# Mental Math Practice: Getting the Most “Bang for your Buck”

In case interviews, you may be asked to undergo mental math gyrations from Minute 1. Consider the following excerpt from this sample case.

This deluge of information might come as a series of facts, sprinkled throughout the case – or it might arrive after a two-sentence prompt citing a client’s desire to reduce their R&D costs.

“*[…] All of the R&D facilities and staff are concentrated in the Boston area. There are approximately 2,300 R&D employees and the total R&D costs are approximately $12B per annum. The client has traditionally received 80% of its total revenue ($24B) from drugs to treat heart disease, blood pressure, and liver disease; 80% of the patents on these [three] drugs will expire by 2015, and the company therefore expanded research into adjacent areas in 2010. Not getting satisfactory results, the company expanded the research portfolio in 2012 *[emphasis mine]

*.*”

If you’re given all this information as a trainee, candidate, or interviewee, you’re likely to be experiencing hand cramps in the near future. Yet the interviewer will also expect that the cogs in your mind are turning, too – that is, that they are doing more than simply holding these bolded numbers in working memory and sending nerve impulses to your forearm to ensure you write everything down before you miss something. To put it bluntly, you should be trying, even as you jot down your notes, to “size” some important quantities, and wonder about things like:

- The total R&D cost
*per employee* - The Boston company’s total revenue
- The time remaining until the patents expire – as well as which patents will expire!

The degree of mental math fluency required to be successful as a consultant cannot be understated. In this article, I’d like to spend some time discussing a subset of mental math skills surrounding (no pun intended) the art of *rounding* numbers, and make suggestions for practice.

## The Art and Necessity of “Rounding” for Mental Math

If you’re reading this article, you most likely know that the art of rounding will be critical to your mental math performance in case interview endeavors. Perhaps you’ve been taught it is necessary for reasons having to do with speed, efficiency, or capacity (what *number of numbers* you can keep in mind for a time, without forgetting). All three of these categories are valid!

Rounding is clearly useful. One motivating example can be found in the multiplication of two numbers — often, and fortunately for us, this can be observed even when a problem is easy enough to solve exactly. Take as an instance the following problem, from a sample profitability case: what is 12% of $12B? Anyone who still remembers their multiplication (or “times”) tables can see that, since 12 x 12 = 144, and one of the numbers is a percentage, we can simply move the hitherto invisible decimal point from behind the second “4” to rest between the “1” and the first “4,” giving an answer of $1.44B. Yet we could have rounded to estimate this one as 10% of $14B (i.e., $1.4B flat), or even 14% of $10B, and we’d have come up with an answer of $1.40B – only 4/100 of $1B (only – yes, only, because $40 *million* is a *comparatively* small amount) off!

By cleverly rounding one answer slightly *up* and the other *down* (alas, this “toy problem” probably represents the upper limit of how much rounding is safe to do; I say so intuitively now, but will make this intuition mathematically precise below!), one can sometimes lessen the work by virtue of dealing with numbers that are more familiar, or easier to multiply, than the original numbers in the problem.

## What about the exact solutions?

That said, there always those dreaded moments, where interviewers asks candidates to provide “exact” answers; and any short-lived affair with your favorite search engine can turn up seemingly contradictory pieces of advice for how to handle such situations, or how to approach and emphasize rounding techniques in your case interview training (possibly, at a detriment to the honing of other skills).

For instance, even when rounding is strongly encouraged, how conservative should you aim to be? One might observe Felix, who had entered her training with good numeracy already, and wonder if her rounding of “90% of $14.4M” to “12.9” (instead of 12.96) was advisable, or if her rounding of “90% of 12.9” to “11.6” was necessary (or whether she could have, instead, just solved “90% of 13”) during her guided estimation of the number of clothing irons in Spain.

Especially if you’re good with numbers like Felix, or happen to be a lightning-fast human calculator like Dr. Arthur Benjamin, it might even seem moot to waste time with fancy rounding techniques in the middle of your calculations: might not it be better to compute exact answers, even if you report them to your interviewer in approximate terms?

I believe that the answers is a firm “no,” for there is a much more compelling reason to become fluent in rounding than is often discussed!

