Skip to main content

Custom scoring models

Jason Ball avatar
Written by Jason Ball
Updated over 2 weeks ago

Using Custom Scoring Models in RightMatch AI

RightMatch AI lets your team create personalized scoring models that learn from how you evaluate resumes and video interviews. By adding custom scores, the platform trains distinct AI models that reflect your unique hiring preferences.

This article explains how custom scoring works, how many scores you need, and what to expect once your models are trained.


What Are Custom Scoring Models?

RightMatch uses two separate custom scoring models:

  1. Resume Scoring Model – Learns how your team evaluates resumes

  2. Video Interview Scoring Model – Learns how your team evaluates interview responses

You can train either model independently or train both—it depends on your hiring workflow.


How Many Scores Do We Need?

To train a custom model, your workspace must add:

  • 25 uniquely scored resumes → trains the Resume Scoring Model

  • 25 uniquely scored interview questions → trains the Video Interview Model

These are independent.
If you want both models, you’ll need 25 of each.

What counts as a “unique” score?

  • A unique score = one resume or one interview question that has been scored at least once.

  • Multiple team members scoring the same resume or interview question still counts as one.


Adding Custom Scores

🏆 Resume Scoring

To add a custom score to a resume:

  1. Open the Enrichments table

  2. Click the candidate’s name

  3. Enter your score in the scoring panel

Each scored resume helps shape your resume-specific model.


🎥 Video Interview Scoring

To add custom scores to interview responses:

  1. Open a candidate’s video interview

  2. Click Review next to a question

  3. Add your score

  4. Repeat for additional questions

Each question you score is a training data point for the interview model.


How Training Happens

Once your workspace reaches the required number of unique scores:

  • RightMatch trains the model automatically

  • There’s nothing you need to configure

  • Updated scoring will begin appearing throughout the platform

Models continue improving over time as you add more scores.


FAQs

Do resume and interview scores count together toward the 25?

No. They are completely separate:

  • 25 resume scores → trains resume model

  • 25 interview question scores → trains interview model

You can train one without the other.


What if multiple users score the same resume or question?

It still counts as one unique data point.


Do we need to manually start training?

No. Training begins automatically once your workspace meets the threshold.


Can we retrain or reset our models?

Models retrain continuously as more data is added.
For a full reset, contact RightMatch support.


What happens if we never reach 25 scores?

The system continues using the default scoring model until the custom model can be trained.


Best Practices for Better Models

  • Encourage team members to regularly score resumes and interview answers

  • Score a representative mix of strong, average, and weak candidates

  • Continue scoring even after the first 25—more data = better performance

  • Add scores early in your pipeline to guide candidate ranking


If you need help building strong scoring data or want guidance on best practices, message us anytime through Intercom!

Did this answer your question?