Interviste
The Siren Song of Internet Micropayments
The Siren Song of Internet Micropayments
Steve Crocker was one of the founders of CyberCash and a co-inventor of
CyberCash's CyberCoin service. In the late 1960's and early 1970's, Dr.
Crocker was part of the team which developed the protocols for the Arpanet
and laid the foundation for today's Internet. In addition to his technical
work, he organized the Network Working Group, the forerunner of the modern
Internet Engineering Task Force, and initiated the Request for Comment
(RFC) series of notes through which protocol designs are documented and
shared. Steve Crocker Associates, LLC is a consulting and R&D company
specializing in current Internet and electronic commerce technologies.
The Vast (?) Market for Micropayments
Would you enjoy making small payments across the Internet? Imagine paying
a nickel for a stock quote, a quarter for a news article, a dollar for
a picture or a song. Over the past several years, a number of Internet
micropayment schemes have been invented, and some of these have been put
into service. Instead of revolutionizing commerce on the Internet, the
total impact has been negligible. What happened? Is the concept flawed,
or were the specific efforts just not right? Is it an idea whose time
has not yet come, or is it forever doomed?
I was one of the founders of CyberCash and a co-inventor of CyberCash's
CyberCoin service. We had great hopes for this service. We imagined a
vast market in which sellers would provide a wide array of digital goods
-- music, information, pictures, games -- and customers would make small
purchases over the Internet -- a page of a book, a single song, or the
cost of sending in a membership. Costs over the net could be disaggregated
into small components rather than bundled into larger prices. The implications
for defining transaction costs and for establishing a low threshold of
entry for small and low-budget vendors (like libraries, for example) were
and remain potentially provocative. The approach, however, required creating
new third party service to handle the financial transactions on behalf
of both buyer and seller.
At CyberCash, we targeted the purchase prices to be in the range of $0.25
to $5.00. We built the system and began service in 1996 in the U.S., and
later in the UK, Germany and Japan. The volume of transactions has been
very low. CycberCash is suspending its CyberCoin operation in the U.S.
but may revive it when there's a stronger market for it.
CyberCash was not alone in this vision of a market, and its experience
is not unique. First Virtual Holdings brought out a system for small payments
in 1995. That business is now gone and First Virtual has transformed itself
into an entirely different business. Carnegie-Mellon University created
Netbill, whose commercial rights were acquired by CyberCash. Digicash
in the Netherlands pioneered a method to make payments anonymously, and
put it into service in Germany and the U.S. Digicash filed for bankruptcy
and has ceased operation. Digital Equipment Corporation developed its
Millicent system, ran some trials and never came to market. It is possible
that Compaq is reconsidering it after its acquisition of Digital, but
no announcements to this effect have been made. IBM developed a system
in its labs but has not brought it to market. Mastercard invested in Mondex
and worked on bringing Mondex payments to the Internet. That effort stalled
and has dissipated. Visa built and tested an Internet micropayment system
but has also not yet brought it to market. Many more schemes have been
invented and some of these have been tested in the marketplace, but none
has gained a secure foothold.
How could so many of us have gone astray? I'll draw primarily from the
experience at CyberCash, but many of the issues are the same for other
systems.
Why Micropayments? Why Not Just Use Credit Cards?
Credit card payments rule the Internet today. An obvious question, then,
is why try to invent something new. The short answer is credit cards are
unprofitable for the seller at purchases below $5.00.
In the U.S. last year, there were more than $900B worth of credit card
transactions including face-to-face, mail order, telephone order, and
Internet purchases. The average transaction was around $80. The merchant
pays a fee on each transaction. This fee is divided among the merchant's
bank, the cardholder's bank, the card association (Visa, Mastercard, etc.),
and the various companies operating behind the scenes to process the transaction
and move the data. The aggregate fee is usually between 1.5% and 3.5%,
depending on the risks and costs involved. The average is about 2.2%,
but Internet companies usually pay more. Merchants with large volumes
of high value transactions with no complaints or chargebacks enjoy a low
discount. Merchants who present a lot of risk, incur a high rate of chargebacks
or whose average transaction size is low generally pay a high discount.
Thus, risk itself -- always an issue in a new venture like net commerce
-- increases the threshold cost. Which is precisely the problem in small
payments.
Although the discount rate is usually referred to as a percentage of the
transaction, often with a few cents added on as well, there's an underlying
factor that is slightly less visible. From the banks' point of view, part
of the cost of processing a transaction is the same for small transactions
as for large transactions. The system as a whole needs to charge an average
of a dollar or more for each transaction, irrespective of whether it's
presented to the merchant as a percentage of the amount or a fixed amount
per transaction. A large portion of this base cost covers the cardholder's
bank's cost of operation -- issuing cards, setting up accounts, billing
etc.
A merchant who creates a large number of very small transactions discovers
his bank will raise his discount rate in order to recover its baseline
costs. Hence, it is not uncommon for a merchant to refuse to accept a
credit card for transactions below $5.00, $10.00, or sometimes $15.00.