**A Better Reason For Learning to “Round” in Mental Math**

My reason is a simple one: many of those “exact” mental math techniques you can learn (and Professor Benjamin’s book teaches a lot of them) necessitate the use of rounding, in order to make them work! A perfect example is the technique of performing subtractions by addition of *complements*: 980 minus 39 can be transformed into “980 – 40 + 1,” where “980 – 40” can be a lot easier to visualize than the original problem, and– crucially – it can easily be corrected, via appending the final “+1” operation. Another, which I’ll display later, is a more subtle technique for multiplying numbers: one that requires the user to decide which *round*number is closest to the original number being multiplied.

Beyond this reason, which deals with mental math “as such,” there are, of course, other reasons to practice rounding for a __case interview__. For instance, interviewers may try to simulate what an executive purports to care about in a given calculation, and the “bottom lines” in these discussions are often expressed as *order of magnitude* estimations (see comments on assessing “competing effects” on the basis of relative size in my recent article 3 Case-Interview Concepts I Learned From Physics). Indeed, in the video from which the “12% of $12B” example derives, the answer itself is even is rounded to $1.5B, since the interviewee cares only about demonstrating that other categories of costs *besides* labor (which makes up the aforementioned 12%), are two times this size in dollar value!

In the subsequent section, I’ll spend some time unraveling a few of the techniques that Art Benjamin espouses in his work. While I’ve never known him personally, his presentations of mental math skills were extremely inspiring for me, as a young physicist, the slam-dunk reason I have provided for “learning to round,” specifically, was one inference, or generalization, that I’d made by reading his __Secrets of Mental Math__, to which I also linked earlier. I will spend time on a few of his strategies that I’ve actually used in preparing for case interviews, and move on in the next sections to both a rule for rounding and some general advice on mental math in business.

**Mental “Mathemagics”: Tips for Playing with Numbers**

Benjamin is the expert, but I (one of many) have profited tremendously from adopting some of his practical tips for negotiating mental arithmetic. So, I’d like to show you some of the tips and tricks for case interview math that I’ve internalized, using examples from his book.

I’ll break these into two categories: the first will be “General Go-To’s,” and the second a couple applications of “Specific Strategies” for multiplying non-round numbers. Since division is, in essence, multiplication by a reciprocal (at least in everyday arithmetic!), both approaches can also be used to engage quotients, ratios, and fractions – you’ll see why after General Go-To #2!

After reviewing these mental math tips, I’ll revisit the topic of trying to approximate an answer by rounding the numbers in a multiplication problem – the multiplicand and multiplier – in opposite ways. In particular, I’ll try to show how you can be confident that their product will be within the ballpark, given how much you’re rounding the individual numbers that are being multiplied. This latter subsection will be more technical – feel free to skip to the final section!

## General Go-To’s

*Go-To #1: How Many Zeros?** *

Even as a trained scientist, my sticking point in a “back-of-the-envelope” calculation can often be the order of magnitude assessment. To parallel what I have stated here, is the ratio of 700 billion to 46 million best expressed in millions? Hundred-thousands? I’d say *ten-thousands* (unless I had mistakenly divided by 460 million and come up with an answer in the thousands!):

700,000,000,000 *divided b*y 46,000,000 *is approximately* **15,**217

The usually touted trick is to express the answer in some scientific notation, or separate an appropriate number of zeros by converting the first few integers to decimal coefficients — as many as necessary to make the problem convenient, and multiple choices that work for a given problem. Yet there’s another, perhaps simpler, strategy you can invoke for your sanity checks… “count” the number of digits you should expect the product of your multiplication (or quotient for any division) to have!

How can this be done? Consider the equation below:

321 x 29 = 9309.

The first number, called the multiplicand, has three digits; the second digit, or multiplier, has just two, and their answer – the product – has exactly four. In general, products that *do not require you to carry over additional digit* will contain a number of digits that’s just *one less* than that of all the digits being multiplied (the *sum* of the numbers of digits in the multiplicand and multiplier, minus one). What about problems that do require you to make a new column?

Try multiplying

999 x 99

and convince yourself that, even when the answer does “spill over,” you will only need to add a *single*additional digit – so when multiplying a number with *X* digits against a number containing *Y* digits, the answer itself have, at its most, *X*+*Y* digits. To be precise, this happens whenever the very first, “leading” digits of each number themselves have a product that is equal to or greater than 10, and occasionally when this product is between 5 and 9. You might impress your friends (and, hopefully, your interviewer!) if you find time to internalize this knowledge, since you can tell immediately that your answer will have the smaller possible number of digits, *X*+*Y-1*, if that product is 4 or less! Of course, something similar holds for division: dividing a number that has *X digits* by another number with *Y digits* yields an answer that has either *X*–*Y* (or *X*–*Y*+1) digits.

