4.3 Numerical Speed & Estimation Under Time

The shared clock: numerical items do not get their own timer

The scored Job Opportunities Assessment is 51 questions in 20 minutes. That is:

[ \frac{20 \times 60}{51} \approx 23.5 \text{ seconds per question} ]

on average across numerical, verbal, and abstract items together. There is no separate Mathematical Ability block with its own 20-minute allowance. Every extra 40 seconds you pour into one stubborn matrix is 40 seconds stolen from a verbal analogy or abstract sequence you might have solved cleanly.

Official performance language matches this design: complete as many as possible, as fast and accurately as possible; finishing all 51 is rare. Numerical speed strategy is therefore not “become a human calculator.” It is “extract as many high-confidence numerical marks as possible without tanking the rest of the paper.”

What “fast” means for series and matrices

These are personal budgets, not official section splits. Adjust slightly to your strengths, but do not pretend a one-minute numerical grind is free.

Estimate vs compute: a decision rule

Compute exactly when

• The interface wants a typed number (no options to lean on).
• Multi-choice options are neighbours (e.g. 54 vs 52 vs 56) where a one-off difference error picks the wrong choice.
• The rule is already clear and the arithmetic is short (single add, single multiply).

Estimate or compare when

• Options are far apart in magnitude (12 vs 48 vs 200) and you only need the right scale.
• You are checking whether a candidate rule is plausible before fully expanding it.
• You are in the final minute and need a best remaining choice on a flagged item with partial understanding.

Worked example — estimation enough

Series: 2, 6, 18, 54, ? with options 100, 162, 200, 250.

You see ×3. 54×3 is a bit more than 150, clearly not 100 or 250. Even a rough 50×3=150 points to 162. Exact multiply confirms 162, but the estimate already killed three distractors in two seconds.

Worked example — estimation not enough

Series differences suggest next term 54 vs options 52, 54, 56, 58. Estimation that “it’s about fifty-something” is useless. Finish the second-difference arithmetic and pick 54.

Worked matrix — mixed approach

11  14  25
19   6  25
17   8   ?

If you suspect first+second=third, estimate 17+8≈25, then compute 25 exactly. If options were 20, 25, 30, 100, the estimate already selects 25; still better to confirm 17+8=25 in one second because typed entry or close options punish slips.

Micro-techniques that save seconds

1. Difference ticks, not essays

For series, jot tiny gaps: +4 +4 +4 or +2 +4 +6. Do not rewrite the whole series twice.

Worked micro-scratch

7 12 17 22 ? → gaps 5 5 5 → 27. Three marks of scratch, under 10 seconds.

2. Ratio glance before hybrid algebra

If terms balloon (3, 7, 15, 31…), ask “near double?” 7 is near 2×3, 15 near 2×7 — test ×2+1 immediately.

3. Row sum first on grids

Adding three small integers is usually faster than inventing diagonal lore. Many JOA-style missing numbers collapse to a uniform row/column sum or a left∘middle=right pattern.

4. Anchor numbers

Know automatic facts so you do not recompute: doubles to 20, multiples of 5, 10× tables, and squares of small integers (if they appear). This is number sense, not a trigonometry syllabus.

5. Avoid fake precision

If the stem produces an integer rule and an integer blank, do not introduce decimals. If you “need” a fraction mid-rule, re-check whether a simpler integer rule fits — often it does.

Flag strategy specifically for numerical items

Chapter 3 covered global flagging; apply it ruthlessly to numbers.

Pass 1 — harvest (most of the paper)

• Answer every series/matrix whose rule appears within one quick classification cycle.
• Flag when: alternating braid is unclear, two hybrid rules compete, or the grid needs more than one full re-scan.
• Do not flag easy items “to recheck later” — that creates a false second workload.

Pass 2 — return with remaining time

• Start with flagged numerical items where you already saw a partial pattern (e.g. “odd positions climb, even positions unclear”).
• Skip still-cold abstract or verbal flags if a half-solved series can finish in 20 seconds — maximise probability of a correct mark per second.

Pass 3 — last minute

• For multi-choice, eliminate impossible magnitudes and pick the best remaining option.
• For typed entry, only submit if you have a rule you trust; a wild typed guess is pure noise. If the interface allows leaving blank, follow on-screen rules; if it requires an entry, use the best verified candidate you have.

Worked pacing scenario (numerical-heavy stretch)

You hit four numerical items in a row at minute 8:

1. Series +4 constant — 12 s, answered
2. Matrix product rows — 20 s, answered
3. Alternating series — 30 s, no lock → flag
4. Hybrid ×2+3 series — 18 s, answered

Total ~80 seconds for four items (about 3.4 average budgets). You banked three marks and parked one. Linear stubbornness on item 3 for 70 seconds might have cost item 4 and the next verbal stem.

Integrity and “speed aids”

Speed must come from your reasoning. Outside calculators, second-device helpers, or another person solving the matrix violates JOA integrity rules and can trigger supervised verification issues. Train so that 23-second decisions are yours alone on a quiet supported laptop.

Building numerical speed before test day

Practice should mirror the assessment’s purpose: natural reasoning under time, not school exam cramming.

When you review, label each miss: wrong family, verification skip, arithmetic slip, or time panic. Different labels need different fixes. Arithmetic slips need slower final checks on typed items; wrong family needs more difference/ratio openers; time panic needs a stricter flag rule.

Estimation drills (optional 3-minute finisher)

Give yourself only 8 seconds per multi-choice series with far-apart options and force an estimate-first elimination. Then spend 5 seconds confirming. This trains the estimate-vs-compute switch so it is automatic on test day.

Putting sections 4.1–4.3 together

Numerical reasoning on the JOA is a toolkit:

1. Series tools — first/second differences, multiply-add hybrids, alternating splits (4.1).
2. Matrix tools — row/column first, then diagonals; typed-entry care (4.2).
3. Clock tools — 23.5-second mindset, estimate vs compute, flag/return (4.3).

None of these require Pythagoras, SOH-CAH-TOA, geometric proofs, or a full algebra course. They require calm pattern spotting, short accurate arithmetic, and the humility to flag when a rule will not appear in time.

Final worked synthesis

Item: series 6, 13, 27, 55, ? options 110, 111, 112, 120; 20 seconds left on your personal budget.

• Classify: gaps +7, +14, +28 → differences doubling, or ×2+1: 6×2+1=13, 13×2+1=27, 27×2+1=55.
• Compute: 55×2+1=111.
• Estimate backup: a bit over 100, not 120 if the +1 hybrid holds.
• Answer 111 and move — do not re-derive a cubic formula.

That is JOA numerical performance: simple verified rules, executed fast, abandoned when stuck, so your Job Opportunities Report reflects real reasoning skill rather than stubbornness on a single cell.