Good mental math is not fast calculation. It is problem rewriting.
When someone computes 94 × 98 in a few seconds, they are almost certainly not running through columns of digits in their head. They are noticing that both numbers are close to 100, and using that structure to simplify the work. The arithmetic they actually do is easier than what you would do if you attacked the problem head-on. What looks like speed is usually a better choice of representation.
Raw calculation is the wrong baseline
Most people, when they try to do mental math, attempt to simulate written arithmetic in their heads. They carry digits, track columns, and try to hold partial products in working memory while generating new ones. This is genuinely hard. Working memory is limited, and written arithmetic is explicitly designed to offload that burden onto paper. Trying to do it without paper is not a test of mental math skill. It is a test of how well you can use your brain as scratch paper.
The experience this produces — effortful, fragile, prone to losing your place — is what makes mental math feel mysterious or talent-gated. It isn't. It's just the wrong approach.
Take 213 − 67. The written-arithmetic path means borrowing from the tens column, tracking the regrouping, hoping you don't lose a digit along the way. But there's a simpler question buried in this problem: how far is it from 67 to 213? Count up: 67 to 70 is 3, 70 to 200 is 130, 200 to 213 is 13. Add those jumps: 3 + 130 + 13 = 146. No borrowing. No columns. The problem went from uncomfortable to easy the moment you reframed it.
Or you can add 3 to both 213 and 67, to make this problem look like (213+3) - (67+3) = 216 - 70, now you only need to deal with 21 - 7 , and you get 146 easily.
Or you can minus 13 to both numbers, to solve (213-13) - (67-13) = 200 - 54, now you have a clean add-to-a-hundred question, you get 146 again.
The crucial part is how to reframe the problem in a way that is easy for you.
The major transformations
There are a handful of moves that cover a large fraction of everyday mental math. They are not tricks in the gimmick sense — each one is a genuine structural insight that applies across many problems.
Decomposition means breaking a problem into friendlier pieces. 47 + 36 becomes 47 + 30, then + 6: you add the round part first (getting to 77) and then finish with the small remainder. For multiplication, 18 × 25 is awkward until you notice that 18 = 9 × 2 and 25 × 2 = 50, so 18 × 25 = 9 × 50 = 450. You've traded one ugly product for one easy one.
Compensation means rounding to a clean number and then correcting. 49 + 38 is harder than 50 + 38 − 1. You overshoot deliberately, then adjust. 201 − 98 becomes 201 − 100 + 2 = 103. The round numbers do almost all the work; the correction is trivial. This move works whenever a number in the problem is close to a multiple of ten or a hundred — which, in practice, is often.
Counting up is what subtraction often wants to be. Rather than 213 − 67, ask: what do I add to 67 to reach 213? This reframe turns subtraction into navigation, which many people find much easier to hold in working memory. It also explains why making change mentally is usually easier than subtracting — cashiers learned to count up, not borrow.
Benchmarking is the move that makes percentages manageable. Anchor on something you know: 10% of any number is just a decimal shift. 15% of 80 is 10% (8) plus half of that (4), giving 12. 37% of 200 is 25% (50) plus 10% (20) plus 2% (4), giving 74. You are never computing 37% directly. You are assembling it from parts with known values.
Near-round multiplication applies when both factors are close to a convenient anchor. 94 × 98: both are near 100, with deficits of 6 and 2. Take one number and subtract the other's deficit: 94 − 2 = 92. Multiply the two deficits: 6 × 2 = 12. The answer is 9,212. This isn't a special trick — it's a direct consequence of how multiplication distributes. But recognizing when a problem has this shape, and reaching for it automatically, is the thing that looks like talent from the outside.
Estimation as a first step deserves its own mention because it changes how you approach every problem. Before computing, ask: how large should the answer be? 47 × 53 is close to 50 × 50 = 2,500. When you get 2,491, you know you're in the right zone. When you accidentally get 24,910, you know something went wrong. Estimation is not a fallback for when you can't compute precisely. It is part of the skill, and often the most important part.
There is rarely just one path
The transformations above are not a menu where each problem has one correct entry. Most problems yield to several different approaches, and part of developing fluency is noticing that.
Take 18 × 25. Two different paths, same answer:
- Halve and double: 18 = 9 × 2, so swap — 9 × 50 = 450.
- Compensation: 20 × 25 − 2 × 25 = 500 − 50 = 450.
None of these is the trick. Each one just notices a different shape in the same problem.
Or take 49 + 38. Two easy routes:
- Compensate on 49: 50 + 38 − 1 = 87.
- Compensate on 38: 49 + 40 − 2 = 87.
Which one you reach for depends on what you see first. A fluent mental calculator doesn't follow a fixed procedure — they scan for the cleanest entry point and take it. Two people can arrive at the same answer by completely different paths, both having done almost no hard arithmetic.
This multiplicity is worth sitting with, because it changes how you should think about being "wrong" when you practice. If you compute 18 × 25 by compensation instead of halving, you weren't wrong to use a different path. The goal is not to learn which transformation belongs to which problem. The goal is to make enough transformations automatic that you always have options.
Why this looks like talent
The restructuring step is mostly invisible to observers.
When someone computes quickly, you see the answer. You do not see the 49 → 50 move, or the recognition that 94 and 98 are both near 100. What gets attributed to natural ability — fast recall, quick fingers, some arithmetic gift — is usually just a person choosing a better method to solve the problem than the original question.
This is also why mental math fluency is hard to acquire by watching demonstrations. The demonstrator is showing the end of the process. The thing to learn is the beginning: the habit of pausing before calculating, scanning the problem for structure, and asking whether there is a cleaner path. That habit is teachable, but it is not visible in the performance.
What looks like speed is usually better representation.
What this implies about practice
If mental math is pattern recognition, the implication is direct: you learn it by seeing patterns repeatedly, not by drilling raw computation faster.
Useful practice exposes you to the same structural families across many different numbers — so that compensation, or near-100 multiplication, or counting-up subtraction, starts to feel like a natural reach rather than a technique you have to consciously invoke. You want the shape of the problem to trigger the right move, the way a chess player recognizes a fork without counting out all the branches.
What doesn't help much: collecting isolated tricks, or running timed drills without attention to strategy. Speed is a byproduct of fluency; drilling for speed before fluency is established mostly builds anxiety.
The next post in this series covers the practical side: how to build these habits deliberately, what to practice and in what order, and how to avoid the common mistake of drilling facts without ever building the strategic layer on top of them.