*Go-To #2: Common Decimal-to-Fraction Conversions*

Who knows how to express ½ as a decimal number? The answer, naturally, is 0.5.

Who knows how to express “5 out of 6” in decimal? You’re correct if you said “.8333…”!

Who knows how to express the ratio 1:7 as a decimal? Nice job – 0.143 is a fine answer! You’re doing quite well, so how about 2:7? At this point, the easiest thing to do might be simply be to “double” your previous answer. Consider that certain fractions are *just worth memorizing*.

The argument here should be self-evident: it’s not that certain fractions appear with an eerily large frequency in consulting (although, you might enjoy taking a look at Benford’s Law if you’re *that* suspicious), but that having sufficiently large repertoires of numerator-denominator pairs can help you break other problems into smaller, more efficiently processed ones.

I’m going to go out on a limb and say that you’ll shine (at the very least among nerdiest companions, if not an interviewing partner!) if you make a concerted effort to commit various fractions to memory, just like you would a multiplication table. I’ve listed some of the fun and useful ones below, but this doesn’t mean that these are the only (or, most) important ones!

- When the denominator is “5,” multiply by 2 and divide by 10 (or, you can reverse it!)
- 1/5 = 2/10 = 0.2 (literally, “two tenths”… etc.)
- 2/5 = 4/10 = 0.4
- …
- 8/5 = 16/10 = “1 & 6 tenths” = 1.6 (simplifies
*improper fraction*, no division!)

- When the denominator is “6,” multiply by 2 and divide by 10 (or, you can reverse it!)
**1/6 = 0.1666**…. (a repeating fraction, sometimes approximated as “0.1667”)- 2/6 = 0.3333…. = 1/3
- 3/6 = 0.5 = 1/2
- 4/6 = 0.6666…. = 2/3 (see 2/6)
- 5/6 = 0.8333…. (the only other one that’s new!)

- When the denominator is “7,” we see a new kind of “repeating number” pattern:
- 1/7 = 0.
**142857**142857**142857**…. - 2/7 = 0.2857
**142857**14285714…. - 3/7 = 0.42857
**142857**1428571…. - …
- 5/7 = 0.7
**142857**14285714285….

- 1/7 = 0.

To know where to start the repeating pattern, multiply numerator by “14”:

- 1*14 = 14
**14**2857…. - 2*14 = 28
**28**5714…. - 3*14 = 42
**42**857…. - 4*14 = 56
**57**142…. - … etc.

- When the denominator is “8,” you can pretend it’s “4” and cut in half… or…
**1/8 = 0.125**- 2/8 = 0.25 = ¼
- 3/8 = 0.375
- 4/8 = 0.5 = 1/2
- 5/8 = 0.625
- 6/8 = 0.75 = ¾
- 7/8 = 0.875

- When the denominator is “9,” decimal conversions are particularly easy:
**1/9 = 0.111**…. (dividing by 9 is similar to multiplying by 11 – already useful!)- 2/9 = 0.222…. (see the pattern?)
- 3/9 = 0.333…. (= 1/3)
- …
- 8/9 = 0.888…. (which is
*very*close to “0.89,” not repeating) - 9/9 = 1 (try “0.999”…. it’s indistinguishable from 1!)

- When the denominator is “11,” decimal conversions are also particularly easy:
**1/11 = 0.090909**…. (a two-digit, repeating pattern)- 2/11 = 0.181818…. (see a pattern forming already?)
- 3/11 = 0.272727….
- …
- 9/11= 0.818181….
- 10/11 = 0.909090…. (again, the relationships of 9 & 11 can be
*super*useful!)

- Other fraction-simplifications may be useful for ambitious candidates:
- 1/12 = 0.08333…. (nothing new, since this is 0.5/6, or 5/6 divided by 10)
- 1/13 = 0.
**076923**076923…. (another long, repeating decimal) - 1/14 = 0.07
**142857**142…. (also nothing new, if you think of 0.5/7) - 1/15 = 0.0666…. (here, think of 2/30 = 0.2/3 = “0.666…. divided by 10”)
- 1/16 = 0.0625 (similarly, think of .5/8)
- ….

