Torre’s Job-Matching Model | Connection strength

4.1. Connection strength

4.1.1.

Purpose

Talent seekers tend to hire people from their network before looking at any other options for hiring. This network is often accessed through WoM or LinkedIn, but the network is a lot larger than that outreach usually goes. With this factor, we rank all candidates based on their multi-degree connectivity to members of the job post.

4.1.2.

Data

We work with a connectivity graph in our backend where every connection between people and/or organizations is stored. The graph is populated with a lot of information:

Google contact information
Address book information
Torre signals
Torre applications
Shared job postings
Messages sent
Referrals
Shared experiences
Shared organizations

4.1.3.

Data validation

Since we have a lot of different types of connections between people and organizations, we can use connectivity weights to determine how likely the people are to actually know each other.

4.1.4.

Algorithm validation

Daily and weekly metrics
Editorial reviews
Direct feedback from talent seekers
Direct feedback from candidates

4.1.5.

Description

4.1.5.1.

Iterate over each person in the network:

Start by going through each individual in the network.

4.1.5.2.

Acquire the individual's PageRank:

Retrieve the PageRank for each person, which is a measure of their importance or influence within the network.

4.1.5.3.

Gather and categorize connections:

Collect all the outgoing first-degree connections for the person, categorize them by type (like friends, colleagues), and store this data.

4.1.5.4.

Calculate the connection quota:

Determine each person's connection quota by dividing their PageRank by the total number of connections in the network. This quota represents the amount of the person's influence that can be distributed among their connections.

4.1.5.5.

Process each type of connection:

For each type of connection (such as friends, family, professional contacts):

Identify the connections of that specific type from the stored data.
Calculate the weight for these connections by dividing the person's connection quota by the number of connections of that type. This step distributes the person's influence across their connections.
Update each connection with the calculated weight, quantifying the influence or importance of that connection in the person's network.

Then for every visible member of the opportunity, we look at all their outgoing connections and see what’s the connection with the candidate of the highest weight. That will determine the score for that candidate.