New data analytics study reveals how to boost educational outcomes


Through advanced analytics and machine learning, a group of experts studied the results of the PISA 2015 test and found which variables were most strongly associated with good test performance.

By Esteban Fredin

In another example of how data science is changing the face of education research, McKinsey & Company, the management consulting firm, has released this year a series of reports on education around the world. Through advanced analytics and machine learning, a group of experts studied the results of the PISA 2015 test and found which variables were most strongly associated with good test performance.

The database of the Program for International Student Assessment (PISA), an initiative from the Organization for Economic Co-operation and Development (OECD), is the largest and most extensive worldwide. The test designed by this organization quantifies the performance of a representative sample of 540,000 15 year olds in math, reading and science in 72 countries. In addition, students, parents, teachers and authorities complete surveys that provide information about family background, socioeconomic status, attitudes and behaviors of students, resources, school leadership, teaching practices, teacher training and professional development; summing a total of 1440 variables, making it the most important quantitative resource to understand the state of education around the globe.

The report provides two very interesting findings that invite us to rethink some widespread assumptions about teaching and education today:

A proper student mindset is even more important for performance than socioeconomic status.

In general, mentality and student motivation are better predictors (30%) of good test results than home environment and other demographics (16%). Also, not all motivation types are the same. What researchers called 'motivation calibration'
—i.e. wanting to hand in a quality homework— turns to be doubly more important than motivation identified by the student —wanting to be a honor roll student—. This study estimates that a well calibrated motivation is equivalent in impact to jumping to the next socioeconomic level.

Furthermore, echoing Carol Dweck’s work about the importance of cultivating a 'growth mindset' to facilitate learning,1 this study found that students with this attitude —those who believe they can improve through effort— had a 12 percent better performance on the test than those with a 'fixed mindset'; i.e. those who believe that their intellectual abilities cannot change.

Inquiry-based learning is no replacement for teacher-directed instruction.

Students do benefit from facing challenging subjects led by a teacher with appropriate pedagogical training. Used alone, teacher-directed learning showed a greater positive impact on test results —12 points above mean scores— than inquiry-based learning  —61 points below average—. Does this refutes years of research and use of inquiry-based learning? Not at all, the best effect, 26 points above the average, happens when the two methodologies are used together, so that what is mainly a teacher-led education is complemented by some inquiry-based learning.

To explain this, researchers put forward two hypotheses: 1) Inquiry-based learning is more demanding for students, so they can only benefit if they already posses a solid knowledge and skills foundation. 2) Inquiry-based learning is a complex process; teachers who have not been properly trained in its methodology can experience serious difficulties.

While acknowledging that the PISA test, like any other standardized test, is only an imperfect measure of teaching effectiveness, this new study by McKinsey & Company shows us the power of data analytics to explore and interpret vast amounts of data and thus extract relevant and actionable conclusions that teachers and institutions can carry into their practice.


1 Dweck, Carol S. Mindset: The New Psychology of Success: Ballantine Books, 2008.