Search engine optimization can have unexpected consequences for a website. In theory, SEO is a set of techniques for tuning a website in order to improve the site's search engine listings. The theory is that better listings will mean more traffic, which will translate into more business. In general, this works. However, I have discovered that it is possible to use SEO to improve the quality of search engine traffic without necessarily increasing the volume. I stumbled upon this by accident, so I don't have fancy theories about search algorithms to explain the whys and wherefores -- all I can do is provide my experience as a case study and let you draw your own conclusions.
One of my clients is London's City Cruises PLC, which runs a chain of tour boats and floating restaurants on the River Thames. One of the restaurant boats, the R.S. Hispaniola, has its own website. With 69 million results in Google for "London restaurant," competition for relevant listings is ferocious. However, having been one of the first restaurants in London with its own website, and with a continuous SEO program now approaching its 10th year, the site does extremely well in the search engines. A few years ago, it started doing a little too well.
Like many restaurant sites, the Hispaniola's contains sample menus, such as a lunch menu, dinner menu, Christmas menus, and so forth. A key USP of the Hispaniola restaurant is that it has one of the best wine collections on the Thames, so there's even a detailed wine menu on the website, with an aim of enticing wine buffs (and French tourists) to the restaurant.
All was going very well until about three years ago, when the conversion rate began to fall and the bounce rate began to rise. Investigation revealed a rise in people coming to the website searching for menu ideas -- suggestions for what to prepare for a Christmas dinner or recipe combinations for a dinner party. There are many websites specializing in this type of information, frequently recipe sites that include the recipes for the items listed in the menu. In other words, looking for menu ideas is a way people search for complementary recipes. The Hispaniola website, by being better search-optimized than most menu and recipe sites and by containing multiple menus, was outranking them for searches such as "Spanish menu" or "menu for Christmas dinner." Within a few months more than 80 percent of the traffic coming from the search engines was what we came to call "menu" traffic.
However, the aim of the Hispaniola website is to gain bookings, so a conversion for http://www.hispaniola.co.uk/ is either a booking or an inquiry. Needless to say, people who came looking for a menu they could cook themselves weren't overly interested in booking a table; as soon as they arrived, it was obvious to them this was not the sort of site they were looking for, and they would leave. The bounce rate for menu traffic was around 85 percent, and the conversion rate was similarly low. In fact, I was surprised to see any conversions at all.
This wasn't a major problem -- it was easy to see what was going on, and easy enough to remove menu traffic from the web analytics data. We could continue to improve the site's conversion ratio (converting visitors into bookings), and the search listings that were used by our target market continued to improve. We simply accepted that there was a little "fog" around the stats that had to be removed each month in order to get a clear picture of what our target audience was doing.
Another key point in this case study is that the Hispaniola website is edited with Adobe Contribute. This is a simple, but excellent, content editing system for websites that don't need a database-driven CMS. But there is a minor downside to Contribute -- over time the code it generates can become messy. As formatting is changed, then changed again, Contribute can accumulate a change "history." For example, you can find this HTML code in Contribute pages: <B><B><B></B></B></B>. This translates as "turn bold on, turn bold on, turn bold on, turn bold off, turn bold off, turn bold off"; in other words "do nothing." This occurs because when you remove formatting, Contribute sometimes decides that, instead of removing the offending formatting command, it will add another command to counteract it.
In contrast, search engines like clean code. After a decade of being edited with Contribute, the amount of "rubbish" code had reached the stage where we needed to remove it in order to maintain the site's domination of relevant search listings. This is a fairly straightforward task, and was accomplished with a minimum of fuss.
Within a few weeks the conversion rate had unexpectedly tripled. Search listings for targeted phrases did not change, nor did the total amount of search engine traffic. What did happen was that the menu traffic vanished almost overnight. Search engines had delisted the site for "menu" phrases. However, instead of a corresponding drop in search engine traffic, it continued at the same level. Obviously something had happened to the mix of visitors the search engines were providing.