Micropayments, microprobability

by Daniel on September 16, 2003

We’re having a good old back and forth slagging off each other’s music tastes and calling each other fascists in the comments section at John Holbo’s site. As you can see, the issue of “whither the music industry in a world of reduced intellectual property” is bound to bring out a lot of interesting opinions; I think this is because a) we don’t know what the heck will happen b) we’d all like to believe that the answer will involve us all owning loads and loads of fantastic music for next to no cost but c) we all suspect that it probably won’t. As you can see if you follow the link, my role in the debate appears to be partly to snipe about obscure, irrelevant and probably wrongly remembered points of price theory and partly to act as the de facto defender of the music industry as she currently stands. I’m not sure that this reflects my genuine views, but in all similar discussions, I have historically ended up in it because of a number of points on which I think people are badly misunderstanding the economics of the music industry. I don’t want to start on a five thousand word thesis which will never be finished on this, so I’ll try to list my points of disagreement one by one in a series of posts. Starting with the easiest point and the one on which I’m most sure of my ground; micropayments are not going to happen any time soon.

It’s an attractive middle ground for a certain kind of reformist position on the Imagined Future Music Industry; one in which the removal of copyright (or its substantial undermining) doesn’t lead to us having to imagine a world in which music recordings are free, but one in which they are just very cheap indeed; say, anywhere from a few cents a song to a dollar an album. Making this sort of assumption allows you to do calculations of the sort “say 100,000 albums at a dollar a pop, figure 50 cents production cost and the musician earns a decent wage assuming that there are no packaging, promotion, legal and other suit-related costs” and come up with a “cottage industry” model of the music world which is often pretty palatable to people who like the kind of bands you hear on college radio – local heroes who might sell 100,000 albums at a pinch and are young enough to be happy with the idea of living the musicianly lifestyle on a low but survivable income.

But it’s just not possible. The fundamental error is in assuming that “cheap” and “free” are close cousins when they’re not; they’re fundamentally different ways of organising your microeconomy. The difference being that once you’ve decided that you aren’t going to operate on the basis of freely provided goods with an abundance of everything for all, you need to have an entire supporting infrastructure of institutions designed to make sure that everyone pays for what they’re taking. And that’s actually surprisingly expensive (I seem to remember that someone wrote an article in the Guardian in the 1980s which argued pretty plausibly that if we knew that everyone in the country was completely honest, roughly half of the workforce would have no jobs to do. Not just policemen but checkout operators, bus conductors, accountants, etc, etc.

You can’t ignore these costs; they’re important. And this is a problem if you’re thinking about a “micropayments” system for transactions of twenty cents to a dollar; the costs of providing a secure payments network have the unfortunate characteristics of rising pretty much in line with the number of transactions (not too many economies of scale) but being pretty fixed with respect to the size of each individual transaction (so they are a much higher proportion of the total cost of making a micropayment). There seems to be a systematic tendency for people to underestimate these costs.

Here‘s a pretty representative price sheet for dematerialised credit card transactions. If we ignore fixed costs (setup costs and monthly flat rate services like billing), which is a reasonable assumption as we would hope to be amortising them over a very large number of micropayments indeed, we only need to take into account the “transaction fee” of $0.20-0.50. This is, to quote, “the costs of the datacenter, network usage, etc”. Let’s assume that there is an element of monopoly rents in here (the big card networks are in general very cagey indeed about the true costs, but it’s such a small oligopoly that the pro tanto assumption has to be that there are substantial rents) and that the true average cost of providing network and data storage costs is 10 cents. But on the other hand, we have to take into account that this is not marginal cost pricing; it’s an “average marginal” cost set by the providers to ensure that data transmission costs are covered on the whole. It’s highly likely that within the system, customers who send a small number of high value payments subsidise customers who send a large number of low-value ones. My guess is that the true marginal cost of a micropayment might be $0.20. So immediately, we see that “a penny a song” is simply impossible. It costs more than that to send the payment message.

Why so much? Email costs fractions of a cent to send, after all. But (see above) email’s free. Free things are much cheaper to administer than things where you have to keep accurate and exact records of who paid what to whom. In principle, the guys behind various Digicash operations reckoned that they’d solved this with “blind signature systems” (basically complicated protocols not unrelated to public key cryptography, for generating special numbers that would be difficult to forge), but they all tended to founder on the fact that when you tried to take the money out of the Digicash economy and turn it into a real bank deposit, all the book-keeping overhead that you thought you’d dodged ended up catching up on you. And even without book-keeping overhead, any payments network has to be many times more secure than ordinary email. I am willing to be convinced otherwise (though be warned, some big names in the business have tried and failed in the past) but it is my current opinion that it is not technologically possible at present to run a payments network materially more cheaply than the credit card companies, without compromising on security.

