Slashdot Mirror
Does Outsourcing Programming Really Save Money?
itwbennett writes "In a blog post titled 'Why I Will Never Feel Threatened by Cheap Overseas Programming', John Larson tells the story of a startup that shipped its initial programming to India, paying $14 per hour, with predictably disastrous results. Larson concludes: 'I have yet to see a project done overseas at that sort of hourly rate that has actually gone well.' But in this not-uncommon tale of outsourcing woe, is the problem really with the programming or with unrealistic expectations?" The comments on Larson's blog post (originally titled "Why I Will Never Feel Threatened by Programmers in India") seem to me more valuable than the post itself.
I once did some contract work for a place that made the mistake of outsourcing a major programming job. My job was to maintain the outsourced code, and keep it functioning (barely) while the internal programming team worked on building a complete replacement from scratch, at half the cost, with the actual system requirements being fulfilled. I spent four months fixing bugs in deliberately obfuscated perl code, at consultant rates, because none of the internal staff they had hired was either able to figure out perl code in general, or willing to even try to sort out that mess. The outsourced programmers in question had the dodgy business practice of deliberately making their code difficult to read, and only including comments like:
# 16426-b
The code in question contained wonderful constructs such as pointless loops where a value would be iteratively divided by the numbers from one through a thousand, then restored to it's original value without being used in the altered form. I started the project with about 6 million lines of perl code, and by the time it was over and the replacement was ready, tested, and brought online, there were only 2 million lines in the outsourced code, including about ten thousand lines of comment code that had been added while I was working on it. I hadn't even looked at about half of the remaining code.
After the initial work was done (poorly), the outsourced programming company announced that their code maintenance fees were being increased, thinking that their poor coding style had essentially locked the client in, and left them unable to get help elsewhere. The only staff member the company had who was willing to make the attempt unfortunately committed suicide after only a month of trying. (Personally, I believe it was unrelated, but the other programmers there claimed she was perfectly fine until she started working on that code... after two months of it I could see why they would think that.)
So yeah, in my experience, outsourcing programming does not save money - if the company I did that work for had just had their own people write the original code, they would have saved a massive amount of money.
His point is that, yes, managers are going to be blinded by that $14/hour price tag and go with that route, but those projects usually fail. They are pitched for too many man-hours up front and they usually run over, and even then the result often isn't up to snuff. The result is that they give up and hire American (Western) programmers to finish the job at market rates. Thus most of the "value" of the overseas effort becomes a cost overrun, but the worst part is that time to market suffers because the initial specification valued cost over time to an unrealistic degree. A company can only get burned like this so many times before cooler heads prevail.
Breakfast served all day!
While I don't really have too much of a horse in this race, and I don't think it's indicative of all outsourcing teams, I remember one gem from an internship I was working during college. We were given a code base developed by a 'guru' from eastern European country, I want to say Ukrainian. The product was an add-on for Outlook, and at one point I was given the problem of 'The app is slow with more then 10 contacts, almost unusable, speed it up.' After a bit of hunting around I found the problem was where the add-on was searching through the contacts lists trying to find a subset.
They'd managed to create a search algorithm with a big O of something around n cubed. I set about rewriting it to get it down to N all while wondering why on earth Outlooks API didn't support this relatively basic feature. It was at this point that I discovered that the API actually did support it, however the programmer had written a wrapper around the API hiding the method in question, then re-implemented it in the train wreck of code I was removing. That was the last time I ever worried about competition from cheap outsourced labor.
lmgtfy:
Research:
1: Babiak P, Neumann CS, Hare RD. Corporate psychopathy: Talking the walk. Behav
Sci Law. 2010 Mar-Apr;28(2):174-93. PubMed PMID: 20422644.
and the summarizing article from Time.