[Solved] Your coding performance with long projects

mpan · 2014-05-04 18:59:15

I'm stuck in a commercial project that had its official deadline 1.5 year ago. I participate in it as a software developer with very limited¹ maneuverability and have to develop a relatively simple application. This quite small piece of code is being re-developed, fixed, modified, dirty-patched, refactorized over and over again for over 2 years now. And has no specification other than talks that happened on a timespan equal to about 2.5 years, plus some mails (usually on backup DVDs on a bottom of some box in my closet).

Now to the point: I will have to give my client some estimates on how fast I can finish another request. The problem is that in my whole career, and I code for like 13-14 years, this is the first time ever I'm stuck in such situation. My performance dropped to unbelievably low levels, because all I do is searching for old e-mails, sheets of paper with notes and analyzing code that while [luckily!] documented, contains lots of q&d fixes and temporary solutions added dozen of months ago (I don't even know anymore what they're and why they were added).

Can any of you give me some estimates on what could be the scale of that performance drop? Based on your experience, if you have any. 2x? 10x? 50x? Is there anyone here, who actually had such case ? I would appreciate any comment.

____
¹ Communication with the client of my client is limited; normal user-developer interaction I'm used to doesn't really exist; "soft" software development things are carried out by other people (who are not developers); for large part of the project it was not even known what should be done and I could do nothing about it etc.

Last edited by mpan (2014-07-05 13:34:28)

ewaller · 2014-05-04 19:20:49

Allow me to be blunt. It sounds like you are coder, not a software engineer. Now that the unpleasantries are out of the way, it sounds like this is an opportunity to step up and be an engineer.

First, formalize what it is you are bidding. Bound it in writing and get buy in from the stakeholders.
Develop an architecture and document it.
Plan the work. Get a good schedule management tool. Figure out what the tasks are. All of them. Design, meetings, reviews, project tracking, coding, integration, testing, release, documentation. Assign realistic time estimates for each step. Plan for failure, allow four rework cycles until you can prove you can do better. Factor in how much time you really get to work an a project a week (hint, it is not 40 hours; most places you are lucky if it is 20).

Coding is not what it is all about. It is a vanishing small part. One who is a coder on a large project is usually doing the grunt work for the real engineers. Off hand, plan 10 hours of other stuff per hour of coding. Plan on using half your time doing non-engineering - sensitivity training, meeting vendors, reading resumes, writing performance reviews, participating in other project's reviews, budgets, planning capitol expenditures, etc...

Good luck.

mpan · 2014-05-04 20:09:31

Well.. thanks for the answer, but it's not exactly what I was asking for.

Believe me, I took a fair share of cookie-eating and tea-drinking even in this project. It just happened that at the end I'm left alone with code to be worked on - something that occurs from time to time. The project has to be finished and no one, including me, cares if doing coding personally suits software developer or not.

The problem is neither with defining who I am nor with my ability to make schedules. It's about problem with estimating how much coding performance could be degraded because of time on which such small thing is spanned. No matter if I'm only designing things, limiting my work to specifications and testing, just hiring others or doing actual coding, I'm used to finishing particular unit/module/part/whatever within a month max. Not 2.5 years! My experience fails completly in that case and I'm unable to give any good estimates. This is why I'm asking others - maybe someone else happened to be in a similar situation.

If the quetion is not clear, let's put an example. Usually I'm able to perform actual work on things for about 12-15 hours a week, maybe more if it includes meetings or other things not requiring good focus. And I can make schedules for that quite well. Now... because after so much time my work is consisting mainly of literally digging into the closet, I can get much less. The problem is I'm unable to determine how much less, as it's pretty hard to measure that kind of performance, especially on short term (and I don't have another year to determine that). It could be as well 2x as 10x less. I'm not the only person who fallen into such trouble, so possibly someone can say that he was so unortunate too and something that should take a week took 6 weeks.

Last edited by mpan (2014-05-04 20:19:33)

cactus · 2014-05-06 07:22:11

mpan wrote:

The problem is that in my whole career, and I code for like 13-14 years, this is the first time ever I'm stuck in such situation.

