|
|
|
|
|
|
| Website Frequentation: prediction until January 2008 |
|
| The activity slowdown of Websites observed since 2006 is still an issue in the third quarter of 2007, with a month of September even less performing than in 2006. XiTi Monitor delivers to you here the latest frequentation data of Websites as well as its predictions until January 2008… |
| Perimeter: |
- Study conducted from to September 2000 to October 2007
- Perimeter of 222,318 websites
|
|
| The growth slowdown of Website visits continues into the 3rd quarter of 2007, with a marked dip in September. |
The table below presents the evolution indicator of Website frequentation for each third quarter (July – September) of the years 2000 to 2007.
On a comparable basis of websites, one thus observes, for 100 visits recorded in the 3rd quarter of 2000, 262 visits in the 3rd quarter of 2007: the visits multiplied by more than 2.6 in 7 years.
But this increase over the long term marks time since 2006. In fact, after having augmented by more than 20% over July – September 2004 and 2005, Website visits “only” increased 8% in the third quarter of 2006, and 3.1% in the third quarter of 2007:
|
This trend is even more marked for the website frequentation indicator over the month of September 2007, which did not increase compared with September 2006:
- Evolution between September 2004 and September 2005: +21.4%
- Evolution between September 2005 and September 2006: +7.6%
- Evolution between September 2006 and September 2007: -0.4%
|
The atypical nature of the seasonal dip of September 2007, clearly more marked than the previous years and thus non-foreseeable, explains the prediction of +7.7% of the website frequentation indicator for September 2007 (based on the average seasonal effects of the 6 previous years) in our study of April 2007 « Websites: frequentation prediction until September 2007… ».
The website frequentation analysis in fact allowed us to observe the existence of audience seasonality for Internet sites, with key moments over the course of the year like the summer dip or the January revival. Thanks to non-parametric methods of deseasonalization, we have been able to extract these monthly, seasonal effects.
The graph below presents the seasonal effects of the months of September 2002 to 2007.
Thus, the seasonal effect of September 2007 is more marked than the average of the seasonal effects of the months of September 2002 to 2006. It corresponds in fact to a dip of -11.9% whereas the average of the seasonal effects of the months of September from 2002 to 2006 is -4.8%.
|
| According to our predictions, a website will record 2.5 times more visits in January 2008 than in January 2001 |
The graph below presents the evolution of the Website frequentation indicator for the visits recorded from January 2001 to October 2007, and our predictions on this indicator from November 2007 to January 2008.
According to our model of prediction, for 100 visits recorded in January 2001, 256 visits will be recorded in January 2008.
|
|
The evolution of the frequentation indicator of January 2008 will then be, according to our predictions, superior to that of January 2007 (but below the three previous years):
- Evolution between January 2006 and January 2007: +5%
- Evolution between January 2007 and January 2008: +8.2% (prediction)
The growth slowdown of Website frequentation is confirmed in this third quarter of 2007 and if the evolution predicted for January 2008 is superior to that of January 2007, it is still far from the years of 2004 to 2006.
This slowdown can be explained by a growth even more significant of the number of pages available on the Web than number of Internet users. The latter, faced with a broader offer, would then be dispersed over a greater number of websites. Hence the priority, for the websites, to orient their actions (in addition to acquisition) towards making their visitors loyal that will be one of the major issues of the “inter-sites war.”
Rendezvous in the coming months on XiTi Monitor in order to follow the evolution of the website frequentation.
|
Methodology:
The indicator used for this study, measuring the number of visits recorded by a comparable basis of websites, reveals the evolution of the activity of Internet sites.
In the absence of official data on the number and the composition of existing Internet sites, it is in fact not possible to present an indicator reflecting the evolution of the total Web consumption of the Internet users.
In this study, the series which we were interested in is the frequentation of Internet sites. It regards the evolution of audience generated by a fixed number of websites. The series of data successively integrates the daily evolution rate, calculated on a comparable over the totality of the websites audited by XiTi. Thus, the evolution of this indicator does not reflect the evolution of the XiTi perimeter, but the arrivals and departures of websites in this latter are integrated into the calculation of our indicator.
The audience recorded on the websites of the XiTi perimeter made it possible to detect very early the existence of seasonal effects, that treatment methods make it possible to extract.
Tools adapted* to the treatment of chronological series allowed us to proceed with the decomposition of this monthly series. The components are:
- The series trend that represents the long-term evolution of the series.
-The seasonal composition representing the fluctuations that are infra annual, monthly, and in our case, those that repeat themselves more-or-less regularly from year to year. It highlights the phases of growth and recession.
A moving average allows a smoothing of the monthly audience; it corresponds to an estimation of the global trend. The moving trend necessitates, in the time t, the measurements of the series in the time t, but also measurements around t.
In our study, a moving average on 13 terms, symmetric centered on t, is adapted in order to measure the annual trend of a series of monthly data. For the moving measurement at one month m, the measurements of the month m, the months m-1 to m-6, and the months m+1 to m=6 are necessary. It is this that explains the obligation of proceeding with an estimation of this moving estimation at the period end. Thus, when November and December 2007 will be over, the moving average calculated replacing its estimate of September 2007, can bring about a slight correction of the seasonal effect estimated for the moment.
Finally, after estimation of the long-term trend and extraction of the seasonal effects, two successive steps make it possible to end up with predictions made until January 2008:
-The estimation, over the prediction period, of the annual trend by successive linear regressions, under the hypothesis of quasi-stability of the evolution of the annual trend for the short-term.
-The re-injection of the monthly seasonal effects, for the estimation of the series over the prediction period.
*Non-parametric method of deseasonalization.
|
|
|
|
|
|
|
|