



Web traffic: forecasts for the end of 2008 

The first and third quarters were more dynamic in 2008 than in 2007, in terms of the way web traffic evolved, based on an equivalent number of websites. What will the full year results for 2008 look like? Here, XiTi Monitor brings you its forecasts… 
Perimeter: 
 Study conducted from January 2001 to September 2008
 Crosssection of 257,170 websites audited by XiTi


September 2008 was an excellent month, after slow growth during the second quarter of 2008 
The table below presents the way web traffic evolved, for an equivalent number of websites, for each first quarter (JanuaryMarch), second quarter (AprilJune), third quarter (JulySeptember) and fourth quarter (OctoberDecember) during the years 2002 to 2008.
The way web traffic evolved, for an equivalent number of websites, during the third quarter 2008 (+7.6%) showed twice the growth of the third quarter 2007 (+3.1%). The same trend was for that matter observed during the first quarter 2008.
The second quarter 2008 was on the other hand, less buoyant with a rate of growth that was five times less than that of the second quarter 2007: +2.4% vs. +12.3%. It is interesting to note that the second quarter 2007 was the only quarter during the year when growth was faster than in 2006. Was this thanks to the effects of the French Presidential elections in 2007 which triggered a lot of enthusiasm for using the web?
As for the fourth quarter 2008 it should, based on our forecasts, record a web traffic growth, for an equivalent number of websites, similar to that of 2007: +6.3%.

The rise in the web visit's rate, for an equivalent number of websites, measured during the third quarter 2008 compared with the third quarter 2007 can be seen during August and even more so during September, two months that were down on 2007:
 Change in July 2008: +6.2% (vs. +10.6% in July 2007),
 Change in August 2008: +4.7% (vs. 0.5% in August 2007),
 Change in September 2008: +11.9% (vs. 0.4% in September 2007, a month that saw a very marked seasonal drop). This is a figure that is very close to our forecast six months ago when we predicted growth of 12.7% for the month of September 2008 (refer to our March 2008 study: " Website traffic: September 2008 is looking good… ").


According to our forecasts and for an equivalent number of websites, the annual change in web traffic will stabilize between 2007 and 2008 
The graph below shows the annual change in web traffic, for an equivalent number of websites, between 2002 and 2008.
After rising for a number of years, hitting +25.2% in 2004 and 2005, the annual rate of change fell back sharply in 2006, to +10.1%, then again to +6.7% in 2007.
This slowdown in annual growth observed for the past two years appears to have stopped this year, as based on our forecast model, the rate of change in web traffic, for an equivalent number of websites, was +6.4%.
Although the annual rate of change has fallen since 2006, overall web traffic, for an equivalent number of websites, is up sharply over seven years: hence, for every 100 visits recorded in 2001, 253 visits will be recorded in 2008.
Just like the first quarter of 2008, the third quarter too saw a rise in web traffic (for an equivalent number of websites) in excess of that of 2007.
According to our forecasts, the overall trend in web traffic for 2008 should be similar to that of 2007: the slowdown in growth observed in 2006 and 2007 is therefore tending to stabilize. Of course, don’t forget that despite this slowdown, web traffic is still up after having grown by +153% over seven years for an identical number of websites!
Be sure to check out XiTi Monitor during the coming months to see how web traffic evolves.

Methodology:
The indicator used for this study, measuring the number of visits recorded based on a constant perimeter of sites, reveals the evolution of website activity.
In the absence of official data regarding the number and composition of existing websites, it is not actually possible to present an indicator that reflects the evolution of Webusers ’total Web consumption.
In this study, we looked at website traffic. This covers the evolution in audiences generated by an equivalent number of websites. The series of data successively integrates daily rates of evolution, calculated on a likeforlike perimeter on all sites audited by XiTi. Thus, the evolution of this indicator does not reflect the evolution of the XiTi perimeter, but website arrivals and departures within this perimeter are integrated in the calculation of our indicator.
The audience registered on sites within the XiTi perimeter very quickly revealed the existence of seasonal effects that processing methods make it possible to extract.
Tools adapted* to processing chronological series enabled us to break down this monthly series. The components are:  The trend of the series which represents longterm evolution of the series.  The seasonal component representing yearly or, in our case, monthly fluctuations that are repeated more or less regularly from year to year. It reveals phases of growth and recession.
A moving average provides an adjustment of the monthly audience; it corresponds to an estimation of the global trend. The moving average necessitates, at time t, measurements of the series at time t, as well as measures around t.
In our study, a moving average of 13 terms, symmetric centered on t, is adapted in order to measure the annual trend of a series of monthly data. To measure the moving average at month m, measurements from month m, month m1 to m6, and month m+1 to m+6 are required. This explains the necessity to estimate this moving average at the end of the period. The average seasonal effects estimated during the previous years and an estimate of the annual trend projected onto the last quarter of 2008 are used to estimate traffic for the period.
After estimating the long term trend and extracting seasonal effects, two successive steps are used to reach the predictions made up until December 2008:  The projection, during the prediction period, of an annual trend based on successive linear regressions, assuming a quasistability in the evolution of the annual trend over the short term.  Reinjecting monthly seasonal effects, so as to estimate the series during the prediction period.
*Nonparametric deseasonalization method.Methodology