To see more like these, visit http://www.asiteaboutnothing.net/p_speed-arithmetic.html#new.

## Specific Strategies

*Specific Strategy #1: Break multiplication into steps, rounding in the most convenient direction*

Consider the problem of multiplying the numbers 41 and 17. At first glance, this might seem “messy” … but rounding can help! Since “41” can be broken into “40 + 1, ” imagine that you just want to do the (slightly simpler) problem of 40 times 17. Really, this is only “ten times the quantity ‘4 times 17,’” and you can do the latter any way you’d like. Then, just add 17!

It looks like “rounding down” the multiplicand, 41, helped quite a bit. What about the other way – rounding up? Benjamin offers the following example:

53 x 89 = (53 x 90) – 53 = 4770 – 53 = 4,717

In this case, he rounded one of the numbers *up* and then *subtracted* the very amount by which this operation caused him to overestimate his answer. He calls this heuristic, by which he adaptively decides whether to round up or down when breaking down multiplication problems, finding the “math of least resistance.” One major alternative, of course, would be to break the multiplication problems into factors that you know how to memorize by heart. For me, finding the closest round factors works quite well – and once you’re practiced, can even be scaled-up.

Try doing it this way:

53 x 89 = (53 x 90) – 53 = (50 x 90) + 3 x 90 – 53 = (5 x 9) x 100 + (3 x 9) x 10 -53

*Specific Strategy #2: Break multiplication into steps, rounding in the most convenient direction*

Consider the problem of multiple the numbers 26 and 23: they are both close to 20, but rounding down the second (23 20) is probably the easiest way to find their product. If I said you that you *had* to round the first number, which way would you go?

While 26 is technically closer to 30, rounding *down *(26 20) still seems quite appealing to me. If we leave 23 as it is for the moment, this might look something like

26 x 23 = (20 x 23) + (6 x 23) = (2 x 23) x 10 + (6 x 23),

and the benefits would include operating with small numbers *and* only additions, which I find to be more intuitive than subtractions – which, admittedly I often do by adding the complement. If we had instead rounded *up*, we might have written

26 x 23 = (30 x 23) – (4 x 23) = (2 x 23) x 10 – (4 x 23),

and that’s still not too bad… but here we are, on paper, when the name of this game is *mental math*. There must be an even easier way to do this multiplication – right?

Try this. Recall that both numbers are near to the number 20; the first is 6 units away on the number line; the second is 3 units away. Now, let’s fantasize about pooling all those units of distance into a single number: 26 + 3 (or 23 + 6) = 29. What is the product of that so convenient, nearby number – 20 – and this new number, 29? We can easily verify that

20 x 29 = 580,

*exactly 6 x 3 = 18 * units away from the true answer of “26 x 23” = 598! Coincidence? I think not!

This approach to multiplication (dubbed the “close-together method” by Dr. B.) works because, if you take out the product of the digits in the “ones” column, 6 x 3 = 18, what’s left is the set of subproblems

(20 x 20) + (20 x 3) + (20 x 6) = (20 x 20) + (20 x 9) = 20 x (20 + 9) = 20 x 29,

and is precisely the first sub-product we computed. This method is powerful, and generalizable.

I’ll walk through another problem to reinforce how the close-together method operates. The name of the tactic refers to the fact that a multiplicand and its multiplier must be near each other. Let’s multiply 107 and 111.

__Step 1__. Figure out how far each number is from their common neighbor, 100.

This can easily be done in your head, at least for this particular case!

__Step 2__. Perform the addition 107 + 11 (or 111 + 7) to find arrive at the sum, 118.

__Step 3__. Multiply 7 x 11 (or 11 x 7) to find the product, 77, and remember it well!

__Step 4__. Multiply the “neighbor” number, 100 with the “ancillary” number, 118.

This should really be straightforward – separate into two subproblems.

In this case, it’s especially easy – one needs only to append two zeros!

100 x 118 = 11,800

__Step 5__. Recall the product of the “ones” digits, and add it to the running total:

11,800 + 77 = 11,877

Apart from the “rounding” that gave us our “nearby” number of 100, this was all exact! With practice, it’s also lightning-quick, compared to its pencil-and-paper equivalent. Yet again, in case interviews, rounding is often the name of the game. How can we make that fit back in?