That sounds like a clear warning sign. Based on what you have written/described, the project sounds awful. Get out if you still can.

drcouzelis · 2014-05-06 12:51:51

This sounds like an awful scenario. My biggest concern is that you will end up with the blame for a failed project that was already destroyed when it was given to you. Please take all precautions to prevent that from happening.

Can you talk with your customer (the person paying for it, your boss, whomever) about scrapping parts or changing the process in areas that just don't work any more? Would they agree to taking all the papers and discs that are in the closet and making a bonfire with them? The goal would be to somehow "start fresh" and do things "your way", which would hopefully make things go faster and easier.

If you want a time and money estimate, I would suggest picking a small task, maybe something you think you can do in one week's time, and trying it. See if you can get it done in one week despite all the hurdles you need to jump over. Whatever the outcome is, tell your customer and see if they want to continue spending more time and money on it. In other words, get someone to pay for you to do a "one week test" then base you estimates on it.

mpan · 2014-05-06 13:50:58

Both of you: thanks for the response and compassion .

If I could, I would drop the project long time ago. However I don't want to, because of loyality to my client. We've been cooperating on other things in the past and I don't want to leave him alone with that mess. It's not his fault either.

The only thing I need is to give him some time estimates in this week. And since it's my first time in such strange situation, I find myself failing at this. Since I have no time to measure this myself, I ask others... maybe someone can provide me with some info, because he or she has been in similar situation. And I ask for scaling factor, not actual time values, because not only each case is different, but in my case some additional issues have to be taken into account.

drcouzelis · 2014-05-06 14:04:05

Oh. Ok...

How about 9?

mpan · 2014-05-06 14:28:44

If only it's based on an actual experience, I'm happy with that .

(and yes, I do realize how stupid my request seems; but the situation is stupid itself)

der_joachim · 2014-05-06 18:00:46

mpan wrote:

If only it's based on an actual experience, I'm happy with that .
(and yes, I do realize how stupid my request seems; but the situation is stupid itself)

Well it is an odd request. This is an intriguing thread. If loyalty is your motivation not to pull the plug, perhaps it would be best to restart from scratch, this time with a workable set of specifications. Otherwise I'd just quit. This sort of dead horse projects is very very draining and nobody actually gets any better from it in the end.

Seen it, done that.

zorro · 2014-05-07 00:03:46

Having to account for my hours weekly, I know I can spend up to half of my time on a high priority task and the remaining hours are used on all the other software engineering activities. I have been working on my current project for >5 years and it is on-going due to new requirements, hardware/software(XP) obsolescence, adopting/replacing other programs etc. The larger the code base becomes the more time is spent maintaining it, resulting in less time to complete other activities on the plan. I can accurately estimate how many hours it will take to complete an activity based on actuals, experience and a multiplier of 3 times the best case. If you are charging for your 'other activities', I suggest you account for this using an additional multiplier x2, if 50% on task.

mpan · 2014-05-12 02:55:55

Thank you all for the feedback. I've multiplied by 3.75 and hope this will be enough.

Arch Linux

#1 2014-05-04 18:59:15

[Solved] Your coding performance with long projects

#2 2014-05-04 19:20:49

Re: [Solved] Your coding performance with long projects

#3 2014-05-04 20:09:31

Re: [Solved] Your coding performance with long projects

#4 2014-05-06 07:22:11

Re: [Solved] Your coding performance with long projects

#5 2014-05-06 12:51:51

Re: [Solved] Your coding performance with long projects

#6 2014-05-06 13:50:58

Re: [Solved] Your coding performance with long projects

#7 2014-05-06 14:04:05

Re: [Solved] Your coding performance with long projects

#8 2014-05-06 14:28:44

Re: [Solved] Your coding performance with long projects

#9 2014-05-06 18:00:46

Re: [Solved] Your coding performance with long projects

#10 2014-05-07 00:03:46

Re: [Solved] Your coding performance with long projects

#11 2014-05-12 02:55:55

Re: [Solved] Your coding performance with long projects

Board footer