2025 GMAC Greater China Test Prep Summit: Key Data Insights & Takeaways
Introduction: 2025 GMAC Greater China Test Prep Summit
不確定自己現在幾分？
12 題免費測出分數區間 →
In September 2025, I was invited to attend the GMAC (Graduate Management Admission Council) Greater China Test Prep Summit in Shanghai. This annual closed-door event brought together GMAC’s psychometric researchers, official prep resource managers, business school representatives, and leading GMAT prep providers from across Greater China.
The agenda was packed, but two sessions stood out: the first featured Dr. Chris Han from GMAC’s psychometric team sharing scoring data and test-taker performance analysis for the new GMAT Focus Edition; the final session covered upcoming reforms to the official prep platform.
For the underlying numbers, including how many points one extra correct answer is worth and how section scores convert to a total, see our GMAT Focus scoring research.
This article distills all the critical data and insights from that day into a structured deep-dive analysis. Whether you’re a test-taker in the middle of prep or a fellow GMAT instructor, I’m confident you’ll find strategic value here.
1. Scoring Scale Update: The New Exam Still Follows a Normal Distribution
Let’s start with the basics: this year’s score-to-percentile charts compared to last year show only minor adjustments. The cumulative normal distribution curve remains intact, which means the GMAT Focus Edition scoring scale hasn’t deviated from a normal distribution—unlike the old exam, which became distorted toward the end. The new exam’s credibility remains solid.
Many students ask me: “Why does my score report show a slightly different percentile every time I log in?” That’s normal. Percentiles are calculated using a rolling three-year pool of test-takers, so you might see minor fluctuations daily. But over the long term, the changes are negligible.
The same pattern holds for individual sections. One thing I noticed: Verbal’s middle range is relatively steep, with an inflection point shifted to the right, meaning more test-takers cluster within one standard deviation of the mean, and the median skews higher. Takeaway: if you can break through the mid-range (medium-difficulty questions) on Verbal, you’ll outperform a larger percentage of test-takers.
2. Chinese vs Global Test-Takers: What the Data Reveals
Percentile Comparison: Who Has the Advantage?
This dataset compares the new exam (Gen 11) and the old exam (Gen 10), with blue representing global data and red representing test-takers in China. While it’s unclear if Taiwan is included, test-takers from Taiwan can use this as a reference given similar educational backgrounds.
Reading tip: when red is higher than blue, it means Chinese test-takers have a higher internal ranking (percentile) for the same scaled score. Quick takeaways:
- DI (formerly IR): Chinese test-takers went from a disadvantage to parity with global averages
- Verbal: Chinese test-takers were significantly disadvantaged before, now relatively disadvantaged, with mid-range slope (discrimination) converging to global norms (previously disconnected)
- Quant: Chinese test-takers still dominate, but the near-perfect score range now has finer gradations—fewer people cluster at the top, allowing stronger performers to stand out
Question Completion Rates: Is the New Exam Really More Time-Pressured?
Did the new exam become more time-pressured? The completion rate data answers this:
- Chinese test-takers’ Verbal completion rate is the same as the old exam, but global Verbal completion rates improved slightly
- Chinese test-takers’ Quant completion rate dropped slightly
- DI, which everyone complains feels rushed, actually has higher completion rates than the old IR—so it’s not the exam, it’s you
Response Time Distribution: What Box Plots Reveal
This box plot is fascinating. Quick explainer: upper and lower whiskers represent max and min values (whisker length = range), box top and bottom are 75th and 25th percentiles, and the line inside the box is the median.
What this shows: response time distribution on the new exam is wider and more dispersed, possibly indicating test-takers receive questions with greater variation in nature and difficulty. The median response time is also higher, meaning questions genuinely take longer to answer.
3. Review & Edit Feature: Data-Driven Effectiveness
Verbal: Early Questions Get Corrected Most Successfully
This chart clearly shows initial response time (white dots) and time spent reviewing/editing later (gray dots) for Verbal questions. The bars represent wrong-to-right changes (gray), wrong-to-wrong (black), and right-to-wrong (white).
Key observations:
- Globally, Verbal time pressure increases as the section progresses
- Reviewing a question takes roughly 60 seconds on average
- Most test-takers choose to review early questions (first 10)
- For the first 10 questions, review catches and fixes mistakes at a notably higher rate—over 50% turn wrong-to-right. (GMAC has since confirmed scoring carries no question-order premium; the benefit here comes from locking in your true final answer, not from any special weight given to early positions.)
