We have recently received inquiries from several hospital or health systems Webmasters who want to know our opinion about Web sites like
Alexa,
Compete.com, and
Quantcast.com. These “third party web data providers” use proprietary algorithms to estimate the number of visitors on any Web site, and have become a major resource for many Web marketers.
Let me start off by saying I am a huge fan of these sites. The possibilities for comparison are limited only by your imagination. Having a friendly disagreement at the ballgame about whose fans are better? Simply point out to your buddy that
ChicagoCubs.com gets 4-5 times more monthly traffic than does
ChicagoWhiteSox.com or
StLouisCardinals.com. Following the Leno-Letterman ongoing ratings war on network TV? Type in
nbc.com and
cbs.com and you can watch the ongoing saga play out in cyberspace.
Most of these 3rd-party data sites use server traffic patterns as the basis for their estimates. (Quantcast.com is unique in that it offers a script to download that will quantify the exact number of visitors, but still must rely on projections for sites that have not installed the script) With so many people using these sites, it begs the question: how reliable are these sites, and how useful are they to hospital/health system marketers?
Using real data from Google Analytics, I was able to compare the actual number of Web visitors to three hospital Web sites with the projections made by Compete.com for a recent five month period. For my comparisons, I chose three different-sized organizations (one small health system, one large one, and one roughly in the middle) located in three different regions of the country (West Coast, Southeast, East Coast).
Large Hospital
Sept Oct Nov Dec Jan
Google Analytics 308830 338770 318100 272040 329780
Compete.com 195605 216997 260732 213240 229912
% Difference 36.7% 35.9% 18.0% 21.6% 30.3%
Medium Hospital
Sept Oct Nov Dec Jan
Google Analytics 53634 54202 55721 45422 54300
Compete.com 23149 33107 25873 16954 21079
% Difference 56.8% 38.9% 53.6% 62.7% 61.2%
Small Hospital
Sept Oct Nov Dec Jan
Google Analytics 16278 16516 14293 13560 15765
Compete.com 6368 4940 4160 4509 6357
% Difference 60.9% 70.1% 70.9% 66.7% 59.7%
My comparison above was hardly scientific. I only compared three hospitals using a single tool metric over a five month time frame. It was not always possible to line up the comparison periods exactly using the free version of Compete.com. If you ran this same comparison substituting Omniture or WebTrends for Compete.com, you would likely see differences, as all Web analytics packages vary in how they tabulate Web metrics.
Nevertheless, it is clear from my analysis that the accuracy of 3rd party data is highly questionable. Reliablity does seem to increase the more traffic a site receives. This makes sense intuitively, as more data should lead to more accurate models. However,until someone can prove to me otherwise, if I were a Webmaster at one of the thousands of healthcare sites out there that get less than 100K visitors per month, I would be extremely reluctant to use a tool like this for any serious analysis.
Fortunately, there are other options available. There is a free benchmarking tool imbedded in Google Analytics that, while it does not let you get very granular in your comparisons (all hospitals are looped into a single group), does have some value. You may also want to consider a free trial Greystone.Net Benchmarking, the program that allows users to make comparisons from data directly imported from Google Analytics and allows segmentation based on hospital size and specialty. (full disclosure: I am the Product Manager for this service) Perhaps one day the “hosted script” model employed by Quantcast.com will take off, while sites like Compete.com will likely continue to improve the accuracy of their models. Until then, however, be careful about relying on 3rd party Web data providers for meaningful comparisons. Unless, of course, you are able to prove that Cubs fans are better than White Sox or Cardinal fans, in which case these sites work just fine.