentation, that always falls to the wayside as well. The truth of the matter is this. The only way to make schedule is to totally inflate your time estimates on the project. We developed a rule of taking the engineers estimates and double them. Of course we're still running a little late ;)

http://religiousfreaks.com/

Re:Techniques don't make up for a bad schedule!

[AuMatar]

I don't disagree that tricks won't help you. I do disagree that getting features done should take priority. If you don't take the time to do things right, it will bite you in the end, either in debugging or maintenance, and likely take longer than doing it right in the first place.

Re:Techniques don't make up for a bad schedule!

[Morris Schneiderman]

We developed a rule of taking the engineers estimates and double them. Of course we're still running a little late ;)

Of course you're running late.

Let me guess - you're a young pup. It's always been the same, since the dawn of history. Every generation laments that succceeding generation has to make the same mistakes, and learn the same lessons, over and over again.

IBM solved this estimation problem back when dinosaurs roamed the earth, virtually all computers were made by IBM, and they came with a "free" (as in beer) systems engineer.

IBM's solution was that every level of management that had to review an engieer's estimate would triple the estimate of the person who reported to (almost invariably) him. So, if the estimate required two levels of managerial approval, the quote to the client was 9 times what the engineers figured it should be. If three levels of approval were needed, the official estimate was 27 times what the engineers came up with. Now you know how IBM was able to afford that "free" systems engineer with every computer.

Actually, there are ways to drasticly increase the likelihood of success on "mission impossible" projects. I successfully led an EDI implementation project where there was no spec, no data, no customer, no access to the computers, the programming had to be done in an obscure language, and there was an arbitrarily set deadline of three weeks. Sometimes you take on such projects because you want to stretch yourself. But that's a story for another day.

Re:Techniques don't make up for a bad schedule!

[MickDownUnder]

"...take the time to do things right...."

It's this notion that not going off and making a total ass over end cock up of a project, and "taking the time to do things right" that is really the problem with allot of developers, and managers attitude to software development.

The quickest way to get a project done is to do it right ONCE! Not five times badly. At developer meetings I've often heard managers or developers call UML and other methodologies a waste of time or unnecessary, it always flabbergasts me to hear such ignorance.

I've been in a situation where an (un)manager told me "We don't have the time to do document and do things the right way", my response was "No, we don't have the time not to do things the right way". We only had 6-8 months to get a large project out, with four programmers, a business analyst and two DBA's. I was told I wasn't to spend time in UML and documentation land, and you know what, I ignored him and spent 3-5 days creating high level class diagrams and interaction diagrams, the core of which lasted us through the entire project.

Any project that you forecast is going to take longer than 3 man weeks ie 120 hours needs some sort of design in place to ensure that the major architectural constructs are already decided before coding takes place.

The key to employing methodologies in projects successfully is unfortunately experience. It takes experience to use methodologies successfully, it takes experience to create good designs for applications and to judge when a design has become mature enough to allow development to commence without the risk of major refactoring later on in the project.

It does in fact take time to acquire this experience (years in fact), but it does not take time to employ this experience once it has been acquired. If you feel like your project needs some sort of methodologies in place, but "you don't have the time", the answer is simple, it's not that you don't have the time, you don't have the right personnel. It's true if you have no one with the experience of delivering the architecture required in a professional and timely manner then taking on methodologies new to everyone on the team is going to take time and lots of it. A good manager however will make sure that the project is correctly resourced.

Printer friendly version

[tcopeland]

Less clicking.

Smart?

[qbwiz]

It seems to me that if you have an impossible schedule, then you've already failed.

Re:Smart?

[Monkelectric]

Thank you. As a Software Architect, I'd like to say you are 100% dead on. I can spot a deathmarch project from the next county, and they *ALL* begin with phrases like "The market window for this product is..." or "We have to complete this by XXX or we all die" or my favorite, "My research shows every month we don't have capability X we loose Y dollars."

The appropriate response to any of these is "You should have proposed this project X months ago (where X is a conservative estimate."

Re:Smart?

[Bacon Bits]

And then you:
a) get fired.
b) get yelled at by your boss for making his company seem slow and incompetent.
c) lose the contract to that company that lies and says they can do it faster than your accurate estimate.
d) all of the above.

Do not lie to your customers when you make a quote for them just to get one contract. Be honest about everything. The most important customers are repeat customers, and you will be #1 on their list if you get them a quality product on time and on budget.

