Methodology

The Coursera Skills Graph connects skills, content, careers, and learners together through machine learning algorithms.

We use the text data of course materials to identify what skills each course teaches and what skills each assessment assesses. Then, we benchmark the skills proficiency of learners using their assessment performance data and skills tagging.

Containing over 38,000 skills in business, technology, and data science, the Coursera Skills Taxonomy combines data from open-source taxonomies, like Wikipedia, and subject matter experts, such as Coursera educators and learners.

The top layer of the skill classification structure is a domain, followed by a competency, then skill, and subskill.

In the Global Skills Report, we measure proficiency at the competency level.

Once we’ve identified what skills each assessment is testing, we use learner assessment data to train machine learning models that estimate how proficient a learner is in a skill and how challenging an assessment is.

Then, once skill scores are computed at the learner level and linkages are formed between users and countries, we aggregate the learner level data to measure country skills proficiency.

This is done by taking a weighted average of individual skill scores—ensuring that scores we’re more confident in count for more in their country’s skill proficiency score than other scores we’re less confident in.

Thanks to the flexibility of online learning, many learners making career changes tend to develop multiple skills at a time. On Coursera, these entry-level pathways to opportunity are defined by job role and associated skills needed.

By combining job posting and course enrollment data, we identify skills related to specific job roles. Then, we analyze learner behavior and skill-to-course linkages in the Skills Graph to understand how long a learner spends on assignments.

This informs how long it may take a learner to reskill for a certain role.