Case interview mental math

Two small habits you can memorize handle almost every multiplication or division that shows up live. The first is round-and-correct. Round one factor to the nearest clean number, multiply, then correct for the difference. Say you're estimating $47 per unit across 8.4 million units. Round 47 up to 50 and compute 8.4M × 50 = $420M. Then subtract what you overcounted, 8.4M × 3 = $25.2M, and you land on $394.8 million. That's the same answer direct multiplication gives, from two easy multiplications and one subtraction instead of one hard one. It works on smaller numbers too. Take 42 × 68 as 42 × 70 (2,940) minus 42 × 2 (84) and you get 2,856, exact, in roughly three seconds.

Which mental math tricks are worth memorising?

The second habit is a small table of friendly-factor tricks for the divisors and multipliers that turn up constantly in business arithmetic. Each one is just a restatement of a familiar factor of 10:

Every one of these is exact. They're just a faster route to the same answer. Spend a few minutes memorizing the pattern (not the individual instances) and it pays for itself in the first case you ever run under a clock. Nothing on the list needs working out in the room, as long as you've seen the patterns often enough beforehand.

The third habit is a small fraction-to-percent table. Case numbers often arrive as "one in six," "one in seven," or "about a third," and converting on the spot beats reaching for long division:

What do these tricks look like in a live case?

In a live case it looks like this. A category has 340 million households, and a brand's penetration is stated as "one in six." The estimate is immediate: 340M × 1/6 ≈ 56.7 million households, with no long division at all. Or a market is sized at ₹49,000 crore and a segment is "roughly a seventh" of it. That's 49,000 ÷ 7 = ₹7,000 crore, exact, and instant once you recognize the fraction instead of computing it from scratch.

Not every multiplication has to be exact to be useful. Estimate 8.37 million units at $46.80 by rounding to a friendly 8.4 million at $47, and you get $394.8 million against a precise $391.72 million. The error is under 0.8%, comfortably inside the ±2% working tolerance used throughout, and you got there in a fraction of the time exact multiplication would take. Part of the skill is knowing when that trade is a good one, and being able to state roughly how big the resulting error is instead of waving at it.

Smart rounding and keeping error bounded across a chain

One rounding decision rarely matters. The risk is a chain of rounded steps, where small errors can compound in the same direction and push the final answer further than any single step suggests. Try a quick check. Multiply three factors rounded to two decimal places, 1.08 × 1.07 × 1.06, standing in for three successive rates rounded from 1.083 × 1.071 × 1.062. The result comes out roughly 0.56% below the unrounded chain. That's small and well inside tolerance, but it's the kind of number worth checking instead of assuming, especially as the chain gets longer or the rounding coarser. Half a percent sounds like nothing, and on its own it nearly is. The worry is what happens when you stack several of them.

To keep it under control, round late. Carry an extra digit through the intermediate steps and round only the final answer, because rounding early lets small errors compound at every later multiplication. Say your rounding direction out loud, too. "I'm rounding that up to fifty for speed" tells the room how much slack you added and which way it leans, so nobody mistakes your approximation for false precision. And know when tolerance has to tighten. The ±2% working convention is fine for narrative and directional numbers. The moment a recommendation hinges on which side of a specific threshold you land (a breakeven test, a hurdle rate, a go/no-go cutoff), stop rounding early and do the tighter pass.