Losing one contract is no big loss, particularly when you know for a fact that your competition cannot possibly make the goals they're quoting. They will look like fools when they ask for time and budget extentions or fail to produce anything usable, and nobody likes to do business with a fool. As they say, people might not know the difference.

Certainly there will always be competition that lies to get the sale and customers that make poor decisions based on unrealistic quotes. Dishonest business and mismanagement are commonplace. That doesn't mean you should compromise your quality or ethics. That's called "professionalism". Customers worth doing business with appreciate that, and you will attract repeat business and new customers through word of mouth.

Re:Smart?

[jrumney]

You will be on nobody's list if you give them realistic schedules.

On the contrary, in my experience companies tend to come back to you to ask you to clean up the mess made by the low bidding cowboys when they realise that you were just being honest with them. At this stage you can often get away with jacking up the rates too, as they have made that realisation that you get what you are willing to pay for.

Re:Smart?

[g2devi]

I was in this position twice, in succession.

In the first case, my company tried to break into the professional market (we were big in the education market) so we announced a major overhaul and several new new features. It was a fantastic vision but unfortunately the president made the mistake of announcing the full feature set and that we would start accepting purchase orders that summer. By the time spring rolled around it was clear that were were going to miss the deadline by several months (very bad when most of your sales are in September), but since the feature set was set in stone, we couldn't cut features. The writing was on the wall but we accepted it and started working 13 hour days nonstop for 6 months. Customer complaints over late orders flooded our phone lines and prevented our previous customers from calling tech support, so our old customers felt neglected and our new ones started cancelling orders. We gave up in christmas and released the complete but buggy product and got more complaints.

The president was replaced, and under his leadership we were able to finish the product we wanted to finish within a reasonable schedule. Depite all the doom and gloom earlier, we survived. Unfortunately, we lost the senior software engineer and another developer in the process.

Then, president hired a vice president which wanted to revamp the whole product and move it to an even higher market. I gave a detailed estimate that said it would take 2 years. The VP said we'd go bankrupt if we didn't release in 9 months. I repartitioned the estimate and said that we could do it in two stages, each 13 months long or three releases each 9.5 months long. No go. It had to all be done in one release, and that release could take no longer than 9 months or we might as well close up shop. I came close to losing my job over the issue, but since I was the most senior programmer there with the most experience about the product, the new senior software engineer told the VP if I went he'd have to give up too. 2 years later, after a complete turn over of the team twice (including the new senior programmer), the product was finally release It had more bugs than we'd like, but it was recieved well by the market so it was "good enough". At release time, I was the last person standing from the original team, but I had enough and left soon after. The product was well received but since the team was new, a new version of the product wasn't released into three years later.

All the VP's threats didn't change an impossible schedule. You can't squeeze 2 years of work in 9 months, not matter how you try to force it. If you try, the most you can do is destroy your team and lose sales in the process. Most of those doom and gloom professies don't pan out. If they *are* true, someone is running the company wrong and development schedules are the least of your problems.

In the later case, had the VP chosen either a more reasonable schedule (like any of the 3 I laid out above), we would have had most if not all of our original team at the end and been able to do subsequent releases at the same measured pace. We would have had regular sales and the company would have prospers much more.

Being a software architect or project manager is hard, but ultimately, the best ones just lay out out all the credible options on the line and stick to their guns, even if it means risking your job. My experience in that job showed me that you can do it even in a hostile environment. If you lay things out properly and tell the truth, you may be resented, but you will be respected, even by those who chose to ignore your advice.

Easy!

[gillbates]

Stop reading slashdot!

Why do people buy into this nonsense?

[gubachwa]

Replace the title with one of "Smart Bridge Building on an Impossible Schedule", "Smart High-Rise Construction on an Impossible Schedule", or "Smart Heart Surgeries on an Impossible Schedule", and I ask you, how many people would want to cross that bridge, live in that building, or get that surgery? No one!

"Smart ... on an Impossible Schedule" is just management/corporate speak for "How to minimize the crapiness of a project when we know we can't spend the proper time required." You can dress it up all you like with sleak catch-phrases, and call it a rose if you want, but it still stinks.

Re:Why do people buy into this nonsense?

[pnatural]

For some reason "Impossible Schedule" in software development means cutting corners instead of increasing manpower.

Because that's all you can do (in general). Adding man-power to a late software project only makes it later, as was shown by Brooks some 30 years ago.

Good On Paper

[Anonymous Coward]

I'm in a similar situation as this, but atm, without the death march schedule. I've had quite a bit of time to consider the things and here's my conclusion.

