Error journal
Every lost mark gets a category — T/F/NG trap, paraphrase miss, spelling, distractor reversal. The Pareto below finds the 20% causing 80% of the damage.
A personal analytics side-quest · IELTS Academic
Instrumenting my own English.
I treat everything like a project — including this. Official baseline: the input channels are excellent. The output channels are the engineering work. This page tracks that work with real numbers, because that's how I'd run it for a client.
- Reading
- 9.0
- the ceiling — there is no 9.5
- Listening
- 8.5
- comfortably above target
- Writing
- 6.5
- the project · target 7.0
- Speaking
- 6.5
- the project · target 7.0
- Overall
- 7.5
- official · Academic
How IELTS computes the overall band: (9.0 + 8.5 + 6.5 + 6.5) ÷ 4 = 7.625, rounded to the nearest half band → 7.5. Quarter fractions round upward, so 7.75 would have been an 8.0. Read the scoring rules before you optimise — requirements first, always.
Gap analysis
The gap, quantified.
Working target: 7.0 in every band. Reading and Listening already clear it. Writing and Speaking sit half a band short — and a half band is a project, not a wish. The tick marks the target.
Telemetry
Progress, plotted.
One data point so far — the official test. Every timed mock from here on lands in js/ielts-data.js as one record, and these charts redraw themselves. No dashboard tool, no export step — the site is the dashboard.
The operating rhythm
Run like a delivery, not a grind.
Writing and Speaking don't improve by "more practice" — they improve by scoring against the public band descriptors, finding the two categories that cost the most, and fixing those first.
Criterion scoring
Writing and Speaking are marked on four public criteria each. Mocks get self-scored per criterion — because "6.5 overall" hides exactly which lever to pull.
Recorded speaking
Part 2 answers recorded and reviewed against the descriptors — plus shadowing for rhythm and linking. Painful to listen to; unreasonably effective.
One fix per session
Each mock ends with a single carried-forward action. A retro that produces twelve actions produces zero.
Error Pareto — waiting on data. Once the first mock's error journal is logged, this panel charts which categories are actually losing marks, worst first.
The log
Every attempt, on the record.
-
IELTS Academic — the baseline
Listening 8.5 · Reading 9.0 · Writing 6.5 · Speaking 6.5 — overall 7.5. Reading maxed out; Writing and Speaking scoped as the improvement project.