I don’t understand why you think that software cost can’t be measured or foreseen. I work for a consultancy firm. We do it all the time and it works.

Bob replies: Hi, Rico, and thanks for commenting. I am a consultancy, and I also do it all the time and it works — for a given value of “works”, anyway. The problem is not so much estimating time as what the estimate is based upon. I tend to work with small companies with very specialized problem domains and often they have difficulty articulating what they want. Sometimes various stakeholders disagree on requirements and these disagreements are not always exposed right away. I also encounter a fair amount of scope creep in the form of statements of fact that turn out to be fancy, not to mention the constant churn of changing business requirements. Given that, I generally steer away from clients who want flat or capped pricing, because such clients generally don’t have the commitment to the project to “go the distance” and work through the various adjustments along the way. One of these days I will post more about that. Rest assured I’m not suggesting that you can’t come up with a reasonable estimate at any point in time; but for any non-trivial project it needs to be understood that this is subject to change and refinement because one cannot do 100% of discovery up front in the real world.

The only way to make any money on RentACoder is to live in China or India, and even then it’s worse than the Western notion of living paycheck to paycheck.

One possible strategy is to write a code generation system in a more expressive language (more expressive than PHP, that is — Lisp, Scala, F#, Scheme, OCaml, whatever) that can quickly handle CRUD apps where you can bang out something in 30 minutes. There’s a fellow by the name of Alex Peake (waybettersoftware.com) who has done this successfully for quite a few years. He gave a demo at the International Lisp Conference a few years ago where with a few hundred lines of declarative application metadata he would generate about 25,000 lines of C#, which would take about a minute to generate and build. Using expressive languages for source-to-source transform with the target being a popular but crappy language like PHP is IMO an underexploited niche that could be useful for learning more about the domain model. I mean, if you spend 30 minutes on a quick throwaway app where the actual generative software artifact is 150 lines of DRY XML, 100 of which you borrowed from a previous throwaway app, it’s no big deal.

The problem is these bids never ask for just a simple database frontend, they want the fucking moon and stars for less than the cost of a Big Mac, and they literally want it yesterday. What a race-to-the-bottom (in terms of wages and software quality) crapfest these code sweatshop sites are.

Excellent post. I’ve been a “provider” on Rent-a-Coder for over 7 years. It should actually be called “Rent-A-3rd world programmer for $2/day who didn’t pay anything for the software he’s using to build your “product” even though it’s not open source”.

Maybe that sounds a little “axe to grind”-ish, but frankly when I first signed up, I was genuinely intrigued that there might be a “quick and easy” way to make some extra $$$ via freelance software work. My Brother-in-law, in the wisdom his extra decade in the business over me has given him (me: 19 years in the business, him, he’s older than dirt, the man chews lisp compilers for lunch) said to me “It’s an online sweatshop, you won’t make money there”.

And, he was right.

Quality as a unique selling point is difficult in software.

How are most average joe non-technical clients to understand that the code I give them is beautifully written, documented and unit tested, rather than some slapped-together rubbish written for $7/hr? All they have is my word, unless they’re going to hire someone else to inspect the code.

Why else do people still buy Maseratis? :)