And the irritating thing is that just because you’re making micropayments, doesn’t mean you can reduce security. As an important plot point of the film Superman 3 illustrated, if I can reliably steal a penny from you in the certainty of not being caught, and if I can automate the process of doing so, then I can steal arbitrarily large amounts of money a penny at a time.

It gets worse. There’s also the ad valorem charges of the card processing networks. These are levied to defray the cost of sorting out fraudulent or mishandled payments (sort of an insurance premium). They’re usually only 2-3% of the transaction, but it’s likely that they would be much higher percentages for micropayments. That’s because the cost of sorting out a mispayment comes in two parts: checking that it was in fact a mispayment, and making the customer whole. The second of these obviously depends on the size of the disputed transaction, but the first is a fixed cost. I don’t think it makes sense to assume that nobody will care about mispayments of small amounts (if they do, they’re making themselves vulnerable to the kind of fraudster discussed above), so we could be talking about much higher percentages.

Bottom line; I’d be surprised if you can send a secure payment over the web at a cost of less than 30 cents, all in, even with the best will in the world and no profits. I’d be interested in any examples of someone offering the service for less, but would tend to suspect that they’re subsidising it from some other source of funds. Which means that the smallest “micropayment” worth bothering with would be about $2. That’s not really all that micro anymore.

Stay tuned for this occasional series; I think my next target will be a few assumptions about what it is that record companies actually do for their money.



Tom 09.16.03 at 6:50 pm

On the other site, you (Dsquared) say:

“Your other points seem sound, but this one had a flavour of “If we sell a toothbrush to every person in China”, “If we capture just 5% of the potential online funerals market” and other golden oldies from my brief involvement with the venture capital industry.”

We want to hear more about your days as working for VCs.


Seth Gordon 09.16.03 at 8:00 pm

Furthermore, even if the cost of processing credit-card transactions goes down to zero, micropayments still won’t become popular, because customers hate them. Clay Shirky has more on this topic.


Doug 09.16.03 at 8:00 pm

1. So why does iTunes work at a dollar a song? And any number of reasonable predictors saying that price is likely to fall, given the demand?

2. Legend has it that American manufacturers of consumer electronics figured what it would cost to make a product, added a little profit, and set their price. Japanese manufacturers looked into what price the market would bear and figured out how to get their costs down to that point. We know which country’s industry won. If it turns out that the market won’t bear any price, then there won’t be an industry at all. If there is some price that the market will bear that’s below the current $18 CD, then on the face of it there is a large incentive to come up with a cost structure that makes a profit at that price.

3. I hope in a future post you’ll attempt the difficult task of working out when a reasonable extension of copyright becomes an arbitrarily long extension of copyright. I think 95 years after the death of the creator vs no copyright whatsoever is a false choice. I also think the longer the various industries push the idea that a century-long monopoly is the right incentive, the more people will slide over to the view of chucking the whole thing.


Jack 09.16.03 at 8:12 pm

Telephone companies, sweet manufacturers and others do manage effective micropayments. Of course they achieve this by solving a smaller problem, reducing the number of potential recipients of payments for example. The music industry could find itself a role by achieving a similar effect — a Universal/Sony/EMI music account perhaps.
There are other possibilities for bundling sales to increase transaction size but keep the penny a song rate. Twenty cents (a dollar even) an album an album say, subscription to everything a band makes digital including live recordings say.
On the actual costs as far as security goes clearly the costs do not simply follow Moore’s law but the greater accuracy and simplicity of creating records will reduce costs, eventually substantially. Already the gap between current practice in the UK or US say and best possible is huge (Estonia’s banks built from scratch or Sweden’s same day money transfer for instance).
Finally although I accept a certain “per payment” fixed cost I don’t think it is a fatal as all that. In principal fraud costs are also per transaction. The splendid Superman example depended upon stealing the half cents and other roundings. This was a real problem for old computer systems but the systems built to prevent round tripping of post monetary union/pre euro currency exchanges already show remarkable efficacy. I guess progress will be relatively slow but I thnk there is a real possibility that Apples iMusic store and Universal’s price cuts may mark the end of the binary opposition between the two views and in the end music will be cheaper and sold differently.