1. Planning documents are a great thing and are pretty much invaluable to a team's success on a project. However, when you complete the document and the very same day, someone in upper management says, "Now this is a great document for you guys" and immediately starts making large changes (true story), there's not a whole lot you can do. Also, as every developer is aware, it is more than likely that the scope, goals, features, etc were never clearly defined in the first place which means you're either going to be wringing it out of management, playing guessing games or both.
   
2. Same thing is true for work estimates, great when you have them nailed down from the start, but good luck if scope creep is already kicking in before you've even started.
   
3. This is probably the only thing that will save your ass. If the whole team has some exposure to each part of the project, it's likely that someone's sanity check will save you days or weeks of time before a bad design decision was made (and on a lower level, coding decisions). Something that I've really noticed myself is that QA starts the first day that designs are put on the table. If a design hasn't been picked at to start with, then it's pretty much guaranteed that a developer is going to have to go back and redo a bunch of things to make a feature work the way it should.
   

So yes, the article does make some good points, but they only go as far as the factors covered are under your control. Even in the best circumstances, a poor programmer or a less than perfect team lead can hamstring the whole team. In the end, I think that the only benefit that articles like these have is to make you really think about process. Process might not save you in the end, but at least going through the effort should make you a better developer.

Getting back on track

[EmbeddedJanitor]

The main problem with getting late is that everyone lies. People only admit to being late when things get really bad and by then everyone is stressed.

Step 1: Admit being late as early as you can. If people start ranting, keep cool. Tell them that ranting won't fix it. The sooner you admit to being late, the sooner you can stop screwing around and start getting the project back on track.

Step 2: Help management prioritise what they want and what can be delivered when and what options are available. If they want a product to sell, they'll need to make some compromises.

Re:Getting back on track

[Nefarious Wheel]

"So, do you want the problem solved or would you like to spend the time apportioning blame?"

-- this little line has helped me throttle down a lot of rants over the years.

fast, cheap, well done

[davidwr]

Pick two. If you are so lucky.

A truly impossible schedule is by definition impossible to meet.

An extremely-difficult-but-possible schedule is by definition makeable if the correct resources are applied... correctly.

If management is giving you an impossible schedule, they are either idiots or setting you up to fail or both.

If they are giving you a difficult schedule and refusing you the resources you need, or do not have the authority to give you those resources, see above.

If they are giving you a difficult schedule and you haven't requested the necessary resources, then they are testing you and you've failed the test.

--

What do I mean by resources? I mean anything that can make the project go faster without sacrificing quality. This may mean additional manpower, the authority to say "no" to new feature requests, the authority to make feature cuts, or the reassignment or hiring of people with key skills to your project. It may also mean late-night pizza parties and family-support to the "code widows" until the project is complete. It can also mean additional or extended time off for the entire team after the project is shipped.

Re:Code reviews

[Bodrius]

Although I think pair programming can be great in many contexts, I think they really have mostly complementary advantages:

a) An in-depth code review (formal or not) has a big difference from pair programming: the analysis is done by a fresh pair of eyes. Pair programming can result in two people who are too close to the code to detect flaws in their underlying assumptions.

In my experience, code reviews are not that useful to detect syntax or style errors, nor do I think they're meant to be. That's the kind of thing you can automate with a tool anyway. If you're spending all your time looking for that, I agree most people will feel it is a waste of time.

But code reviews can be extremely useful at a higher level: to detect design gaps and flaws in the implementation, or just get improvement suggestions that a fresh set of eyes ban provide, once you see the whole solution.

Often the very process of explaining the whole solution top-down after implementation allows you to detect issues. Other times you can identify common patterns and needs in different features that we're not visible before ('hey, I had to do the same thing elsewhere and we have a common utility class/method for this that avoids this obscure bug here').

b) I don't see how it should be that different regarding the attitude problems you are describing.

If people get so easily offended by a code review, it seems to me it'd be because of team problems bigger than a code review process: either the programmer thinks their code is perfect (which is something to fear) or they don't trust their peer enough to have honest conversations about code.

I guess this is the one advantage of pair programming, it forces people to work closely and practically guarantees that trust. But I do not think you need to lock two people in a room and do everything together for them to trust each other professionally in a healthy work environment.

Heck, code reviews should be fun.

