Auto-Matching is an intelligent process that works behind the scenes in the following matching processes:

• Admin-led
• User-led with 1-sided approval 
• Roulette

You can read about each match process in Matching Process Overview.

Admin-led

Admins can click Run Auto-Matching to optimize all remaining unmatched Users, taking into account how much capacity they have left. The results require Admin review and can be approved in bulk or one-by-one.

User-led with One-sided Approval

The same Auto-Matching algorithm is used, but the results are sent to Users as recommended matches instead of being sent to Admins. This is part of the Recommendations feature.

When matching is turned on, eligible unmatched Users receive an initial invitation to match. After that, Together continues to generate Recommendations for eligible unmatched Users on a recurring cycle.

Recommended matches are optimized across the full matching pool, so a User’s Recommendation may not always be their highest-scoring individual match. This helps maximize the overall quality of Recommendations across the Program.

How Recommendation Cycles Work

Recommendations are automatically generated and sent to unmatched Users in a program after the initial invitations to match are sent out.

Recommendations are sent every 2 weeks only if all of the following conditions are true:

• The matching process is currently ON for the Program.
• The User has previously received an invitation to match when matching was turned on.
• The User is currently unmatched in the specific match type. For example, a User who is already matched in a peer-to-peer match type may still receive Recommendations for a mentorship match type in the same Program.
• The User’s role is eligible to match. For example, if Mentees can request matches is turned off, mentees will not receive Recommendations.
• The User is eligible to match and has not been explicitly marked as ineligible by a Program Administrator.
• The User has not turned off match recommendations for that match type in their profile settings.

Recommendations are recalculated once per day using the current pool of Users in the Program. The Auto-Matching system optimizes Recommendations across the entire matching pool, meaning a User may receive a Recommendation that is not their highest individual match score if their top match was recommended to another User in order to maximize the overall quality of matches across the Program.

As a result, Users may occasionally receive no Recommendations during a Recommendation cycle if all of their eligible matches were recommended to other participants. This does not mean the User will never receive another Recommendation.

Every day, Users who currently have no Recommendations or whose previous Recommendations expired more than 1 day ago are added back into the Recommendation pool. Recommendations may change over time as:

• Other Users accept or decline Recommendations
• Existing Recommendations expire
• New Users join the Program

Users are never recommended the same match twice. Once all eligible matches have already been recommended, the User may stop receiving additional Recommendations.

Program Administrators can review the most recent Recommendations sent to Users in order to validate that Recommendations are working correctly. Administrators can also manually move recommended matches into Staged, Active, or Awaiting Admin Approval states. These actions will send email notifications to the affected Users.

Roulette

The same Auto-Matching algorithm is used, and results are announced to Users on a regular cadence as their new match. There’s one twist: past matches are excluded until you’ve matched with everyone else available.

Note: The Auto-Matching algorithm is not used in User-led with two-sided approval. In that process, Users create a shortlist of favorite matches using percentage match scores as a guide.

How Automatching Works

Auto-Matching analyzes all possible combinations of matches across all Users and then assigns the optimal set of final matches.

Example:

Mentee 1 with Mentor A: 8 Mentee 1 with Mentor B: 5 Mentee 2 with Mentor A: 9 Mentee 2 with Mentor B: 4

Starting with Mentee 1 and assigning Mentor A would be incorrect, because Mentee 2 would be stuck with a poor match (4). Auto-Matching looks at the whole picture to avoid this.

Now imagine doing this for 1,000+ Users  - it’s a complex optimization problem.

Auto-Matching also factors in capacity: how many mentees a mentor can take. If someone already has matches (or matches staged by an Admin), those count toward capacity before Auto-Matching runs.