≡ Menu

Dumb myths about software, part 342

From TechCrunch

“The marginal production cost of software is zero. That’s what the price should be.”

Uh huh, sure it is.

Comments on this entry are closed.

  • J L Smith

    Ian, are you confusing unit cost and marginal cost here? Unit cost is total costs / total number of units, which has a number.

    Marginal cost is just how much it costs to churn out one more unit.

    In software that’s not exactly zero, but it it is close enough to make the glib statement above true for the first part, if we’re talking about packaged software.

    Where TechCrunch go wrong (and they are the biggest bunch of Bubble 2.0 wankers going) is by confusing it in the reverse direction, so you’ll never get to recoup your outlay. That’s the only way the ad-driven model their so fond of can be made profitable: by pretending that development, hardware, indemnities etc don’t exist so every ad view does not have a correlating cost.

    Also, the marginal cost is 0 only works for packaged software, for a service that is disguised as software (as all web apps are) there are costs per transaction.

  • http://profile.typekey.com/ianbetteridge/ Ian Betteridge

    No, I’m not confusing the two at all. It’s a myth that the costs of software don’t go up the more copies you make, because it neglects:

    1. Support costs

    2. Bandwidth costs

    If you’re running a web-based application, the more people who use it, the more it will cost you to run. If it’s a non-web application, it will cost you more to support it. Neither of these are neglible – and both ramp up steeply if you end up with a hit product.

  • J L Smith

    I am actually sort of agreeing with you, actually. The problem is with the sloppy language used by people like TechCrunch (which won’t take my comments for some reason.) An instant messaging service is not software in a way that makes the marginal cost = zero equation true.

    However, you have to be really careful with marginal cost: support and bandwidth costs do not necessarily count. Marginal cost is how much it costs you to make the next thing, it’s not an average.

    The marginal cost of a Starbuck’s espresso comes from the coffee beans and the electricity to drive the pump. The barista’s wages, rent etc cannot be counted towards the marginal cost unless you have reached capacity. You’ve already sunk that cost, it’s gone. However, if you need to hire another barista to make the next espresso then the *whole* cost of hiring that person is the marginal cost of that espresso. So espresso 1’s marginal cost is 5p, espresso 2’s is thousands of pounds.

    So here’s a conundrum: if you’ve hired a support person and paid for your bandwidth upfront then the marginal cost of your web app is pretty much zero. If you’re paying per transaction via something like Amazon Web Services then you do have a real marginal cost.

    And the barista example extends here. If you buy, say 100Gb of bandwidth from your ISP and your app uses that in under a month, the payment you have to make to extend the cap becomes the marginal cost of the next transaction.

    Because marginal costs are so wildly variable you’d be a cretin if you extrapolated them into a simple pricing decision. (“Hey, this espresso is gonna cost me 10k to make because I have to hire someone to make it. Make all the espressos 12.5K so I can make a profit!”) Where they are useful is (fairly obviously) in decision making.

    The fact that you get shit like this out of a supposed business mag like TechCrunch is a clear sign that the next DotCom bust is just around the corner. Bugger. I’d just got a gig I like as well.