## How Much Rounding Can I Get Away With in My Mental Math?

If we had rounded the numbers in the above problem *both* to the nearest hundred, that is, we had solved instead the problem 100 x 100, we’d have been about 15.8% below the actual answer. If we rounded the first number *down* to 100, and the second *up* to 120, we’d have got to within 1% of the answer – amazing! But could we have known this would happen in advance, before invoking the close-together method? How can you predict the effect of liberal rounding?

To answer this curiosity (and, finally, lay to rest the woes of your most skeptical friends about the dangers of severe rounding errors), we’ll switch from an arithmetical approach to an algebra-based formalism. The idea will be to figure out how far we’re off, exactly, even if we do not know in advance the amount of the rounding we will do; we will write out the equation that describes a multiplication in which an *arbitrary* degree of rounding is applied to each number.

*Algebra Interlude [Optional]*

Let the first number, the multiplicand, be represented by *A*; the variable for the second number will be *B*. Rounding either number means adding or subtracting a small amount, which can itself be expressed as a fraction of the number being rounded. In mathematical “language,” we can call the “rounded” version of the first number “A + (a x A),” where the lower-case letter a is the ratio of *the number of units by which A has been increased or reduced to the value of A itself*. For now, say “a” is a positive number – A is being rounded up. In other words, if you add a certain number of units to “Big A,” then divide by the value of A, you get “little a.” Rounding down works the same, but in that case (since we are subtracting from A) we assign a negative sign to the front of “little a” – this works out identically, though, because subtraction is really the same thing as *adding* a negative number. We’ll do all of this for B as well, so we’d write:

A x B [ A + (a x A) ] x [ B + (b x B) ],

that is, the right-hand side is the new multiplication problem that forms for rounded numbers.

What happens when we rewrite the multiplication using the laws of algebra, such as the Distributive Law? We see that the final answer to the *new *multiplication is given by

(A x B) x [ 1 + a + b + (a x b) ].

Here, “little a” and “little b” are both fractions, and we know that multiplying a fractions against any number gives a smaller number. This means that multiplying two fractions gives an *even smaller* fraction; since these fractions describe a “rounding” process, and rounding usually means adjusting by only a small number of units anyway, we can justifiably assume that (a x b) term doesn’t affect our answer enough to be included in deciding about “too much” rounding. Thus, we can assume that, by rounding, we changed our multiplication problem like this:

A x B (A x B) x [ 1 + (a + b) ].

Let’s do one more simplification step. Distributing again, we see that our new answer is

(A x B) + (A x B) x (a + b).

Said another way (i.e., English), the “rounded” answer is approximately (A x B) x (a + b) “more” than the original answer. Since what we really we want to know is not *how much* the rounded will deviate by – a million dollars might matter if you only make 2 million dollars per year, but will be entirely negligible if you’re looking at a 50-billion dollar proposition – but rather the % error in the product of A and B after all the rounding and multiplication has been carried out, we will make another fraction. We divide that “error” amount (A x B) x (a + b) by the actual value of the product (A x B) to find the important quantity, which turns out to be (a + b).

*Mental Math Sanity Checks: Sizing a Rounding Error*

Whether or not you read the last section, the upshot of the algebra is that it’s actually relatively simple to calculate the “percent error” you can expect in the product of two rounded numbers (as compared with the value of the “exact” product). The way we can do this is define two numbers – “little a” and “little b” – as the fractional number of units, respectively, by which the multiplicand and multiplier are rounded. So, if I round the first number in the multiplication *up *by 5 units, I divide this 5 by the original value of the first number to get “little a.” Both “little a” and “little b” should also keep track of the “sign,” or direction of rounding – so, if I round the second number in the multiplication *down* by 7 units, I divide -7 by that number to get “little b.”

Now, *either* “little a,” or “little b” (or *both*) can be positive or negative. So let’s “plug in” some values and see if this approach works. Before, we multiplied 26 x 23 to get 598. If instead we had done 25 x 25 = 625, we would have landed about 4.5% too high. Here, “little a” is equal to -1/26, and “little b” would have a value of 2/23; adding them together, we see (a + b) = 4.8%.

Not too shabby! Let’s check this on our other close-together multiplication, 107 x 111. If we’d treated this as 100 x 110, our error would be about 1.8% – we could have anticipated this by computing (a + b) = (-7/107 – 1/111) = (-0.065 + -0.009) = 7.44%! What if we’d gone instead with 105 x 105? Then we’d have had (a + b) = (-2/107 – 6/111), still only about 7% error. Note that, in last two these examples, we didn’t even try to round in opposite directions!

