Amazon.com ? 2 August 1999
3
revenue, and 2) the company?s existing customers are
buying less per customer than they did in comparable
quarters last year and the year before. As will be discussed
below, both of these trends were expected, and neither
means that our aggressive projections for Amazon.com?s
future growth (which call for revenue of $3 billion in 2000
and $10 billion in 2003) are necessarily too optimistic. For
the company to achieve these projections, however, the
second of these trends?declining revenue per customer?
will have to stabilize or reverse itself within the next 12-18
months. If it doesn?t, we will almost certainly have to trim
our aggressive forecasts.
Flattening customer acquisition velocity?and increasing
customer acquisition cost. There are two basic drivers of
Amazon?s revenue: number of customers and revenue per
customer. For all of the company?s history thus far, the
number of new customers added in each successive quarter
has increased significantly (better than 20%) each quarter,
allowing the company to maintain strong revenue growth
even as revenue-per-customer has declined. In Q2,
however, the number of new customers added was only
5% greater than the number added in Q1 (versus 36% last
year), and we would not be surprised to see the number of
new customers added decline for the first time in the
company?s history in Q3. As a result of flattening of new
customer growth, existing customers are contributing a
greater percentage of the company?s revenue each quarter.
Existing customers contribute less revenue per customer
than new customers (because the average existing
customer makes a purchase less than once a quarter,
whereas all new customers?by virtue of being new
customers?make exactly one purchase per quarter), and
the mix is contributing to a decline in overall revenue per
customer. At the same time, Amazon?s customer
acquisition cost appears to be increasing (our estimates
suggest that it rose from $13 to about $20 from Q1 to
Q2?the lowest in the industry, but almost the highest it
has ever been), so the slowdown in new customer
acquisition is not likely to be rectified by increased
investments in marketing. Based on the last quarter, we
think it is likely that Amazon.com is reaching a point
where the growth of new customers will stabilize in the 2
million ? 2.5 million range?a prodigious rate of growth,
but a flattening in the trajectory.
Declining revenue per customer, both new and existing.
We model Amazon?s business in part by tracking and
projecting quarterly revenue per customer, a metric that we
regard as being similar to the same-store-sales metric in
traditional retailing. On a year-over-year basis, revenue
per customer has been declining for the past two years?a
trend that we expected and are not necessarily concerned
about (we will discuss the reasons for this decline below).
For the company to hit our aggressive projections,
however, and begin to benefit from the leverage inherent
in spreading customer acquisition cost across multiple
product lines, revenue per customer must stabilize and
then begin to increase within the next 12-18 months. If it
doesn?t, our aggressive projections will prove to be too
high.
To really understand the revenue-per-customer dynamic, it
is necessary to divide Amazon?s customers into two
groups?new and existing (new customers are those that
make their first purchase within the period in question).
Of these, the far more important group is existing
customers, as the new customers will continue to constitute
a smaller and smaller percentage of the active customer
base in any particular period. As many Amazon.com
observers have noted recently, not only revenue per new
customer but revenue per existing customer has been
declining for the past two years. Amazon?s goal is
ultimately to gain greater ?wallet share? with each of its
customers, by enticing them to buy many different kinds of
products. At some point, therefore, the revenue-per-existing-
customer trend will have to reverse itself (in our
aggressive model, this happens in Q3, 2000; in our official
model, it happens in Q1, 2001).
To more easily assess the likelihood that the revenue-per-existing-
customer trend will stop declining and begin
increasing, it makes sense to analyze why it is declining in
the first place. In our opinion, the best explanation stems
from a comparison between the habits of the early-adopter
customer and more mainstream customer. Two years ago,
in Q2 1997, we estimate that the average Amazon existing
customer bought $41 worth of stuff (books) in the quarter.
A year later, the average purchase had fallen 21% to $32,
and this year, it fell 21% again to $26. Over the same time
period, Amazon?s base of existing customers grew from
340,000 to 8.4 million and its product categories increased
to include books, music, videos, auctions, and gifts. We
believe that the main reason revenue-per-existing-customer
is falling despite the addition of additional product
categories is that the customers added more recently
simply buy less on average than the early adopters who
discovered the company back in 1997. If this trend were
to continue indefinitely, obviously, the company would be
in trouble?each new customer would be worth less than
the last, and the company would get stuck in the law of
decreasing returns. As the more mainstream customers
come to represent a greater and greater percentage of the
total base, however, we believe that revenue-per-existing-customer
should begin to stabilize (according to
management, this is beginning to happen).
The recent addition of Toys and Electronics to Amazon?s
product offerings, as well as a reported stabilization in
revenue-per-customer of recently added existing
customers, should help stabilize and ultimately reverse
Amazon.com?s overall revenue-per-customer trend. In Q3,
for example, our aggressive projections call for revenue-per-
customer to decline 18% year-over-year to $28
(slightly more than the 17% decline in Q2), and then
decline only 7% in Q4 and 3% in Q1. As mentioned, our
aggressive model, which calls for $3 billion in 2000
revenue, assumes that revenue-per-customer will begin to |