Jack 09.16.03 at 8:14 pm

…and what doug said while I was typing.


PG 09.16.03 at 8:26 pm

I think it’s mostly the fault of the generation that grew up in the ’80s: Music for nothing and information for free


Abe 09.16.03 at 8:51 pm

Whole credit card argument is bunk, all you need is a prepaid system where you can refill your micropayments in $5 increments. BitPass does this already [ http://www.bitpass.com ]

Scott McCloud is selling comics online for $.25, heard its going ok. He’d make a lot more selling them for $1 though… [ http://www.scottmccloud.com/comics/trn/intro.html ]


--kip 09.16.03 at 9:15 pm

I don’t know the carefully guarded details of BitPass’s system, but they’re convinced they’ve solved the technical problems noted above–so convinced that it’s in a successful beta test, and the arguments against it swirl around users’ psychology and competing with free. Scott McCloud effectively responds to Clay Shirky’s dismissals; Henry Jenkins offers additional commentary.

But again, none of these address the basic concerns Daniel brings up, and I can’t do much more myself than say they say they’ve solved it, and that Scott McCloud, who’s been looking for a viable micropayment system since before Jakob Nielsen got a starry pie in his sky, said “They’ve finally figured it out.” So. Time will tell, and all that.


anon 09.16.03 at 9:27 pm

Hurricane Electric (a well known hosting provider) offers a $300 package, with 300 GB of bandwidth. If each transaction is 100kb (not unreasonable, given a decent protocol), that’s $.0001 per transaction. Of course, that’s assuming 3 million transactions a month, but Apple has already had nearly that rate at a buck per song. It’s also assuming the server can handle 3 million transactions per month, but I don’t see why not, given that this is only slightly over 1 transaction per second. Finally, it assumes that it’s OK to batch transactions with the credit card companies in some way (since CC companies want to take $0.20 per transaction).

If it’s so easy, why has nobody done it? Well, they have. Bitpass is a new system which does the batching thing. It seems to actually work, with almost the level of transparency one wants.


The Philosophical Cowboy 09.16.03 at 9:43 pm

Complete digression, but based on what you’ve said:

I suspect that a number of the roles that the Guardian article suggested would disappear would still remain (in lesser numbers), even in a totally honest world.

I take “total honesty” to mean – “doesn’t knowingly break the law or cheat others”.

However, that would still leave a number of traditional policing functions (traffic direction, people whose speedo’s are broke, people driving dangerously/carelessly without knowing). It would leave a role for checkout staff who could make change and ID what the costs of your goods were (unless moving to a purely cashless society), as it’s too hard to tot up yourself*.

Moving onto my bit, the accountants – wouldn’t need auditors, per se, though there would be substantial room for system error leading to mistakes, which they do. You’d have a role for tax accountants, as a complex, but honestly played, system would still be hard to deal with. You’d have a role for accountants preparing accounts, as the purpose of them isn’t just to say if managmeent are doing their best – it’s to show if it’s the best practical or if investors would rather invest elsewhere, or let them make decisions, etc.

Bus conductors – well, it’s a convenient way to pay the fare.

Examples about payment/cash handling rather depend how full blown a revision you expect. But, e.g., on a busy bus, it may well be more efficient to have a person taking money and making change than any other solution. Individuals doing it will be slower, and if you had pots of change (say) in seats, that still leaves standing passengers.

I.e. – the transaction costs in society aren’t all about honesty – some are about convenience to all parties. It may be easier to have cash taking specialists, than to have to do it yourself in time pressured situations…

* could have scanners as go round, etc, but would require human back uP)


Matt McIrvin 09.17.03 at 1:36 am

How does iTunes do it? Speculation I’ve seen (based on credit-card bills) suggests that Apple has some clever way of chunking multiple transactions by the same user, so that while they look like a bunch of individual one-dollar-and-change purchases to you (including tax), they look bigger and less frequent to the credit card company.

Of course they need to handle the case in which you buy one song and never buy another, but to reduce their overhead all they need is to increase the average transaction size. It may be something as simple as grouping the transactions a few days at a time, since people tend to buy several songs at once.

I suspect that this algorithm, whatever it is, is a large part of what makes the iTunes store possible.


