Long-Tail Queries

Introduction

The searches that bring up your web site are such that the distributions of queries is made up of a few common queries (hopefully your chosen keywords)  and many more so called long-tail searches.  Long-tail search queries are very specific queries that may only occur once however, the total number of these individual queries is such that they form the bulk of the total search. You will often hear about the 80/20 rule, i.e. 20% of the total number of search queries will be made up of the most common key phrases with the remaining 80% given to the long-tail.

Modelling the Long Tail

Long-tail searches can be fitted using a power law (also known as Zipf’s law) of the form y=c x^-k. Another distribution called the  lognormal distribution can also fit the results, although care must be taken in interpreting the results as these two curve agree while the lognormal will fall off more rapidly than the power law.  To see the results clearly, we use a double logarithmic scale on which a power law distribution is a straight line.

Using Google Analytics data for three websites, I took the keyword data and the number of visits from three different web sites and plotted a log/log graph to illustrate the power law dependence of the data. Site I takes keyphrases and page views over a period of a year. In this time there were more than 52,000 keywords. Analytics only allows you to see 500 at a time so later results are plotted at important points to illustrate the relationship.

For another well optimised website which has less traffic but shows a reasonably good fit with the data. More than 5200 key phrases making the long tail in a one month period.

Over one-month the graph has less data it still has the power law distribution but the fit is not so good. This may be due to the time scale over which the data is taken. The steps toward the right-hand end of indicate that there are many queries that each generate a single pageviews.

For a very low traffic web site over a year. The results are as follows:

Long tail search queries are very specific and varied by their very nature. All three of these sites have lots of text content and so do well in picking up long-tail search terms. Any large amount of text content will naturally be suitable for many such search queries so to catch the long-tail the most effective way is to spread your net wide by including lots of text content and include variations of key phrases and synonyms where possible.

The long tail is also illustrative of the competitiveness of the key phrases for a well optimised website,  since there is a direct relationship between the competition for a key phrase and the traffic it receives. The long-tail can be very useful for bringing highly selective traffic in a competitive market.