Subscribe to the RSS feed (english)
The newspaper NetEco has made an effort to publish an article comparing "site-centric" and "user - centric " approaches. The article concludes on “an a priori objective study”, which uses data from the ISPs to US. Let’s call this method "network".
Unfortunately, none of these three methods is perfect. To complete the article by NetEco, here is a small comparative article on the concept of unique visitor.
To begin with, what is represented by “a unique visitor”? One tries to catch the person behind the computer. This may not be a bad idea, but I am not sure it is so meaningful.
I have already spoken about the difference between identity and personality (in french). From a marketing point of view, each of my personalities is important. And I am not sure that we can so easily merge them into a single identity. Let me explain. I am both a professional and an individual. I use the Internet for work and for leisure. I am not in the same state of mind, I do not have the same availability, or the same interests when I work or when I don’t.
The Laurent Nicolas at work who receives an ad for a discount on diapers is not the same as the one who receives the same publicity at home. Advertisers are trying to reach a holy grail which they perhaps don’t even need? I ask the question here, but I do not think they want to get an answer today ... So, let’s pretend that it is interesting to talk about unique visitor. The following table shows when each measurement method may overestimate (+) or (-) underestimate the number of real * unique visitors.
In conclusion, the network-centric method faces the same problem as user-centric (representativeness of the sample), but does not benefit from its precise measurement (mouse-keyboard activity for example, identification of members of a home ...).
The main reason that justifies why such an approach has been developed is certainly not methodology. We must look on the side of patents developed or purchased by Nielsen: it is now almost impossible to make the user-centric measurement and difficult to do site-centric systems without paying royalties. People must therefore look for other solutions.
To conclude on proposing a possible method, let me switch back to the notion of personality. Ideally, the user should have an interest in identifying (to be sure they do correctly). The UV defined on the basis of identified users would be closer to reality, whatever the place of connection, the number of users of the computer, and so on.
Does this seem interesting?
I have already spoken about the difference between identity and personality (in french). From a marketing point of view, each of my personalities is important. And I am not sure that we can so easily merge them into a single identity. Let me explain. I am both a professional and an individual. I use the Internet for work and for leisure. I am not in the same state of mind, I do not have the same availability, or the same interests when I work or when I don’t.
The Laurent Nicolas at work who receives an ad for a discount on diapers is not the same as the one who receives the same publicity at home. Advertisers are trying to reach a holy grail which they perhaps don’t even need? I ask the question here, but I do not think they want to get an answer today ... So, let’s pretend that it is interesting to talk about unique visitor. The following table shows when each measurement method may overestimate (+) or (-) underestimate the number of real * unique visitors.
| Site-centric | User-centric | Network-centric | |
|---|---|---|---|
| (+) | If it uses several navigators / computers, the UV is counted several times. * If the user deletes the cookies | * If visitors of a site are overrepresented in the sample, the projection exceeds the reality * If the sample is underrepresented on a profile (gender, age, profession) positively correlated with users of the service, the method of correction of representativeness (weighting) will exaggerate their numbers | As long as the analysis does not use all the ISPs (illusory), network-centric methods face the same problems of representation as the user-centric ones. For example, the audience of the ISP’s portal is overestimated. |
| (-) | * If multiple users share the same computer (in a home, an internet cafe, a school, etc.). * If IPs are used to estimate the number of VU | * If a portion of the audience comes from a place (foreign countries, offices, university, cafes), which is not represented (or underestimated) in the sample * If members of the sample do not identify correctly ( in the case of Nielsen, a window asks which member of the household is the current user) | * In the same case that the site-centric * In the same case as the user-centric |