- Later questions, even when reviewed, mostly result in wrong-to-wrong changes—likely because time pressure during the initial pass prevented adequate comprehension, so revisiting doesn’t help much
Quant: Question One Shows Remarkable Correction Rates
The Quant data is even more interesting. The first 10 questions take significantly longer than the next 11. I raised my hand during the session to point out: the GMAT system tends to load REAL (word problem application questions) in the first 10 questions. While the professor said there’s no such setting, I see enough score reports daily to know better.
Another fascinating finding: question one gets reviewed by a huge number of test-takers, with high success rates for wrong-to-right corrections. My theory: many people enter the Quant section feeling nervous and not fully warmed up, but the question itself isn’t particularly difficult—so revisiting it later usually results in getting it right.
I also asked the professor if this relates to the Wainer Strategy—a deliberate adaptive test gaming tactic where you intentionally miss early questions to lower difficulty, then go back and correct them. Interesting hypothesis.
DI: MSR First Questions Are Review Hotspots
DI also shows the pattern of more time available early, less available late. Test-takers spend roughly 1-1.5 minutes reviewing flagged questions. Notably, questions 1, 4, and 12 are commonly reviewed.
Question one follows the same warm-up hypothesis. Questions four and twelve, the professor believes, are very likely the first questions in MSR sets.
Across all three sections: wrong-to-right correction rates for early questions are at least 50%. Flagging and reviewing is a genuinely useful feature—but the pattern isn’t about early questions carrying more scoring weight. GMAC has confirmed there’s no question-order premium; what actually happens is that a miss early on nudges the CAT toward easier follow-up questions, so review works best when test-takers still remember the question well enough to self-correct, which fades as time pressure builds.
4. Retaking the GMAT: Does It Actually Help?
This dataset answers one of the most common questions: does retaking the GMAT work?
My first reaction: should I open a Dustin GMAT branch in Korea?
But seriously—the next two charts contain a lot of information, but the core conclusions are clear:
- Lower-scoring test-takers show the largest improvement, and gains diminish as scores increase
- 655+ retakers show almost no significant improvement, sometimes even regression
- For 655+ test-takers, more retakes yield diminishing marginal returns
- Chinese test-takers follow the same global trend, but the curve shifts slightly upward, meaning Chinese test-takers perform better than the global average on retakes (especially in the 405-655 range)
Bottom line: if your first attempt was disappointing (e.g., 605 or below), retaking has strong potential for significant improvement. But if you’re already scoring high (e.g., 705+), you need to carefully assess whether retaking makes sense.
5. Section-Level Deep Dive: Strengths and Weaknesses of Chinese Test-Takers
The following series of charts uses box plots to compare accuracy and response time for Chinese test-takers versus global averages.
Quant: Dominant Across the Board, But Word Problems Slow Things Down
Quick takeaways:
- Chinese test-takers are strong in Quant—faster and more accurate
- The variability in Quant ability among Chinese test-takers is lower than globally
- Even when questions are word problems (REAL), accuracy doesn’t suffer
- But word problems increase median response time by roughly 30%, for both Chinese and global test-takers
From an arithmetic vs algebra perspective: global test-takers struggle more with algebra (lower accuracy), while Chinese test-takers excel at algebra, with higher accuracy and faster response times.
Breaking down by content area, Chinese test-takers still dominate, but note:
- Chinese test-takers are noticeably stronger in equations and inequalities, noticeably weaker in counting and sets
- The widest ability range is in value calculation/ordering, with long 25th-75th percentile boxes
- Slowest response times with the highest variability are in rate/ratio problems
Verbal: Chinese Test-Takers Trail Globally, Reasoning Ability Is Key
The common trend: CR is weaker than RC. Chinese test-takers’ CR/RC performance lags the global average significantly, with lower accuracy and longer response times.
Quick primer on Verbal’s four foundational categories:
CR question types:
- Analysis: Boldface, two-party dialogue, etc.
- Critique: Weaken, strengthen, evaluate, etc.
- Plan: Questions involving proposals or plans
- Construction: Fill-in-the-blank, assumption, inference, explain, etc.
