Forum Moderators: DixonJones
I've explained that this is just a maths problem because some of the unique users are the same between different months. Because you can't count them more than once, the total uniques for a quarter is NOT the sum of the individula months.
My client cannot get his head around this and just sees this big discrepancy.
I personally think that the average PV per unique user will get more and more inaccurate (and higher) the greater the period of study. So if you take a whole year, the average will look ridiculous compared to a monthly average.
So, my point is should we only calculate average PVs per unique user for a monthly period, and no longer ? Is this more accurate ?
So, my point is should we only calculate average PVs per unique user for a monthly period, and no longer ? Is this more accurate ?
I would say that average PVs are pretty much useless over any timescale.
One problem with the using the word "average" is that there are 3 different types of average - Mean, Mode and Median. They all have their uses in different circumstances.
Mean - This is the measurement that most people mean when they refer to an average. Simply add up all your page views and divide by the number of uniques.
Mode - The number of pages that most of your "uniques" viewed.
Median - The number of pages that the middle "unique" viewed
Take a simple example of a day on a quiet site:
User 1 - 10 page views
User 2 - 1 page view
User 3 - 1 page view
User 4 - 200 page views
User 5 - 8 page views
Mean = 220/5 = 44 views
Mode = 1 view
Median = 8 views
So which of these averages is most useful. The answer in this case is probably none of them. Why, because Users 2,3 & 4 are all spiders (guess which one is Alta :))
So firstly you need to strip spiders out of any stats involving uniques.
In the case above all three averages would then be 9 (the mode might not be, but bear with me it was years since I was at college).
Suppose we have stripped the spiders out of our stats on a busier site and have the following figures:
User 1 - 10 page views
User 2 - 8 page views
User 3 - 1 page view
User 4 - 2 page views
User 5 - 1 page views
User 6 - 25 page views
User 7 - 3 page views
User 8 - 4 page views
User 9 - 8 page views
User 10 - 2 page views
User 11 - 2 page views
Gives Mean = 6, Mode = 2, Median = 4
Are any of these "averages" useful ?
Possibly the Mode is, it tells you that most people view 2 pages and then leave. But if User 1 hadn't hit refresh twice and User 10 had viewed one more or one less page, then the mode would been 8 pages, which is quite different from 2.
The average published by most stats packages would be 6. Almost completely useless.
If your software is capable I would produce a graph charting page views against uniques as a percentage.
Group your ranges of pages views in to something relevant to your site.
In this case group as follows:
1-2 page views - 45% (of visitors don't like the site)
3-6 page views - 18% (of visitors quite like the site)
6-9 page views - 18% (of visitors really like the site)
9 + page views - 18% (of visitors love the site)
I would suggest that this information is better than just publishing the number 6. However these numbers are still open to interpretation. Perhaps those one and two pages visitors found exactly the information they were looking for quickly and the 9+ viewers were flapping around looking.
In summary if you cannot produce a histogram I wouldn't bother with publishing this average just publish the total uniques and the total page views.
Again, defining what a session if leads to many problems, but this method would perhaps satisfy your client more.
I'm just going to use this to keep the client happy. Personally, I agree with Mark, the whole concept of averages is useless in this context.
Thanks for your help guys :-)
Fom