They're peer conversations about code, and if you enjoy programming it should be fun to argue techniques and solutions with your colleagues, even (specially?) when there are strong disagreements. Both sides typically learn something from it; even if a lot of times it is not something you want or need to change in the current code, it will be useful knowledge elsewhere in the future. Just like college or academia, for that matter, back when coding was done for fun.

12 Rules

[johnBurkey]

1) Make sure you know what you are building. Many project delays are because the "customer"- the manager, corporate head, you? doesn't actually know what they want.

2) Make sure you only work on things that you need to ship version 1.0 of that.

3) Make sure you keep the prototype running always.

4) Show Demos every few days to make sure noone is confused about what is going on.

5) Tell them they can ship it whenever they want, they write the check.

6) In the meantime, work towards the goallline like a football player, do not circle it like a lion waiting for it to die.

7) Don't make any project your time to show how clever, cute, or interesting you can be...

8) Keep Teams/Egos/Methods/Files/Modules/Projects/build times small. Small is good.

9) If someone is not clicking with the rest of the team:

- talk to them privately

- reassign them

- if this person is you, read #11, and consider if you want to build this project, or do something else. Follow your heart.

10) Do the riskiest part of the project first.

11) Remember that the enemy of the better is the best.

12) Don't worry about it. If you are working hard, and follow 1-11, you are doing your part.

That's enough to chew on. As homework, go build a paper mache model of the project, complete with testers whizzing around, filing bugs that are actually feature requests.

The EA method...

[abigsmurf]

Companies should follow EA's tactics. Say to the coders "you'll have an easy schedule during the majority of a project but at crunch time, you're expected to work huge hours with minimal overtime pay" You start the coders on a project nearing crunch time, have them slave away till it's finished then move them onto another project that's nearing crunch time. If no projects are nearing crunch time move ahead the schedule on one knowing full well you can't meet the new date but still force the overworked coders to give 110% to attempt to meet this impossible target. You keep all the experienced programmers doing the stress free base work and once the outline is done, you move them to start another project. The vital experienced coders stay happy and stick with the company and the dime-a-dozen average coders have all the work they could possibly give squeezed out of them for minimal pay. Gotta love an industry with few/weak unions...

double time ain't enough.

[fprog]

Seriously, doubling the time is NOT enough.

Top 12 - Rule of thumb says:

1) Triple the estimated time + 10%.

2) Add 2 weeks to that amount for a complete code review.

3) Any changes by the customer means "adding 2 weeks to the schedule", even if it's one line of code... why? because, it must pass documentation, Q.A., validation and be reviewed by the entire department, without accounting for "double bug", bug induced by another bug and stuff like that or bugs that are in the core library and means retesting "everything", "every module", etc.

4) Any changes must be approved, reviewed and means adding delay (normally a big NO-NO) and therefore, 99% of the case thoses changes are left for future phases or abandonned by the client when they realise the implication. If not, it's your objective to discourage them or force them to reconsider by any means. =P

5) Don't give any feedback to the customer unless you must! If you do, downgrade any progress to minimum to reduce expectation and refrain the customer from adding new stuff to the TODO list.

6) Which means, each phase must be clear, consise, humble, realistic, feasible, with lots of buffer time for fixes, reviews and testing. Exagerating how complicated it is to a customer is always a good idea, because in the end, everything that seems easy is actually very difficult.

7) Do minimum documentation, UML to get started... They'll get rewritten at least 30 times, before everyone in the group agrees after what 40 meetings?

8) Once the phase somehow works and the thing is somehow settle, start documenting it, so you don't forget. It's actually a very good way to detect logical mistakes, misbehaviors, bad coupling, bad cohesion, missing corner cases, bad design choices, usability problems, etc.

9) Best teamwork is small team of 3-5 senior people working toghether hand-in-hand, sometime helped by 1-2 junior, which can do much better than 120 junior who are completely clueless and never deliver on time...

10) For big projects split things up in component and/or phases that a small team of 3-5 people can do, keeping in mind of the big picture so its scales up, but ignoring any meaningless future details that don't matter "right now".

11) Rush the people to do it in "the simple 1/2 time delay", keeping in pocket the "double time" remaining for any arising issue and reworking the core libraries, overhaulin the code, reviews, fixing bugs, etc. This means that if you are really late, you are actually late on your "buffer time". If things goes well, then the project will be done before it's expected, which will impress any client.