RC question types:
- Stated Idea: Main idea, detail questions
- Inferred Idea: Inference, application, evaluation (including function and logic questions)
Key findings:
- Chinese test-takers underperform the global average across all four major Verbal categories
- Globally, RC performance exceeds CR performance
- Chinese test-takers only perform relatively well on RC – Stated Idea (pure literal comprehension, no inference required)—as soon as reasoning is involved, performance drops noticeably
- Response time is dragged down primarily by RC inference/application/evaluation questions, and CR plan/fill-in-the-blank/assumption/explanation questions
My observation: Chinese students desperately need to develop higher-order reasoning skills beyond literal reading comprehension. This is a core emphasis in my teaching approach.
DI: DS Dominance, MSR/TPA Are Achilles’ Heels
The DI data aligns perfectly with what I see on the front lines:
- DS (Data Sufficiency): Chinese test-takers crush it, with accuracy and speed significantly above the global average—because DS is mostly math
- MSR and TPA: Chinese test-takers lag significantly. Both question types are text-heavy and “bloated” (stuffed with irrelevant information), testing reading speed and the ability to filter relevant information from large volumes of text
- Table Analysis: Chinese test-takers unexpectedly underperform, with lower accuracy and slower response times
My hypothesis: most Chinese test-takers only read literally and don’t make higher-order inferences, and they tend to passively absorb all given information rather than filter it. That’s why they excel at math-related questions with minimal text and necessary information, but struggle with TPA/MSR questions packed with distractors. Instructors should focus on teaching “how to quickly filter irrelevant information in TPA/MSR.”
6. GMAT vs GRE: Validity from the Admissions Office Perspective
A hot debate in recent years: should you take the GMAT or GRE? Most test-takers approach this selfishly—”Which is easier? Which helps me hit my target faster?” But this time, GMAC showed us the admissions officer perspective: which exam better reflects applicants’ true abilities?
Score Discrimination: GMAT Wins Decisively
- Left chart: GMAT Verbal’s score distribution more closely resembles a normal curve, with fewer test-takers clustering at high scores. In admissions officers’ eyes, a high GMAT Verbal score is more meaningful
- Middle chart: The difference is even starker—1/3 of GRE test-takers score near-perfect on Quant. What discrimination power does that offer schools?
- Right chart: GMAT tests data interpretation and business insight (DI section), which are directly relevant to modern business management, but GRE has no equivalent assessment
Statistical Reliability and Measurement Precision
GMAT has very high statistical reliability—if the same test-taker retakes the exam without any change in ability, the scores will be highly consistent.
While GRE’s Quant reliability is also high, the asterisk note is the key: 1/3 of test-takers score at the ceiling, making the scores meaningless for differentiation. In GMAC’s words: “Reliability only matters when it provides unique and valuable information.”
Additionally, SEM (standard error of measurement) indicates the gap between a test-taker’s score and true ability (lower is better)—GMAT comprehensively outperforms GRE here.
Response Time Comparison
GMAC claims GMAT provides significantly more time per question than GRE, emphasizing depth of reasoning rather than rapid calculation.
That said, I have to note this slide cherry-picks information. Whether test-takers feel time pressure isn’t just about nominal time per question—it depends on question format. Does a sentence completion question take the same time as a critical reasoning question? So take this slide with a grain of salt—it’s a bit self-congratulatory.
Vocabulary and Prep Focus
Final comparison dimension:
- GMAT requires roughly 6,000-7,000 words, while GRE requires 9,000
- GMAT allows longer average response time per question across all sections, meaning GRE emphasizes speed more
7. Scoring Myths Debunked: GMAC Sets the Record Straight
“If I mess up the first two questions, will my score tank?” This is probably the question I’ve been asked most often.
GMAC’s answer is unequivocal: GMAT scores are calculated based on multiple factors, and the sequence of your responses is not one of them.
Factors that influence your score include: characteristics of questions you got right (e.g., difficulty), characteristics of questions you got wrong (e.g., difficulty), number of unanswered questions (guessing is better than leaving blank).
The following are false myths:
- Taking too long on a question negatively impacts your score
- Changing a correct answer to incorrect at the end penalizes you more than getting it wrong initially
- Certain question types carry more weight than others
I’ve actually covered these in separate videos if you want to dive deeper: Complete GMAT Scoring Mechanism Explained.
