|by Sebastian Heath|
This month’s column takes a broad look at the usage patterns of the ANS web-site. But before I get into the details I need to give some background on how this is done. The ANS web-site is divided into two main parts. All the informational pages, such as opening hours and descriptions of the collection, are hosted by our Internet Service Provider (ISP). The machine that makes the searchable databases of the collection and library catalog available is kept at the ANS. Users of the site don’t know it but sometimes they are accessing the informational pages at our ISP and other times they are pulling information directly from the ANS’ own servers.
Each time somebody reads a page or performs a search in one of the databases, a line is generated in the log file of the machine providing the information. This log file lists, among other things, the address of the machine making the request and the file or search that was requested. Before going any further it is very important to say the ANS does not try to track individual users or do anything to determine the specific identity of those using our site. The main purpose of the log is to allow us to make sure that the site is running correctly but looking at the information it provides in aggregate allows some interesting patterns to emerge.
The last full month for which the log files are available is April 2003. Over the month’s thirty-one days, the ANS web-site received a total of 166,933 hits. This doesn’t count use by the ANS staff. That number is deceiving by itself, however, because a single person reading a single page can produce many hits. For example, the results of a database search that has twenty coin images will produce twenty-one hits: one for the page and one each for the images. A more useful indication of traffic to the site is the number of “page requests,” meaning the number of pages read without also counting any images those pages include. For April, there were just over thirty-five thousand page requests from twelve-thousand individual “hosts”. A host may represent an individual computer sitting on someone’s desk or it may be a machine that is providing access to many users. An example of the latter is the machine “cache-mtc-ac09.proxy.aol.com” that accessed the ANS site on eleven different days during the month. While it would be possible to record whether or not these hits came from the same AOL user, the Society doesn’t try to gather such individual statistics. The page request number is also useful because it eliminates most of the traffic from search engines that usually just check whether or not a page has been updated.
Thirty-five thousand pages served to twelve-thousand hosts is still a very coarse measure of the usage pattern of the Society’s web-site. One interesting measure is where the traffic is coming from. Forty percent of our traffic comes from addresses that end in either “.com” or “.net” and 8 percent from “.edu” domains. One can sometimes tell more about the origin of a .com or .net address if one looks at individual entries in the log but it is more useful to look at the addresses that explicitly indicate which country they come from. By this measure, ninety-two different nations are represented in the log. Other than the United States, the top countries by the number of hits are: Italy, Portugal, Spain, Belgium, and Canada. Almost every European country, from Albania to the United Kingdom appears at least once. The same is true for Latin America. In the Middle East, Lebanon shows up but Jordan does not. India, China, and Japan are well represented, of course, but it is interesting that Kyrgyzstan, Turkmenistan, and Brunei also appear. The only region of the world not well represented is sub-Saharan Africa, with South Africa being an exception.
One good indication of what people are doing on the site is the database usage. There were 8,058 searches of the coin database and 1,924 searches of the library catalog. In the coin database 1,587 different key words were used in searches. However, only 82 of these terms where used more than 10 times and the majority of these pertain to Ancient numismatics with a strong bias towards Roman emperors. The top ten search terms are: aachen, constantine, avg, coin, coins, genio, alexander, bithynia, ii, and providentia. This list again indicates that most of our users are interested in the coinage of the Ancient world.
The final statistic to look at in this column is the number of coin images viewed. In our test month of April the ANS served 81,403 images of coins. At the time we had about 3,500 images available so that most coins are being viewed multiple times. The five most viewed coins are all in the Greek and Roman departments and are all on display in the Federal Reserve Bank of New York. Figure 1 shows the “winner” of this contest, a denarius of 48 BC celebrating Caesar’s campaigns in Gaul that was viewed 380 times in April.
Without implementing a more sophisticated tracking system, there are limits to the amount of information we can gather about our users. But the ANS web-site is not a commercial venture looking to squeeze every penny out of its audience. Rather, it is one component of our overall effort to promote numismatics to an international audience. In this regard these numbers, while not large when compared to commercial sites, are encouraging.
Figure 1: Roman Denarius of 48 BC. Bearded bust of Gallic chieftain, shield to left. Biga with charioteer and warrior. (1937.158.248)