12) Finally, but not the least, there's no such thing as a bug, it's just a "small improvement", a "new feature", "code overhaulin", "mispelled requirement" or a "security enhancement". It keeps customer smiling, it's less depressing for everyone and overall keeps the mood up on everyone face with a laugh or two! ;-)
Furthermore, no ones want to hear that the code is "full of bug", but saying that a group of people are always "enhancing, overhauling and improving the code base" also means bigger bonus! =)

Hope it saves you from any future project delay or cost overrun!

Release early, release often

[mcrbids]

I can't stress this one enough!

As architect of a small software company, the most frustrating aspect of designing software is the knowledge and understanding that there's no way to know how to do it right until you deliver something that's wrong. People almost never seem to know what they need, but they do know that whatever they have isn't it, and they'll tell you why.

So, we have a very different tack, similar to Agile Programming. When we implement a new feature or functionality tidbit, we release very early - pretty much as soon as it more or less works without major errors, - with as much fanfare as we can manage on the cheap, and then wait for the feedback.

We're very up front about it, and openly welcome the feedback and ideas. This takes any conflict out of the relationship and turns it into a sort of partnership. Customers love being listened to, and feedback we get, in droves. When a customer sees THEIR idea implemented a week or two after they suggest it, even when it's something stupid-simple (such as having a particular button selected by default to avoid a common mouse-click) then they're your advocate for life!

It's that continuous, iterative cycle that's resulted in our young, 3-year-old codebase having eye-popping features, and remarkable stability. The software updates itself at the client's discretion, so nobody seems to mind much. They update as often as they like.

With this model, there is no due date. It takes me about 15 minutes to issue a release of our software. The idea of a "release" means almost nothing - we've done more than one in a day! (we've released 46 official releases in the past year alone, with too many "unofficial" releases to count)

Faced with this "impossible" situation, (I've lived "impossible" schedules for years now) I'd step UP the release cycle and start pre-annuoncing alphas/betas of the product the instant that something that appears demonstrable compiles and can be stuck on a dev server someplace. Invite comments and download. Call people to ask about feature X or Y. Let them know it's really early, and make sure that they have a place to bitch about the problems they find.

When you do, you'll be surprised how much of what you thought was required was, in fact, completely un-necessary - or at least could be put off until next March for a future release. But, you'll find some simple, straightforward ABSOLUTELY GOTTA HAVE that takes a man-month to code that the users would sacrifice their firstborn to have.

Agile software methods will find this "gotta have" pretty quickly. The waterfall model of software development would take a decade.

You decide.

This model won't work for all products and/or markets. And it's very important not to take away functionality that the customer previously had, or they get the feeling that you're taking something away, and that's bad. But for us, it's been very, very successful, and the relationship we have with our clientelle is very friendly, close and intimate.

PS: Maybe it helps that we're an ASP.

Re:Code reviews

[Anonymous Coward]

The message your coworker was giving you, that you missed, is, "Hey, this code is illegible! For starters, how about splitting up this monolithic function into something remotely comprehensible, and then maybe I could give you some useful feedback."

No charge for the clue.

Just say no (and more)

[janolder]

I've been in the business for over 20 years, as a developer, as an architect, as a team lead and as a manager.

From architect on up, one of your key job responsibilities is to push back on features, schedules and so on, and to set expectations right from the get-go. Early on, I used to laugh out loud when being told proposed dates by marketing. That didn't go over too well, of course, so I've adopted a more diplomatic way of saying 'no' since. :-)

The gist of it is that many executives believe in Spanish management (very well explained in Peopleware). This boils down to setting ridiculous schedules, asking for continuous overtime, etc. The idea being that every minute an engineer spends more will get the product out the door faster. Of course, this is not the case as Peopleware will tell you in great detail. It is also matched by my own experience.

However, if you push back with data in hand (such as a detailed sizing) and a constructive proposal what to do differently, you may very well end up with a more reasonable schedule, a good product and happiness all around.

A few more gems in Peopleware:

• Schedules are counterproductive. Teams that don't have schedules ouperform those that do by up to 50%.
• Overtime is only productive for one week.
• Cubicles are sinkholes of effectiveness. Why do Microsoft and IBM only have offices for engineers?

Peopleware is rather sobering. Every other page you think "wait - we're doing exactly what is being described here." The good news is that you can do something about it once you can recognize the signs.

For those of you with a humorous bent, I highly recommend checking out Joel Spolsky's articles on project management. A few highlights:

• Human task switching considered harmful
• Joel's test to identify bad software companies

As far as the TFA goes, once you've accepted an impossible schedule you're already hosed. If you can't push back, leave. The job market is good right now.