Russell L. Carter 09.17.03 at 5:51 am

The entire post can be summarized as:

A byte stream of size ‘x’ transmitted over the TCP/IP protocol ought to have a cost ‘y’ associated with it, billable to somebody, and all civilization will fail without that. (Implied: we need to bulk up the caching somehow so that intermediate vampires get their share).


dsquared 09.17.03 at 7:10 am

iTunes: that’s not a micropayments system. You make one, quite large payment and then consume the goods in small units.

Bitpass: well, maybe, but as I say, I’ve seen it before and I am somewhat sceptical at present, particularly as they have not released any technical details at all. Many systems have worked in beta tests; the point is that there’s all the difference in the world between a system confined to a small group of vetted users, and one available to the general public. How many professional credit card fraudsters formed part of the test group? According to their website, they’re “live” at the moment, but scalability is everything in payment systems, and there’s no evidence that they’re breaking even.

Scott McCloud: I think he is just part of the test group for Bitpass.

Note that it’s easy to come up with all sorts of subscription-type models, but if you’re doing this, you should be thinking about subscription models for the entire music industry rather than assuming that the payments would look reasonably like normal payments but smaller.


Kieran Healy 09.17.03 at 7:37 am

Tunes: that’s not a micropayments system. You make one, quite large payment and then consume the goods in small units.

It’s not? You pay 99 cents a song, and that’s it. You don’t have to deposit any money in an “iTunes account” or pay a subscription fee or anything of that sort. Where’s the “quite large payment”?


dsquared 09.17.03 at 9:07 am

Hmmmm maybe I’ve got the wrong end of the stick. I’ll have a look and find out what the wrinkle is.


dsquared 09.17.03 at 9:34 am

Right, a bit of digging later (not much though, because I use a proper person’s computer and thus couldn’t have a proper look at it) reveals that these are actual credit card payments (not micropayments in any techie sense) and thus Apple is wearing the 25c minimum fee. I think that they are either

a) relying on bundling the transactions on a daily basis and taking advantage of peoples’ tendency to buy a load of tracks at once (you only have to buy three 99c songs in one go for it to turn into a macropayment)


b) losing money and charging it to the iPod marketing budget. At 500,000 downloads a week, assuming that Applie is wearing the entire 25c (unlikely as they are also trialling a few very aggressive DRM measures for the record labels), that’s a cost of $6.5m, which doesn’t strike me as necessarily way out of line; it would buy you 16 30-second slots during “Will and Grace” or maybe an ad a week in a less expensive show. (Christ this document is interesting.)

My guess is more a) than b); offering the 99c single-song option makes the whole service more palatable to people who end up making macropayments. I’m not sure the idea scales to yer class micropayment environment, and it also depends on Apple having access to 200,000 songs that *everyone* wants, rather than 200,000,000 songs that *someone* wants.


Kieran Healy 09.17.03 at 12:37 pm

these are actual credit card payments (not micropayments in any techie sense) and thus Apple is wearing the 25c minimum fee.

OK, I see more clearly what you mean by “micropayments” now. You’re right that when I buy from iTunes I often buy more than one song, but the fine-grainedness is one of the things that makes it so attractive: this song and this one but not that one, etc. I don’t know whether 90% of Apple’s sales come from 1% of its songs or something. Possibly.

At the same time, if Apple can make a profit on an online store that sells 99c songs and allows me to buy them one at a time — well, music just got a hell of a lot cheaper and easier to buy from my point of view. So even though penny-a-song can never work, it’s still a lot closer to the micropayment vision than hiking in to HMV and dropping $18 on a CD.

(not much though, because I use a proper person’s computer and thus couldn’t have a proper look at it)

Should I read “proper person’s computer” to mean “Wintel Box”? I fear so. I pass over the flamewar invitation in silence, because I know better than to get into arse-kicking contests with porcupines. Besides, I need to go buy a song or two — I’ll tell you what they are and you can hike off down to HMV, buy the CDs, and play them on Windows XP, assuming it’s not on the blink again.


Andrew Edwards 09.17.03 at 12:48 pm

The ITunes point is a fair one.

Why not bundle together a large number of songs into a single transaction. I.e. I pay you $20.30 for 40 songs I’ve selected off a list. That’s a single transaction, and the margins can clearly cover the transaction costs, and it’s only $0.50 a song.

Or you can charge a bandwidth fee. An mp3 is around 3.5MB, so charge a fixed amount (Say $20) per hundred MB I download through Kazaa. Use a Neilsen-style homescan sample of 5000 people to determine what’s being downloaded, and distribute the funds to artists accordingly.



