Drill the archetypes: free
The Insight Bay practice portal includes SAT-aligned question sets covering every archetype in this article, graded across four difficulty levels so you can train at race pace.
Open SAT Math practice →Below 700, SAT Math is a content test: learn more, score more. Above 700, it quietly becomes a different exam: an error-management test where a handful of question archetypes and a handful of avoidable mistakes decide everything. Studying harder in the old way is exactly how scores get stuck.
Here is the arithmetic nobody tells students: on the Digital SAT's Math section, the difference between 700 and 780 is typically only a few questions per test. Near the top of the scale, every single error is expensive, which means the skills that got you to 700 (knowing the content) are not the skills that get you past it (not giving points away). The plateau isn't a knowledge problem. It's a precision problem wearing a knowledge problem's clothes.
The Digital SAT Math section is short: two modules of 22 questions, 44 in total, of which a couple are unscored experimental items. A 700-level student is already answering the large majority correctly. The remaining gap to 800 is concentrated in perhaps six to ten questions, and those questions are not randomly distributed. They cluster into recurring archetypes, and into error types that have nothing to do with mathematical ability.
The Digital SAT is section-adaptive: your performance on Module 1 determines whether Module 2 is the harder or easier variant. The top score range is effectively reachable only through the harder Module 2, and the harder module is precisely where the eight archetypes below concentrate. Two consequences follow:
Across official practice tests and the students I coach in the 650-780 band, the questions that separate 700 from 780 collapse, again and again, into eight skeletons:
| # | Archetype | What it really tests |
|---|---|---|
| 1 | Quadratic-linear systems | Finding intersections / "for what value of k does the system have exactly one solution": discriminant fluency. |
| 2 | Vertex form & extremes in context | Translating "maximum height" or "minimum cost" into completing the square or vertex reading. |
| 3 | Exponential vs linear models | Recognising percent-change language ("increases by 4% per year") as a growth factor, not a slope. |
| 4 | Equivalent expressions with parameters | Matching coefficients: "ax² + bx + c is equivalent to…, what is a + b?". Disciplined expansion. |
| 5 | Circle equations | Completing the square to extract centre and radius from the general form. |
| 6 | Similar triangles & trig ratios | Spotting similarity inside composite figures; right-triangle trig without a diagram. |
| 7 | Statistics interpretation | Margin of error, sampling validity, and what a study design does and doesn't justify concluding. |
| 8 | Constraint word problems | Building a system from sentences with units and rates: translation under load, the SAT's favourite disguise. |
From 700 upward, every review session should classify each lost point into one of four causes, because each has a different cure:
Why the log works: error-specific feedback followed by immediate corrected retrieval is among the most reliable interventions in the testing-effect literature. Students remember corrected errors unusually well, the "hypercorrection effect", but only if the correction happens while the attempt is still fresh. Same-day review isn't a preference; it's the mechanism.
The built-in Desmos calculator is the single most under-exploited resource in the 700 band. The right policy is neither "always graph" nor "real mathematicians don't": it's a decision rule:
| Weeks | Focus | Sessions (3 × ~50 min/week) |
|---|---|---|
| 1 | Baseline + taxonomy | One full adaptive practice test. Build the error log with the four-cause taxonomy. Identify your 2-3 archetypes. |
| 2-3 | Archetype drills | One archetype per session: 10-12 questions, increasing difficulty, Desmos decision rule practised explicitly. Five-question retrieval warm-up from previous sessions. |
| 4 | Module-1 discipline | Timed Module-1-style sets with a checking protocol: underline targets, re-read final sentences, verify one in three answers in Desmos. Goal: zero careless losses. |
| 5 | Hard-module immersion | Hard Module-2 sets under slight time compression (race-pace training). Error log by cause, not topic. |
| 6 | Full rehearsals | Two full adaptive tests on separate days, exam-condition. Compare error-cause counts with week 1: the score follows the cause-counts. |
The Insight Bay practice portal includes SAT-aligned question sets covering every archetype in this article, graded across four difficulty levels so you can train at race pace.
Open SAT Math practice →One honest caveat: plans like this assume the content base is real. If the audit above scattered ticks everywhere, the plateau story isn't yours yet: a content-first programme (and possibly a structured SAT course) will move you faster than error discipline will.
The free assessment includes a timed SAT Math section, a cause-classified error review, and a written plan: the same diagnostic I run with every new SAT student.
Book the free assessment