Internet advertising metrics - difference between cookies, unique browsers and people - Source: Nielsen Online
Common practice is that traditional principles of media advertising are applied on the Web, however, new media requires a different approach and analysis. Therefore in this article we will explain some Internet specific metrics such as cookies, unique browsers and why is challenging to use try to transform cookies into people and to use this for online campaign measurement and planning.
Cookies are a very common metric for advertising and are accepted as standard globally. Not because the Audience Measurement company tracks cookies (in fact Audience measurement can also track people, using a Panel) but because the advertising platforms use cookies. Majority of the internet advertising is delivered to internet users by an Ad-Server (a server dedicated to serve advertising). This ad-server uses cookies to recognize users (in order to avoid, for example, that the same user is exposed to the same ad 100 times). Whenever a campaign is planned, it is planned taking into account a frequency metric. Based on that, the frequency is calculated by the ad-servers using cookies, not counting the real number of people.
Cookie facts:
-
Cookies can be accepted or not accepted by a user.
-
Once accepted, cookies can be deleted or not by a user.
-
There is software that helps users manage their cookie acceptance and deletion.
-
A cookie is linked to the browser that accepted it.
-
If a person uses different browsers (for example Internet Explorer and FireFox) this person will generate multiple cookies (in this case, two).
It’s very easy to understand why the number of total “unique cookies” is way beyond the total market population: because Unique Browsers (metric based on cookie count) were never intended to be people. This is why in markets where Nielsen provides the common industry currency based on tags & cookies, the total number of Unique Browsers per month is higher than the total internet population or even, in some cases the total population of the country.
Now think about how you use your computer on a daily basis: do you delete cookies? Maybe you do. How often do you delete cookies? Some would say once a month, some once a week, a few will say every day. But let’s pretend you NEVER delete cookies. How many different browsers do you use in a month? Maybe you are a loyal IE user, but sometimes you use Firefox, or you check your webmail using your mobile: you have “generated” 3 different cookies, even without deleting them. So also cookie generation is something that needs to be considered in the Audience Measurement scenario, and it is even more important than the simple cookie deletion factor.
Unfortunately there is no simple way to take into account the cookie deletion and cookie generation factors with a pure tagging only methodology. Even crossing data with a panel is a hard task, considering that cookie deletion and generation vary a lot over time and there is not a “fixed adjustment rate” that can be applied. To achieve this it requires a very specific type of high quality panel, both in terms of sample recruitment and management, and in terms of the technique to track the individuals under measurement. Seasonality, Audience composition and Site type effect a lot the variation of deletion and duplication phenomenon.
Several approaches have been explored with cookies only, or with lower quality panels, however the standard error is too high to be considered solid and reliable, and or on the other site the user privacy can be put at risk because too intrusive methods are chosen. But let’s focus on one small fact: the chances that a user generates or deletes a cookie in a single day are lower than the chances that a user generates or deletes a cookie in a week or a month interval.
Let’s do a very simplified example; User “Mario” is loyal to the site, he surfs on that site every single day (we all wish our users are like Mario, don’t we?). He is also a privacy concerned user, and he deletes his cookies once per day. So if we try to draw a small graph with the real number of people surfing for 5 days and the number of different cookies surfing for the same interval (cumulated data) the result would be similar to this:
The Red line is the real number of people (1) and the blue line is the number of unique cookies (Unique Browser) surfing on the site. So if we abstract from this example and generalize the normal website audiences over time, the graph that we obtain would be something similar to the following:
The way Average Daily Unique Browsers is calculated by Nielsen.
Why 28 rolling days? Because we need to make sure we analyze the same amount of the same day-type. Average Daily Unique Browser metric is the closest metric to the real number of people based on pure tagging methodology.
Why not using information straight from a Panel to transform cookies into people?
Internal research has demonstrated multiple times that although it is possible to transform cookies into people using a Panel, it’s not a simple process, and certainly requires far more than just applying a conversion ratio. The huge variations on behavior of the users over time and across demographic dimensions make hard to create those simple ratios. And this is mainly not due cookie deletion, but due to the cookie generation process.
While it is possible to understand the cookies age, it’s impossible to know if that cookie is “new” because the user just deleted one or because it’s a real new user. Therefore assumptions have to be done based on the real number of people.
Moreover, with tagging methodology only, it’s impossible to know how many people are behind a browser and how many browsers are in front of a single person.
Cookies are good and transparent.
Cookies have many advantages in terms of audience measurement, they are easily understood, there is a massive knowledge about how to use them, and are a shared tool across multiple Internet technical solutions. They are also the most common metric for the ads, Audience measurement, and, with the Average Daily Unique Browser, it’s a transparent metric, providing figures close to the real number of people.