8. Official Prep Resources: Future Roadmap
The final session featured GMAC’s prep resource manager sharing feedback from test-takers and prep providers on official products, along with future reform directions.
Pain Points from Test-Takers and Prep Providers
Test-taker pain points:
- Why do different products use different platforms?
- Taking a full-length practice exam without preparation is overwhelming, and seeing terrible results is demoralizing
- Don’t know how to start preparing
- Want to know effective study methods when time is limited
- Want more practice exams and more granular practice question tags
Prep provider pain points:
- Want more difficult questions (especially DI) to help stronger students
- Want teacher-side access to students’ practice exam data
- Want to assign customized practice sets to students
- Current official explanations are too rigid, many practice exam questions lack explanations
GMAC’s Two-Phase Reform Plan
Phase One: Integrate official practice questions like the OG with practice exams, creating a single platform with seamless user experience, integrated with the GMAT website.
Phase Two:
- Build diagnostic functionality: beginners start with brief skill assessments rather than diving straight into full-length practice exams
- Platform provides personalized feedback and study recommendations based on diagnostic results
- Test-takers can practice based on recommendations and review progress, creating a continuous improvement cycle
- Create a dedicated interface for prep providers, allowing teachers to view student performance and build customized practice sets using an improved tagging system
Expected timeline: Phase One (platform integration) by end of 2025, Phase Two by mid-2026. Based on my experience with GMAC—take it with a grain of salt.
Appendix: Exam Rules and Security
One session featured a pre-recorded remote video on exam rules and security measures. I assume most of you are law-abiding test-takers who don’t need this section, but here are a few slides for reference.
Frequently Asked Questions
Q1: Is the GMAT Focus Edition scoring system reliable?
According to data released at the summit, the new exam’s score-to-percentile relationship still follows a cumulative normal distribution curve and hasn’t become distorted like the old exam did toward the end. GMAT’s statistical reliability also outperforms the GRE, meaning the same test-taker’s scores across multiple attempts show minimal variation. Overall, the new exam’s scoring credibility is well-supported.
Q2: Does retaking the GMAT actually improve scores? When is it worth retaking?
Official data shows that lower-scoring test-takers (below 605) see the largest gains on retakes. Chinese test-takers in the 405-655 range even outperform the global average on retakes. But test-takers scoring 655+ show almost no significant improvement on retakes, and those at 705+ need to carefully evaluate. Bottom line: if your first attempt went poorly, retaking makes sense. If you’re already scoring high, think twice.
Q3: How should I use the Review & Edit feature most effectively?
Data shows that for all three sections, early questions (especially the first 10) have a wrong-to-right correction rate exceeding 50%, because test-takers still remember those questions well enough to self-correct. Later questions show poor returns mainly because time pressure during the initial pass prevented adequate comprehension in the first place. Important: GMAC has officially confirmed there’s no scoring premium or penalty tied to question order—a miss early is not inherently harder to recover from than a miss later. Suggested strategy: flag uncertain questions (of any position) during your first pass, and use remaining time to revisit them—prioritizing early questions isn’t about chasing extra points, it’s simply where your memory of the question is freshest.
Q4: What are Chinese test-takers’ weaknesses on the GMAT? How should they improve?
Data clearly shows Chinese test-takers lag the global average across Verbal, especially on reasoning-heavy question types (CR Construction/Critique, RC Inferred Idea). For DI, MSR and TPA are Achilles’ heels because these question types are packed with distractors, testing information filtering and reasoning rather than literal comprehension. Recommended focus areas: advanced logical reasoning, information filtering, critical thinking.
Q5: How do business school admissions officers view GMAT vs GRE?
Summit data shows GMAT has significantly better score discrimination than GRE for both Verbal and Quant (GRE has 1/3 of test-takers scoring near-perfect on Quant, offering almost no differentiation). GMAT’s standard error of measurement (SEM) is also smaller, meaning scores more closely reflect true ability. Additionally, GMAT’s DI section directly assesses data analysis skills needed in business contexts, which GRE doesn’t cover at all. For applicants targeting top business schools, GMAT scores carry more weight.
(Updated August 2026: Revised the Review & Edit section to align with GMAC’s official clarification that scoring carries no question-order weighting.)
