Torre’s Job-Matching Model | Recommendation weight

4.5. Recommendation weight

4.5.1.

Purpose

The skill recommendation-weight ranker algorithm's purpose is to evaluate the reliability of the genome and how much others have liked working with the genome.

4.5.2.

Data

For this factor, we take into account all recommendations on the platform. We compare all recommendations made to the person with other people’s recommendations.

4.5.3.

Data validation

We validate recommendations through experiences. While giving recommendations, users have to tell with which experiences they have worked with the person they are recommending.

4.5.4.

Algorithm validation

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

4.5.5.

Description

The weight of each recommendation is determined by:

The longevity of the affiliation between the recommender and the recommended. Longer relationships will yield greater weight.
The weight of recommendations received by the recommender. So the more recommendations you have, the greater the weight of the recommendations you give.
The type of relationship between the recommender and the recommended.
The number of recommendations given by the recommender. With each recommendation you give, the overall weight of your recommendations diminishes proportionally. This means that if you recommend selectively, your recommendations will have greater weight than those given by someone who doesn't recommend selectively.

4.5.6.

Mathematical description

The mathematical formula for this process is as follows: t is the recommendation weight of the recommender. When the recommender is verified and has a recommendation weight below 100 (for example, new users), becomes 100. When the recommender hasn’t been verified yet, becomes 0. Lr is the longevity of the affiliation between the recommender and the recommended. Fr is the relationship type factor. For most relationships, the factor is 1. For recommendations where the recommender is the subordinate (for example, if you were led, hired, taught, mentored by - or received investment from - the person you’re recommending) the factor is the golden ratio (φ = 1.6180339887…). d is the damping factor. A new recommendation can alter the weight of many other recommendations both directly and indirectly. Because of that, recommendation weights must be calculated cyclically. To make recommendation weights easier and faster to compute, the dampening factor slightly reduces the weight of a recommendation. The damping factor is 0.9.