How Do GMAT Scoring Algorithms Affect You?

I have worked with hundreds of prospective business school applicants during their preparation for the Graduate Management Admission Test (GMAT). Recently, I used my extensive experience in this arena to help design the GMAT Exam Simulator software distributed by Upstart Raising to GMAT Preparation providers and tutors worldwide.

Since we began offering this educators' software direct to students, I have received many inquiries from MBA candidates interested in the finer points of how we successfully emulate GMAT scoring within our software. In the next few paragraphs, I will attempt to shed some light on this subject.

The actual calculations for raw, scaled sectional, and cumulative scores are not published by Educational Testing Service (ETS), the current publisher of the actual GMAT exam, and the goal of the GMAT is to produce scores that are consistent year to year. In other words, a 650 from several years ago is supposed to be equivalent to a 650 today. To meet that goal, the scaled sectional scores continually fluctuate with regard to the associated percentile.

For example, average quantitative scores have been rising for several years and it now takes a scaled score of 51 to reach the 99th percentile while a 48 now represents only the 87th percentile. By comparison, in the verbal section, there is a 7 point difference between the 99th (a score of around 45) and the 87th percentile (around 38).

Since these types of scoring adjustments are not announced publicly, the only reliable way to monitor these scoring fluctuations is to use observations of actual test scores. So clearly software algorithms must be continually monitored and adjusted to provide accurate scoring assessment.

At Upstart Raising we devote the large majority of our effort towards the reliable reproduction of the total cumulative score with regard to test taker performance. We continually adjust our formula for cumulative score generation along with question pool content to reflect the observations and feedback of actual test takers. In stark contrast, many inferior GMAT preparation software products use static formulas and question pools that bear no resemblance to the current exam.

It is important to note that there is no simple relationship between the number of questions answered correctly during a single GMAT session and the associated cumulative score. Each question in each section of the GMAT is assigned a unique weight and earlier questions in the section are worth more than later questions while quantitative questions are generally worth more than verbal ones. Students should be aware that the cumulative score on the GMAT is calculated in increments of 10 points.

So, the difference between a 690 and a 700 is only one "tick". This difference could be attributable to a difference in the number of correct responses, but that is not necessarily the case. 10 points is so narrow a margin in the cumulative GMAT score that the order of your incorrect responses could be the determining factor in whether you receive a 690 or 700.

For example, if you miss 3 verbal questions towards the end of the section during one GMAT session, but miss the same 3 questions in the middle during another session, your second session could easily be scored 10 points lower. This is one reason why we coach students to give more attention to the earlier questions on each section. Quite simply, the earlier questions are worth more.