Estimating: The Skill Behind the Skill
Exact answers get the attention. Estimates do the work.
Almost everything useful about being good with numbers comes down to knowing roughly what an answer should be before you have it — because that is what tells you when something has gone wrong, and going wrong is far more common than being slow.
Why the rough answer is the important one
A calculator will not tell you that you typed 4700 instead of 470. A spreadsheet will not mention that you picked the wrong row. Neither of them knows what you meant. The only thing standing between a slipped digit and a decision made on nonsense is somebody thinking “that looks about ten times too big”.
Most consequential number errors are not arithmetic errors at all. They are input errors that nobody challenged.
Round to one digit, then fix
The fastest estimate keeps a single significant figure and tracks the size.
317 × 48→ 300 × 50 = 15,000. The true answer is 15,216, so the estimate is within 2%.6,240 ÷ 79→ 6,000 ÷ 80 = 75. True answer: just under 79.19.6 × 41→ 20 × 40 = 800. True answer: 803.6.
Notice what happened in the first one: rounding 317 down and 48 up pushed the errors in opposite directions and largely cancelled. That is not luck, it is the technique — when you can, round one number up and the other down.
Count the zeros separately
The digits and the magnitude are two different problems, and mixing them is where decimal slips happen. Do the digits first, then count.
0.004 × 2,500: the digits are 4 × 25 = 100. Now the size: 0.004 is 4 thousandths, 2500 is 2.5 thousands, and the thousands cancel the thousandths. So the answer is 10, not 100 or 1.
Doing it in that order means a mistake in the size never contaminates a correct set of digits.
Anchors worth carrying
Estimating quickly depends on having a few reference points that need no calculation.
| Anchor | Useful for |
|---|---|
| A year is about 500 working hours per day-of-week | Time and staffing |
| 1 in 8 is 12.5%; 1 in 6 is about 17%; 1 in 3 is 33% | Reading proportions |
| Doubling seven times is roughly ×128 | Growth and compounding |
| ×1.1 five times is roughly ×1.6 | Repeated small rises |
Fermi questions
How many piano tuners are there in a city? Nobody knows, and that is the point — the value is in the chain. Population, households per person, pianos per household, tunings a year, tunings a tuner can do. Each step is a guess you can defend within a factor of two or three, and the errors partly cancel.
The habit transfers directly to work: how many support tickets will a feature generate, how long will this take, is this number plausible. You are not looking for accuracy. You are looking for whether something is roughly ten, roughly a thousand, or roughly a million — which is nearly always the decision that actually matters.
Estimating inside a game
Multiple choice is estimation made explicit. If the options are 306, 314, 286 and 412, you do not need the product — you need to know it is a little over 300 and ends in a 6. That is two facts, both quicker to get than the full multiplication.
This is the largest single source of speed in Carry the One, and it is the one that transfers furthest outside it. The techniques get you the exact answer when you need it; estimating tells you when you do not.
How to practise it
Guess before you calculate. Every time, out loud if you can, even when the exact answer is a second away. The guess costs almost nothing and it is the only way the instinct develops — an answer you did not predict teaches you nothing about the next one.
More like this: all the mental arithmetic guides.