The implications are clear. The transaction costs implied in the credit
card financial structure mean they cannot be used directly for very small
purchases. If each transaction costs the merchant a dollar just to process
the credit card payment, he can hardly afford to sell anything for a quarter!
A related but no less important "cost" is the time it takes
to authorize a credit card transaction. In the U.S., credit card authorizations
take between 6 seconds and 90 seconds. Outside the U.S., the authorization
time is often longer. A customer who makes an $80 purchase waits patiently
and usually makes only a few purchases in a single day. In contrast, a
customer who makes small purchases may make several in a single day and
will not want to wait more than a second or two for each one. Imagine
waiting 6 to 90 seconds 10 times a day in an environment that seems to
promise real-time interactivity. Then, imagine waiting during periods
of peak use -- at 6 p.m. the week before Christmas.
Designing a Micropayment System: Aggregation Is the Key
Thus, micropayment systems posed several kinds of technical requirements.
The ventures mentioned above all use some form of aggregation that reduces
both the cost and the time delay for any single transaction. In CyberCash's
CyberCoin system, the user created an account, loaded it with funds --
usually $5.00 to $20.00 -- from his checking account or credit card, and
then spent against the funded amount. Each transaction involved an interaction
among the user, the merchant, and the CyberCoin system, but individual
transactions did not require any interaction with the banks or the credit
card system.
The CyberCoin system was designed to make the purchase transaction very
fast and efficient. Of particular concern was the number of messages sent
between the user, merchant and CyberCoin system, the number of cryptographic
operations needed to assure authenticity of the transaction, and the number
of database operations needed for each transaction. We chose each of these
to match what we perceived the need to be. We chose cryptographic algorithms
which were strong enough to protect small value transactions but which
could be carried out reasonably quickly on our computers. We reduced the
number of messages sent among all the parties, and we chose a database
design that allowed us to respond very quickly when a merchant forwarded
an authorization from a user.
From a technical perspective, these strategies appear to have been effective.
When the system operates at high volumes, the cost of an individual transaction
is low, perhaps a fraction of a cent. Further, the system adds only about
a second to the total transaction time.
Assured Delivery Reduces Customer Service Cost
Another source of expense in any payment system is the cost of handling
complaints and questions from the consumers. A customer service operation
requires many people, telephones, and computers. Customers are usually
confused or irritated, so the interactions are far from efficient. Each
customer service call costs a few dollars, sometimes more. If there is
no way to resolve a dispute amicably and quickly, there might be several
calls associated with the same transaction.
Customer service becomes a serious burden in the micropayment world. Even
if the nominal profit margin is high on individual micropayment, the absolute
margins are rather thin. A single customer service call can easily wipe
out an entire year's profit on that customer account. And an unhappy customer
can put a dent in company's profitability. Therefore, one of the design
goals in each of the micropayment systems was a reduction in the number
of customer service calls.
One potential source of dispute in online transactions is that the customer
may claim he did not receive the digital goods he paid for. Several micropayment
systems included some form of "assured delivery," approximately
equivalent to the U.S. Postal Service's certified delivery service. To
assure delivery, the payment process is intertwined with the delivery
of the digital goods. For example in the CyberCoin and Netbill systems,
encryption is used to assure delivery.
When the user requests the information, perhaps a page from Gray's Anatomy
for a medical student who needs it for a paper, the "goods"
are sent back to the user's system in an encrypted form before the user
pays. This puts the information on the user's computer but does so in
a form the user cannot read.
Once the encrypted information is on the user's computer, the user's system
sends back an authorization for payment, and the merchant's system forwards
this authorization to the clearing center. The merchant's system also
includes the key necessary for decrypting the digital goods. The clearing
center debits the user's account and credits the merchant's account. When
the merchant's system receives acknowledgement from the clearing center
that payment is complete, the merchant's system forwards the key to the
user. The user's system then decrypts the digital goods and displays or
stores the result.
If the merchant's system fails to deliver the decryption key to the user's
system, the user can contact the clearing center to obtain the key. The
user's system maintains a copy of recent transactions, so it can prove
that it received the encrypted information and that it authorized payment.
This information is sufficient to show that the user paid for the goods
and is entitled to decrypt them, thus providing protection from failures
in the merchant's system or potential fraud by the merchant.
How well does this scheme work to reduce customer service costs? No one
knows. There has not been enough volume in any of the micropayment systems
to gauge how large customer service costs are or whether assured delivery
is an important way to reduce those costs.
What Went Wrong?
As noted in the introduction, neither the CyberCoin system nor any of
its competitors has made much progress in the market. What went wrong?
We know some of the reasons, which I list below. But far more important
is what we do not know. I am convinced that it is not sufficient to simply
fix the problems we know about.
Too much authentication
The early micropayment systems were clumsy to use, mostly because the
designs erred on the side of caution. In the early implementations of
CyberCash's CyberCoin system, the user was presented with multiple screens
for each transaction. This level of caution is appropriate for higher
value purchases, but it slowed down the interaction. The user did not
find it a compelling or gratifying experience.
