More important than the variety of features is, of course, the quality of search results. These will, not surprisingly, vary substantially from site to site. We have built viaLibrito give the most comprehensive results possible, particularly for early and rare books, and our testing shows that our search results are nearly always the most complete available. Sometimes the differences between our results and those of other sites can be dramatic. To demonstrate this we suggest performing the following tests.

General Searches:

        Differing results are especially likely when using keywords to find material on a specific subject, or when searching for rare editions of popular or collected authors. This can be demonstrated as follows:

        Search for anything that is NOT the author and/or title of an uncommon book (examples: Central Park; Antilles; Canterbury Tales). Take the results from each site and sort them by price/descending. Count the total number of results above $1000, or some other meaningful threshold. (If a site does not allow sorting by price you will just have to count them). Compare the results.

        In nearly all circumstances you should find that viaLibri will produce more matches than the other sites. In some cases the difference may be dramatic. For example, on a search for the keywords "Central Park" viaLibri found 134 items priced at $1000 or more just within its first set of results. Meta-search "A" found 22 items and meta-search "B" found 13 (tests run on Dec. 3, 2006). Comparable differences would be found at lower thresholds and for any other keyword searches where the selected criteria will match a large number of books.

        The difference in these results mostly reflects the difference between viaLibri's focus on finding rare books rather than cheap ones. Unlike other sites, however, viaLibri does not stop searching after its first set of results. If there are more possible matches viaLibri can continue to look for them. There are, for example, over 5000 books on the internet that will match the key words "central park"; viaLibri can display all of them for the user who has the patience to continue searching to the end.

        But that type of patience is not really necessary. Two other functions, unique to viaLibri, can help sort out the wheat from the chaff. Up to this point in our experiment we have been matching books which include both the terms "central" and "park." The two words could have appeared far apart in the description of a book that has nothing to do with the "Central Park" we are interested in. By enclosing the two words in quotation marks, ("central park"), we can filter out many irrelevant matches. This will reduce the total number significantly. It is also likely that a collector with an interest in Central Park (or any other subject) will have a cut-off date after which books will no longer be of great interest. If he only wants 19th century material he can set the date fields to return only books after 1850 (when the park did not yet exist) and before 1900. Using these filters will produce a set of relevant matches that is less than one tenth of the original total. The same general principles will apply for all but the narrowest collecting interests.

        On the other hand, for the user who is interested in finding an inexpensive copy of a specific book, an alternative method is also available. In this case, choosing the "I want... Bargains" option on the search form will assure that the least expensive copies will (with few exceptions) be the first to be displayed.

First Edition Searches:

        When searching for first editions by a collected author the usual method is to enter the author and title while also checking a box marked "first editions." The number of accurate results will, however, vary with each of the different meta-search sites. This can be demonstrated by the following test:

        Search for the first edition of a classic work of modern literature using the "First Edition" checkbox as part of your criteria. Sort the results by price/descending and then count the books which appear and which are, in fact, first editions. In nearly all cases viaLibri will have the most complete set of listings for true firsts, often including books missed by other sites. For example: on a search for James Joyce's Ulysses, with the "first edition" attribute checked, viaLibri found 16 copies, meta-search "A" found 12 copies and meta-search "B" found 8 copies. (Test conducted on Dec. 4, 2006).

        However, the "first edition" attribute assigned by the various listings sites is frequently inaccurate and using it as a search attribute will sometimes eliminate items which should be included. This is especially true of books from booksellers whose book descriptions are not written in English. With viaLibri a better selection of first editions can often be found by leaving "first editions" unchecked. When the books are sorted by date the first editions of the same title will be found together in the same year, making them easier to find. These results can sometimes be further refined by including the publisher's name in the "Imprint" field.

Searching By Date.

        While it is not possible to compare the reliability of viaLibri's' date searching features with that of similar meta-search engines (they don't include it), it is possible to compare it with that of the individual listing sites. Of the fourteen sites we include in our searches, nine offer date searching on their own search forms. The results of our own testing reveal, however, that the dates assigned by viaLibri are in most cases more accurate and reliable than those supplied by the listing sites themselves. The following tests will help demonstrate this:

        Go to one of the listing sites that allow date parameters as part of their search queries. Search using keywords that are likely to produce only books that have been dated by their cataloguer. "ESTC" (the abbreviation for "English Short Title Catalogue") is a good example, since any book that has been referenced to ESTC will almost certainly have a known publication date that would be included in its description. First, perform a search for keyword "ESTC", without including date parameters, and then record the number of books found. (For the larger sites adding a second parameter, such as "king", will give a more manageable number). Next, perform the search a second time, this time setting a date range between 1450 and 2008. On most sites this will reduce the number of matches, even though all the books returned by the first search should also be returned by the second. On a few sites the reduction will be dramatic. Repeat the same steps from viaLibri. Leave unchecked all the sites except the one you are testing. Count the different results and compare them. There should be little or no difference between the two viaLibri results.

        And because viaLibri performs its own date extraction it is able to supply dates for books even when the listing site itself does not. Some errors are inevitable in this process, but overall accuracy has been shown to be very high. The usefulness of accurate dating and sorting will be self-evident.