Abstract: Described herein are techniques for generating for each company in a set of companies a rating score that represents the desirability of employment with the company. The rating score for each company is derived by analyzing the member profiles of members of an online system to identify employee transitions between companies over certain time periods. At the beginning of a first time period, each company is assigned a baseline score. Each transition represents an employee departure (loss), for one company, and an employee hire (win), for the other company. With each transition by an employee to/from a company, the rating score of the impacted companies are adjusted—increased for a win, decreased for a loss—in some proportion that is relative to the current rating scores of the impacted companies. Accordingly, company A hires an employee from Company B, the rating scores of companies A and. B will impact the level of the adjustment that is made to their respective rating scores as a result of the transition.

Abstract: In an example embodiment, a machine-learning algorithm is used to train a talent peer model to output a score indicating a likelihood that one organization is a talent peer to another organization. Scores above a predetermined threshold indicate that the organization is a talent peer to the other organization.

Abstract: Methods, systems, and computer programs are presented for generating company-comparison reports based on a company ranking for hiring and retaining employees. One method includes an operation for determining transitions of users of a social network based on their profiles. Each transition comprises a change of employment from a source to a destination company. A transition graph is created, for a group of companies, including a node for each company and links between the nodes. Each link comprises a number of employees that transitioned from the source to the destination company. A weight, calculated for each link in the transition graph, is based on a number of employees of the destination company and a number of users transitioning between companies (both directions). A company score is calculated for each company based on the transition graph and the weights of the links. A report based on the company scores is then presented.

Abstract: In an example embodiment, a machine-learning algorithm is used to train a talent peer model to output a score indicating a likelihood that one organization is a talent peer to another organization. Scores above a predetermined threshold indicate that the organization is a talent peer to the other organization.