Difference between revisions of "Unique Device"

Revision as of 13:41, 2 August 2012

Unique Browser (formerly Unique Visitor - one of the IFABC metrics) is an unique and valid identifier (= IP + User-Agent). [1]

Unique Browser not Visitor

This metric does not measure a person but is rather a measure of the device through which a person interacts with a website or network. Formerly it was called the Unique Visitor, but it was misleading as it made people believing that they measure by this way real people. [2]

Difficulties to count people

We may therefore reasonably assume that number of Unique Browsers somehow corresponds to people visiting a website (the trends corresponds well to the increasing or decreasing of the popularity of a website).

However, the very same person accessing a website from the office during the day, from the smart-phone when commuting, and from home the evening will be counted 3 times - because using 3 different devices. On the contrary, when sharing a PC (at home, internet cafe, etc), the visits of several persons are counted as only 1 Unique Browsers (as they access over the very same device).

Understanding aggregation of the Unique Browser

A typical error that is done while treating with Unique Browsers/Visitors data is that of taking the figures for a specific period and then of making a simple sum of single periods for counting a different timespan (for example, summing up unique browsers figures for 12 months to obatin a year period's unique visitors). Unfortunately, such sum does not have any real meaning as it does not correspond to the number of Unique Browsers for the needed period (in the previous example, for a year).

example with the unique and repeated clients in a shop

Imagine to own a small grocery and you want to know how much clients do you have:

• Monday come Jane Dee (morning) = 1 unique client
• Tuesday comes Jane Dee (afternoon) = 1 unique client
• Wednesday comes Jane Dee (morning) and she returns still the afternoon = 1 unique client (and 1 repeated client)
• Thursday comes Jane Dee (noon) = 1 unique client
• Friday comes Jane Dee (afternoon) = 1 unique client
• Saturday comes Jane Dee (morning) and John Do (afternoon) = 2 unique clients

(Sunday you are closed)

When measuring number of clients per day, your measurement starts at 00:00 and ends 23:59. On Monday you had 1 unique client. Then the next day you set counters to zero; and Tuesday you had 1 client again and so on and so far.

When measuring number of clients per week, you start measure on Monday 00:00 and you stop on Sunday 23:59. Then you had Monday 1 client and Tuesday the very same client came you again (repeated client). And it was always the same loyal client coming every day. Only Saturday came 1 different client. So per week you had 2 unique clients and 1 repeated clients.

In other words you have during the week 2 unique clients, Jane Dee coming every day (repeated client per week) and John Do coming once a week (unique client per week). (You might try to make a simple sum of number of clients per day (2+1+1+1+1+2=8) and believe that you have 7 clients, but it is just because you would count Jane Dee 6x and John Doo 1x.

If your clients keep the same behaviour, so Jane coming every day and John coming once a week, the number of your unique clients per month will be also 2 (and also number of the repeated client will be 2 - John came every Saturday in the month), as well as the number of the clients per year (Jane came every day in the year and John every Saturday of the year).

Example with the Europa webnest

Following examples show the number of unique and repeat browsers aggregated per day.

To get the number of unique and repeat browsers aggregated per week, there has to been done new aggregation.

The sum of aggregations per day gives wrong number, because this way are counted again the unique browsers which have been counted already.

All what has been mentioned above, can help to understand the number of unique and repeated browsers on EUROPA webnest by month and by year. (The conclusion that in the year 2010 there have been 157 593 711 unique browsers = the sum of the monthly aggregation is wrong, because some unique browsers could be count even 12times).