The audience estimation framework allows the publisher to annotate a set of URLs that a presumed segment of users are more likely to visit, compared to the overall population of users. Labeling subsets of a site according to the expected audiences allows us to estimate the presumed audience segment for a user, based on the page consumption on the site. To exemplify, we can imagine the site entertainment4all.com. This site contains articles on every imaginable aspect of entertainment, including feature articles on a variety of musical genres with a wide specter of artists. To better understand the users and their preferences, the audience annotation framework can be used to divide the users into well-defined audience segments. This is achieved by labeling subsets of the site with specific audience tags.
This may imply that entertainment4all.com decides to label all their Justin Bieber articles with the following audience tags: "gender: female", "age: 15 or below", "music genre: pop". Further, they may choose to label their opera section, art section, etc, differently, hence introducing a range of audience segments. When users consume articles on entertainment4all.com, we can then estimate the most probable audience segment for the user, based on the pages they consume. This will enable the publisher to target the user with specific recommendations and ads based on the audience segments for the individual users.
Injection of audience data
An estimation of the presumed audience for a page may be injected as metatags on a URL. The method for inject meta data to a URL is further described here. Essentially, if there is an expectation that a given URL is consumed by a specific audience, this can be reflected though the use of metatags. Considering that you have a wide variety of content targeting consumers of various age groups. For each age group we can define a metatag associated with this group:
Pattern:
<meta name="cXenseParse:<prefix>-audience" content="<value>"/>
Examples:
<meta name="cXenseParse:xyz-audience" content="age/15-25"/>
<meta name="cXenseParse:xyz-audience" content="age/25-40"/>
<meta name="cXenseParse:xyz-audience" content="age/40-60"/>
<meta name="cXenseParse:xyz-audience" content="age/60+"/>
For content where we have an expectation that a majority of the consumers may fall into a specific age group, this can then be reflected by inserting a metatag matching this age group.
Syntax
A well-formed audience keyword is injected with the group name <prefix>-audience, and will appear in the content profile of the page as a regular content keyword.
The <prefix> is defined as the Customer Prefix.
The keyword <value>, injected as content above, is expected to capture an audience type, and the specific value for the type. The type and value is separated with a slash, as in age/15-25. Note that there are no external restriction to the name and naming conventions of the type and values when these audience keywords are injected, as long as they abide to this syntax. Multiple audience tags may be inserted on the same URL.
Individual audience estimation
When a user visits a URL with a specific audience tag, this audience tag is associated with the user. Periodically, the page visits of a user, and hence the audience tags associated with the various URLs, are analysed. The user will be associated with the most probable audience tag, based on the analyses of the page visits for this user. This will be injected in the user profile for the user, and appear as a keyword in the user profile.
|
Group |
Item |
Weight |
|---|---|---|
|
|
The most probable value for a give type. E.g. |
The probability for the user being a member of the given audience segment type. E.g. |
The audience keywords for an individual user can be inspected with a profile/user query.
Aggregated audience statistics
Aggregated statistics an be retrieved with a /traffic/user/keyword query. This could e.g. provide statistics for a population of users, e.g. the proportion of users which are assumed to belong to the various audiences that are defined.