Jack 09.17.03 at 2:03 pm

Even supposing that Apple haven’t negotiated a better deal than 25 cents a pop, it’s still better than the music industry’s normal physical distribution costs which have to bear inventory, packaging, transport and the huge cut given to retailers. They’ll still be able to make money even with an average of say ten per cent going to the banks.

Telephone companies know that the fixed costs identified for micropayments can be reduced if they do not have to be everyone to everyone.

even single song iTunes sales are larger than ‘true’ micropayments,
the money doesn’t go directly from customer to artist but from customer to Apple via an existing financial relationship (.Mac subscription often) to small number of music company acocunts and then via a prexisting channel to the artists where Apple will provide excellent statistics compared to physical invoicing.
record companies do not have to extend credit to Apple.

Music companies may actually be under most pressure from the record stores, they take the biggest bite out of the price of a CD and are the biggest sales channel. CDs have to compete with DVDs and video games. If they can’t provide revenue they’ll lose shelf space and sales.

If record companies find new channels, like iTunes, that work, they will be less frightened and if those channels are more efficient than record shops they will not be under the same pressure to maintain the current per track price. Then the price elasticity arguments may start to tell. They will also no longer need to focus on a limiting the product range to make inventory management easy.

I’m sure even now Apple could charge less than 99 cents and still make money and, just as happened with tapes and video recorders, record companies will relax when they are making money from the new world. Current hostility is surely partly due to pressure from their existing channels since many of the arguments would benefit the record companies as much as consumers.

Maybe Apple will be Dell to the music industry’s Compaq and IBM.


dsquared 09.17.03 at 3:37 pm

Even supposing that Apple haven’t negotiated a better deal than 25 cents a pop, it’s still better than the music industry’s normal physical distribution costs which have to bear inventory, packaging, transport and the huge cut given to retailers. They’ll still be able to make money even with an average of say ten per cent going to the banks.

Not wanting to be a dick about this, but 25c out of 99c is more than ten per cent.


Jack 09.17.03 at 3:54 pm

and the mean transaction involves more than one track


ethel g 09.17.03 at 5:26 pm

Isn’t Public Lending Right (*) an actual, existing, micropayments system? While it’s based on a central fund personally I don’t think that should necessarily disqualify it.

The interesting bit about PLR is that it doesn’t claim to be exact, but (I believe) authors and lenders are broadly happy with the results.

Similarly but more rigorously, one of the RSA (crypto) chaps has proposed a probabilistic micropayment system which I think does address the transaction costs issues: precisely because you’ve got _lots_ of micropayments, you can use the law of large numbers to get good enough estimates of amounts due.

(*) For non-UK readers, PLR is a system whereby authors of books in public libraries receive money. I think monies received are dependent in some (monotonic, increasing) way on the number of times a book is loaned.


Robert Nagle 09.17.03 at 6:10 pm

This is a fascinating piece. I wrote about similar issues on sharethemusicday.com .

If it is cheaper to run the system as “free” than as “for sale”, then that seems to be a reason for supporting a voluntary system of compensation.

Actually all this bickering about pricing ignores one fact: the overwhelming majority of music out there in the world is available for free on IUMA, Amazon.com and mp3.com. Young artists at the moment don’t have any realistic hope of getting a deal with a major label (if any label at all).

The only ones selling songs are those backed by big marketing machines.

One unique aspect to music as content is that its value cannot be assessed until after the music is heard once or twice. That is another reason why a voluntary system works best–the transaction is made after the song is downloaded on the PC (usually).


Thomas Dent 09.18.03 at 11:15 am

This may be a bit iconoclastic, but I don’t understand why recorded music is such a good thing after all, or at least why we should automatically accept that it is.

Classical music has been suffering from recordings for decades. The more recordings, the fewer practicing musicians and the less musical tradition and culture there is. (I’m not claiming this correlation as a proof of causation though.) Recordings also promote standardised interpretations, the false ideal of the ‘perfect’ interpretation, and the idea that it is acceptable to listen repeatedly to the same piece played exactly the same way. Live performing artists even get inhibited by having to ‘live up to’ their own recordings.

I say this as someone who owns hundreds of excellent classical music recordings. Unfortunately, many of the best of them are from the 60’s and earlier. It is somewhat depressing that in the vast majority of places live classical music is entirely overshadowed by recordings.

Comments on this entry are closed.