Webtrends stores the processed data from your log files in multiple "buckets" that are broken down by a specific time period. These buckets fall on yearly, quarterly, monthly, weekly, and daily boundaries. When you request data for a report, Webtrends pulls this data from the largest available bucket to fill the report. If you try to sum the contents of some smaller buckets to the contents of a larger bucket, the visit counts will most likely not match. For example, if you were to try to compare a monthly bucket’s data to the sums of the four weekly buckets that make up the same time period the numbers will be different.
The reason for this is because of the way Webtrends calculates visits. A visit is defined as the time period for which a visitor is active on your site. A visit starts when a visitor first loads a page on your site, and terminates 30 minutes after that visitor stops requesting pages (i.e., 30 from the last entry in your logs made by that visitor).
When you look at the buckets, you are most likely going to have visits that span the cut-off boundary for the bucket. For example, a visit that is still active at 11:59 pm on Tuesday, and doesn’t stop until 2:15 am on Wednesday, spans a boundary cut-off for your daily buckets. In this case, that visit is counted in the totals for both Tuesday and Wednesday. This is accurate since when you look at any single day, you want to know how many visits were active on your site on that day. However, if you simply sum the visit count from the days for the whole week, you are counting some of your visits twice if they span a day boundary. Thus, we have a weekly bucket that has data for the whole week which does not count visits that cross a day boundary twice since we’re now looking on a weekly scale rather than a daily one. The weekly bucket will have an accurate count of how many visits there were for the whole week. It is because of these boundary issues that we created the various bucket sizes. Each bucket is accurate for the time period that it represents.
Webtrends gives you the ability to create custom ranges within your report. In order to get the most accurate count for that range, Webtrends pulls as much data for that time period as it can from the largest bucket that fits in that time range. After that it tries to fill in the holes with the smaller buckets.
If you request a report on Saturday, July 1, 2009 to Sunday, July 9, 2009, Webtrends will pull the numbers from the weekly bucket for the week of 7/2/09 - 7/8/09, and then add in the numbers from the remaining two daily buckets for 7/1 and 7/9. As discussed above, there may be a little overlap between the ends of the week and the two day boundaries, but this will be a small percentage of the overall numbers. Since Webtrends is meant to be trending software, the exact values are not as important as the overall trends, and since the overlap will be relatively constant over time (increasing along with the rate of increase of your overall traffic), its effect on the trends will be constant as well.