If you can get to a reasonable schedule (by way of reducing features, adding time or people), the TFA is a bit limited in its scope. I have a few recommendations that have worked for me in the past (your mileage will vary, and you should read Peopleware anyway):

• You need to have the spec in writing.
• You need to use source control, even as a single developer.
• You need to have a single button build.
• When the build breaks you roll back the change that broke it or you stop everything until it is fixed.
• You need to build daily. Or better yet, continuously.
• Unit tests are your friend. They're less useful on GUIs but for logic they're a godsend. Personally, I can crank out high quality code about twice as fast with unit tests (if you consider debug time). Reduced maintenance and improved sleep is a bonus.
• In the same vein - automated system testing (if possible) is a wonderful way to improve shipped quality. Your testers (if you have any) cannot test everything.
• Code reviews are great. We use code reviewer with great impact on code quality.
• Hire the best people only and fire those you have no hope of redeeming. It may sound harsh, but allowing an ineffective developer to remain on a team is a great way to kill both the team and the project.
• You must have a bug tracking system.

Re:That's funny, but...

[StrawberryFrog]

Learn to use copy and paste

Nooo!

I wish I could beat with a crowbar all the cut-and-paste programmers who make my life difficult - I have to maintain the piles of repetitive crap code that they produced - they knew how to use Ctrl-C and Ctrl-V but could not or did not bother to do the most basic of programming taske: eliminate repetition by factoring code into meaningfully-named, paramterised procedures. This was best practice in 1970 already but so many people still don't get it.

A new word for managers' vocabulary

[hey!]

Sustainability.

I've taken development teams down into the Valley of Death and out the other side. It can be done. But there are consequences, and you're kidding yourself if you think you can avoid them.

The first consequences is that you can only do this once. If you're smart, you hire smart people. They know when you are ordering them to jump into a pile of shit. And if you're smart, you hire craftsmen, and craftsmen have pride. They resent being told to do that. What that means is that if the second time you try it, you end up losing all the people who can find a job easily -- in other words your most valuable employees.

Another word that should be in managers' vocabularies is investment. Developers create assets that produce future revenue and expenses. You want programmers working as hard as they possibly can, but not simply to get the job done. To get the job done right, which means producing something that generatesw more revenue and less expense. When you focus on just getting the job done, you end up with something that barely breaks even. That's the minimal criteria for "not failing", not the maximum possible success.

That's why I hate questions that start with "How hard would it be..." because they focus on the present, not the future. It should be "I think we can make X dollars a year; how much would it cost to support this feature, and would the amortized development costs be justified by the net revenue given the other things we could put our efforts into?" It's not as pithy as "How hard would it be...", but it covers the things managers should think about: the future impacts on revenues and expenses, the opportunity costs of the road not taken.

Can anyone really estimate accurately?

[Anonymous Brave Guy]

We would take the estimates, double them and add 30%, that is how I bid out all work.

It's interesting to look down this thread and see how many different rules of thumb are employed, apparently by quite experienced developers/managers.

It seems to me that we can make two reasonable assumptions based on this:

1. Development usually takes longer than most engineers estimate up-front.
2. No-one has yet found a reliable way to assess how much longer up-front.

One might hypothesise, based on these assumptions, that any project plan that quotes hard figures up-front will be inaccurate most of the time, and without allowing a huge defensive overhead that inaccuracy will often be in the wrong direction.

Now, working from unreliable estimates is in no-one's interests. If a company overestimates by too much on a fixed-price contract, it risks being undercut by a more realistic competitor. If it underestimates, it will look bad when things over-run, its profits will fall, and its reputation will suffer as well.

So what conclusions can we draw from this?

Firstly, if we're looking for realistic management, the way to go is almost always some sort of time and materials contract, with floating deadlines based on when features are actually ready. This allows good developers to do their job properly, which in the long run will be more efficient, and it means that other groups such as sales, marketing and customer support can be kept informed and given realistic expectations.

Secondly, fixed-price contracts are only a good bet for a software development firm if it can include enough slack to be safe, without including so much that it can be undercut by more realistic competitors. The corollary, of course, is that if you're a client, you're almost inevitably going to be screwed by agreeing to a fixed-price contract, even without considering the lack of flexibility it entails.

It is a rare project indeed that truly requires hard deadlines. Smart clients and smart development shops will reach a more flexible arrangement, and come to respect each other for having realistic expectations and providing realistic updates as time goes on.