Computer-adaptive tests are still a new feature of the educational landscape. Those of us who have interviewed for work have had the experience of taking computer-based tests. But computer-adaptive tests represent at least one level-up from the usual experience. Here's why.
THE SCORING ALGORITHM
The nerve-wracking feature of the GMAT exam is its scoring algorithm. Computer-based tests deliver a range of questions, at a range of difficulty levels. But CATs, as computer-adaptive tests are known, deliver their questions based on a running estimate of your performance. A sequence of mistakes will cause a CAT to reduce the difficulty level of the next questions you see. As a consequence of this, computer-adaptive tests rarely ask any truly easy questions. After the first few questions at a medium level of difficulty, all the rest are chosen to challenge you.
It is important to complete the entire test, or else the scoring algorithm will apply a penalty downgrade, on the assumption that your performance on the questions you answered was improved by the time saved on the questions you didn't. If this penalty does not deter you, consider that spending too much time on the early questions means you are also likely to generate longer strings of mistakes in the final stretch. That may be enough to convince the algorithm to downgrade its estimate of your baseline performance. The best response to the timing constraint is just to practice until your technique is strong enough to work well under pressure.
Since the adaptive test algorithm is adjusting itself to your performance, it will not allow you to literally skip questions, or to go back and revise your answers to previous questions. You may ask yourself, What should I do when I can't answer one of the questions? As a final resort, choose a random answer, just to move the algorithm forward and avoid tanking the entire test. In fact, you can assume that the test already has a good estimate of your final score, so that any one particular question is not that important. Skipping it - by choosing a random answer - won't make a big difference to your score, provided you return to your baseline performance on the next questions.