IP-popularity is an indicator used to evaluate links from different servers quantitatively. Links have a higher value if they are from different servers or different C-class networks.
IP-popularity is usually talked about along with domain and link popularity. While the former is based on the number of links from different domains, the latter is a parameter for the number of incoming links of the same domain. Therefore, we have to distinguish between the number of links from a domain (link popularity), different domains (domain popularity), and several different servers from different C-class networks (IP popularity).
The background to this is link building in off-page SEO. When a backlink profile of a web page is generated, it often happens that many links originate from the same host. The servers of this host often have the same C-class masks, also called subnet masks. C-class networks are IP addresses that have the same network address range.
IP addresses always consist of two parts, whereby the first digits of this address designate the network and the last ones, the computer which is connected to the network. When data is sent via the Internet, it is first passed to the subnetwork and then routed within the network to the computer.
To gain IP popularity, the backlink profile should not include the same subnet as much as possible, in other words different server addresses from different hosts. Only then will this parameter be high and the ranking of the corresponding webpage be better than is the case with links from the same servers with the same host addresses. To determine IP-popularity, backlinks have to be analyzed. There are both paid and free tools that do the job. Backlink checkers check the number of inbound links and their origin.
Google probably uses IP popularity as a ranking factor to exclude that many links from one link network direct to a particular web page. Google tries thereby to prevent strengthening backlink profiles through proprietary networks (see Black Hat SEO).
If you set links from just one network to your website, Google suspects unnatural link building and counts these links to your site as just one link. The page may even get penalized when this type of link building is obvious and is undertaken on a large scale.
Natural link building, however, is characterized by different network addresses to match the natural recommendations from people. The network addresses don’t have to be geographically far apart, but should have different C-class masks. The underlying assumption is that sites in the US, for example, which refer to each other by means of links have a higher popularity, than a reference from Thailand to a US website, which perhaps came about through a link exchange and is not a sensible recommendation, because the website is of interest for specific users.