A more onerous hurdle was the loading of the user's account. In the CyberCoin
system, we experimented with two forms of payment, checking accounts and
credit cards. The credit card payment was fairly easy, but many users
were, nonetheless, reluctant to pre-pay $20.00 to load their CyberCoin
accounts. Payment from a checking account was significantly worse. Before
we could give users access to their checking account online, we required
a canceled check and other proof of identity by mail. Very few people
went to the trouble of setting up a CyberCoin account tied to their checking
account.
Later, these barriers were lowered. The payment process was streamlined.
In other systems, a user was permitted to accumulate charges first and
then have them charged against his credit card. These were a big improvement,
but they didn't prove sufficient.
Too few merchants
As with all network systems, a critical mass of users, sellers as well
as buyers, is required. Bob Metcalfe suggested the value of a network
grows as the square of the number of users. All of the early micropayment
systems failed to gain a critical mass of merchants and a critical mass
of consumers. These two groups reinforce each other, of course. In all
of the early ventures into micropayments, there was neither a single compelling
digital good available only through the micropayment system nor an overwhelmingly
attractive variety of digital goods.
An Internet micropayment system might yet come into widespread use if
pushed by a large marketing power. We can imagine that AOL, Visa, Mastercard,
or a collection of large banks might have sufficient marketing power to
entice a substantial number of merchants to come into the market and make
their digital goods available in small transactions. On the other hand,
this may be exactly backwards. The banks or other financial services companies
are more likely to provide a successful micropayment system when the merchants
demand it. Perhaps the critical mass needs to come from a collection of
large publishers who view micropayments as essential for their growth
and health.
Subscriptions and Advertising -- Alternatives to Micropayments
Once upon a time we thought micropayments were the only way merchants
would deliver small quantities of digital goods. However, two very important
alternatives emerged: subscriptions and advertising. Both of these have
substantial advantages from a merchant's point of view.
Subscriptions provide guaranteed income. If a merchant were to choose
between selling digital goods on a one shot basis and selling on a subscription
basis, the merchant will choose subscriptions. Although the higher price
for a subscription will mean that some of the potential buyers will be
lost, it also means that a higher price will be extracted from a number
of buyers. Moreover, the guaranteed income gives the merchant much greater
ability to plan, provision and grow.
Subscriptions have another and perhaps even more important quality. They
provide considerable information about the consumer and establish a relationship
between the consumer and the merchant. Consumer information is extremely
valuable in its own right, as it enables to the merchant to target further
sales pitches more accurately -- "up-selling" and "cross-selling"
-- or to sell this demographic information to others for similar purposes.
The relationship also generates renewals, thereby continuing the income
stream in future years.
Compared to subscriptions, selling individual items on a one-shot basis
is a secondary concern. Micropayments are useful as a way to draw in new
customers and to serve casual customers who will not pay for subscriptions,
but it's unlikely to be preferable to subscriptions.
Advertising, too, has been an extremely successful alternative for funding
the delivery of small digital goods. Stock quotes and news stories, in
particular, are easily available on the web free of direct charge. Web
searches are also free, though they cost the search companies quite a
lot of money to provide. Instead of a few cents for each transaction,
the users pay attention. That attention may be far more valuable to the
merchant and his advertisers than the few cents the merchant might collect.
In this respect, the web is similar to broadcast television.
Reverse Micropayments, The Internet Surprise
All of the early micropayment systems were aimed at collecting small amounts
of money from large numbers of consumers in exchange for digital goods
of (presumably) real value from a relatively fewer number of merchants.
As described, none of these has achieved any large volume.
An interesting reversal seems to have occurred, however. Several companies,
Cybergold, Netcentives and ClickRewards among them, have been making small
payments to consumers instead of collecting small payments from them.
These payments are either for attention to ads (Cybergold) or rewards
for patronage (Netcentives and ClickRewards). Compared to the several
micropayment systems, each of these "reverse micropayment" systems
is overwhelmingly successful.
Cybergold is a particularly interesting case study. It offers users anywhere
from fifty cents to five dollars to read an ad. Users have to register
and describe themselves. The payment varies according to the material
and the demographics of the user. An ad for a Cadillac will be worth far
less to a college student than to a middle-aged, high income executive.
Cybergold reports that it has hundreds of thousands of users and has accumulated
around a million dollars in payments, most of which is sitting in its
accounts waiting to be disbursed to the users or consumed in some fashion.
In an effort to provide additional outlets for its users to use the funds
they have accumulated, Cybergold is now entering the classical micropayment
business. It remains be to seen whether it reaches critical mass and creates
a sustainable system.
That reverse payments may prove the first commercial success of the micropayment
model should not surprise me. The trajectory of the Internet has surprised
us all, even those of us who worked on it when it was an experimental
system among a relatively small group of researchers. That it may transform
the way we work and our economic structures is likely. What we have learned
from micropayments is that new technologies do not change people and organizations
overnight, that change is hard, and that the future holds many surprises.
Released: April 22, 1999