To start, we ran a few numbers. Forrester Research Inc., the most frequently cited forecaster of online sales, suggests the U.S. online-consumer market will exceed $184 billion by 2004 — whopping growth over last year's $20 billion to $30 billion estimate. (No one has exact figures; the government only recently began tracking this new phenomenon.) But according to Forrester, only 60 percent of the 1999 total required physical delivery of goods. Digitally delivered goods, such as airline and event tickets, online brokerage and banking services, plus "researched goods" like automobiles, which have a separate delivery network, accounted for the other 40 percent. The physically delivered categories contribute much of the future growth and account for $132 billion of Forrester's 2004 forecast. To put this in perspective, that's 2.3 times the $57 billion that consumers currently spend annually on catalog purchases. That is enormous growth in a short period of time. But we worry that even that much home-delivery volume will not provide enough sales density to alter fundamental delivery economics.
For further perspective we built a forecast model that highlights the two key drivers of local delivery economics: sales concentration and population density. Specifically, we selected a range of U.S. cities — many covered by announced expansion plans from the key local deliverers listed in Exhibit 1 — and examined both current sales density and a forecast for 2004. Our model builds on Forrester's Internet sales forecasts, published data on Internet penetration rates in key cities from Scarborough Research, population and median income data from the U.S. Census bureau, and the one constant: the land area of the cities.
For those not familiar with the consultant's perennial favorite information graphic — the bubble chart — Exhibit 2 warrants more explanation. The horizontal axis indicates the number of Internet users per square mile in each city on a logarithmic scale. (Note that the logarithmic scale represents a tenfold increase in user density for each increment along the scale, rather than the more typical increase of a fixed linear amount.)
Population growth (and in some cases decline) and Internet penetration rates drive the migration to the right between 1999 and 2004. For example, in 1999 an estimated 60 percent of the 4.4 million inhabitants of the Washington, D.C. greater-metropolitan area had Internet access — the highest penetration of any U.S. city. Since the Washington, D.C. area covers nearly 3,500 square miles, those 2.6 million users produce a user density of around 750 per square mile. By 2004, increasing Internet penetration and population growth drives the user density to more than 1,300 users per square mile. The vertical axis estimates the Internet purchases of those users. Again using Washington, D.C., as the example, and looking ahead to 2004, we adjusted the average online sales-per-person to reflect the above average affluence in the district. This yields a projection of online physical-product sales of nearly $700 per user in four years — well above the overall 1999 U.S. average of around $125.
The size of the circle captures an important third variable: market size. Though high user density and high sales-per-user drive sales density, the overall market size also plays an important role.
Exhibit 2: Internet Sales-Density Analysis
For example, Denver, with a user density of nearly 1,300 per square mile and projected 2004 sales per user of $800, will have $1.3 million of sales per square mile per year. Washington, D.C. fares worse with sales per square mile of only $900,000 — but offers a bigger prize due to the greater total population.
The curved lines on the chart highlight different combinations of sales levels and user density that yield the critical value of sales per square mile. The $14 million in sales per square mile in New York City offers an attractive market; even dividing by 365 days per year for the 24/7 Internet economy, the revenue potential totals more than $39,000 per day per square mile.