A TOEIC total of 735 looks precise. It can support an application threshold or show change across two test dates. It cannot tell a learner what to do at 7:20 tomorrow morning. Study plans should start from the separate skill scores and the tasks underneath them. For a 31-year-old project coordinator in Hsinchu, a reading strength cannot repair missed details in a supplier call. The next 4 weeks should follow the listening evidence, even when the total score looks respectable.
The sections matter.
The total compresses different performances
The TOEIC Listening and Reading test has 200 multiple-choice questions. The current ETS test information specifies 100 listening questions in 45 minutes and 100 reading questions in 75 minutes. Each section receives a scaled score from 5 to 495; the reported total runs from 10 to 990. Two candidates can reach 735 through very different pairs. One may score 420 in listening and 315 in reading. Another may score 330 and 405. The total supports the same administrative statement for both people. Their Tuesday practice should look different.
ETS publishes separate score descriptors for listening and reading. The descriptors distinguish tasks such as inferring a speaker's purpose, connecting information across a text, handling paraphrase and understanding less common vocabulary. They describe typical strengths and weaknesses around score bands, not a diagnosis of one person's every error. A section score also contains measurement uncertainty. It deserves interpretation beside actual work samples.
GEPT makes the skill structure visible in another way. The official GEPT information brochure reports listening, reading, speaking and writing and maps its levels to the CEFR. The GEPT site also explains that listening and reading raw answers are converted to scaled scores so results from different administrations can be compared, while speaking and writing responses receive trained human ratings. The official scoring FAQ says each speaking or writing response is rated by two teachers, with an additional senior rater when their scores differ beyond the stated tolerance.
Different skills produce different evidence.
Start with the largest useful gap
Write the four skills on one page. Beside each, record the latest formal score, one task the learner can complete, one recurring failure and one piece of recent evidence. A score without a task stays too abstract.
For the hypothetical project coordinator:
| Skill | Recent evidence | Working decision |
|---|---|---|
| Listening | Loses dates after a speaker corrects them | Practise corrections and number sequences |
| Reading | Finds contract details accurately | Maintain with two short texts each week |
| Speaking | Gives updates, pauses inside noun phrases | Record one 90-second update weekly |
| Writing | Clear emails, weak request softeners | Revise two real-style requests |
Choose the gap that affects the learner's actual purpose and has evidence behind it. The smallest score may come from a skill the learner rarely needs. A graduate student preparing for seminars may place listening ahead of a slightly weaker business-writing score. A customer-support worker who writes 23 tickets a day may decide differently. I would give one weak skill half the study time, two supporting skills the next 40%, and maintenance the remaining 10%. Those numbers are a planning preference, not a finding from ETS or LTTC. They force a decision while leaving room to revise it after 2 weeks.
Turn a descriptor into a task
“Improve listening” is too broad. If the score report points to difficulty connecting information across utterances, build a task that requires that connection. Use a 70-second workplace recording containing an initial delivery time, a correction and a final reason. Listen once and write the final time plus the reason. Listen again, mark where the correction began, then compare with a transcript.
Our article on opening the transcript after the second listen gives a sequence for keeping the first attempts auditory. Use that sequence for 11 minutes, three mornings a week. Change speakers and situations while keeping the target operation stable. The reading version could use two short emails with a changed deadline. The speaking version might require a 90-second handover that includes the problem, current status and next owner. The writing version might ask for a request to a senior colleague, then compare the message with a rubric for clarity, reason and degree of imposition.
Now the skill label has become observable.
Keep the comparison fair
A full official test and a six-item classroom check do different jobs. Use the classroom task to steer weekly practice, reserving another official or well-matched form for the point when the learner needs evidence of score change across equivalent conditions. Repeating the same practice set can improve familiarity with that set and create a flattering graph. Track three small measures: accuracy, time and independence. Suppose the learner catches the corrected delivery detail in 4 of 9 recordings during week 1, then 7 of 9 during week 4. Also note whether a second listen or transcript was needed. Seven answers obtained only after reading the transcript do not show the same listening performance as seven first-listen answers.
Scores move for reasons beyond instruction. Sleep, test form, topic knowledge and guessing add noise. One section dip should trigger inspection, not a complete course redesign. Look for the same pattern in the score descriptor, the learner's errors and a second task.
Rebuild the plan every four weeks
At the end of week 4, repeat a parallel task. Keep length, number of speakers and information density close to the first check, while changing the topic and wording. If corrected details now survive the first listen, reduce that strand and move time to the next supported gap. If they still disappear, inspect the cause: number recognition, word boundaries, working memory, or unfamiliar correction phrases such as “make that Thursday the 17th”. The total score still has a job. A university, employer or scholarship committee may need one stable summary for a decision. The learner needs a map with more streets on it.
Tomorrow at 7:20, the Hsinchu coordinator can play one 70-second call, write the final date and reason, and wait until the second listen before opening the transcript on the train to work.