*Mental Math Implications for Case Interviews*

Even without conducting these numerical tests, you can see immediately that it *pays* to round in opposite directions, especially when you’re rounding by comparable amounts: (a + b) is closest zero when “little a” and “little b” nearly cancel one another. Of course, all of this logic also applied to division: the important quantity for a quotient is 1 – (1 + a) / (1 + b), which is the percent error for A divided by B, when both A and B are rounded as described above.

Maybe even more importantly, it is also now quite simple to picture how much rounding is “too much.” If one’s interviewer is comfortable with a 20% error in an approximate product, a rule of thumb could be to make sure that one does not allow (a + b) to exceed 20. You might be rounding each number up by 10%, or one up by 30% and the other by down by 10%, but that is your threshold. It’s actually not so strict when you look at it like that; and not having to *guess* in advance what your overall error will be for a pairwise product helps to eliminate some anxiety.

Check yourself on this! Watch this video on YouTube – but first, try to think about how you would move to approximate the answer to 819,593 x 7,669 (*Mbbmath*’s solution begins at the 55-second mark). Then, figure out the value of “little a” by counting how far from “819,593” you’ll get on the number line when rounding that multiplicand, and dividing (yes, roughly!) that figure itself by 819,593. Do the same for 7,669 to find the ratio we’ve been calling “little b,” and finally perform the sum (a + b). At the 1:54 timestamp, he flashes slightly different values than I would have calculated with my algebraic formalism; I get (-0.024 + 0.043) = 1.9% – cozily within the 20% range, and exceedingly close to the rigorously hand-calculated error value of 1.8%!

## BONUS Advice: Why can Mental Math seem so Difficult within the Context of a Case?

Mental math “hacks” are great, but can you pull off such wizardry during a case?

One might imagine three reasons that performing feats of mental math becomes more intimidating than it otherwise might be, in the specific setting of consulting case interviews:

- The candidate is ill-prepared (simply not well-practiced enough at negotiating numbers)
- Nerves impair clear thinking (including inhibitions, like “tip-of-the-tongue” phenomena)
- The
*context*is too unfamiliar

While the first two might seem more obvious, I believe the third requires special attention. I would like to spend the remainder of this article not on generic advice, or even further methods that I’ve enjoyed learning from FirmsConsulting’s training programs, but an honest-to-goodness unpacking of that last point – truly, context is key.

*Mental Math With (Almost) No Numbers*

Let’s switch, for just a moment, to a different kind of “mental math” question – one that tests logic, rather than focusing purely on numeracy skills.

Pretend you’re sitting at a table with friends out at a restaurant or social club. On the table there are four ordinary playing cards, with one minor modification: on the reverse side of each card has been printed a letter of the alphabet. You are told: “if there is a vowel on one side of the card, then there is an even number on the other side.” Glancing briefly down, you note the following arrangement.

A P 2 3

The youngest (and most precocious) of your friends asks you which cards would need to be flipped in order to discover whether the “even number-consonant” rule had been violated. Did you figure it out yet? She claims that you need to turn over two of the cards.

The answer, of course, is that you must check the numerical value of the card imprinted with “A,” as well as the letter that’s the printed on the back of #2. As straightforward as it may seem once you’ve arrived at the solution, it’s admittedly unfair of me, the author, to ask such a question without a warm-up. If you subscribe to “better late than never,” try an easier version.

*Social Problems are Simpler*

While you’re sitting at the same table, your server changes shift, and a new server comes to introduce himself (and take another round of drink orders). He has not a clue who was drinking what earlier; for all he knows, you and your friends have swapped drinks, or have been sharing. This is of concern for him, since your mathematically-inclined friend is clearly underage, and he cannot tell from afar what she is drinking. From his eyes, your table looks something like this:

Underaged College-aged Drinking Coke Drinking beer

Who does he have to check on? Clearly, he needs to make sure your underaged friend isn’t drinking something alcoholic, and by rights he should also ask for your beer-drinking friend’s ID (you’re probably the one drinking soda because you have a case interview tomorrow, for which I sincerely hope this last-minute read comes in handy). Not so hard, is it, this problem? Yet, you might have noticed that the *structure* of the question is identical to that of the 4-cards problem. The only thing that’s different is the *context*, and somehow this matters tremendously. Humans are often much happier solving problems *that look like what’s already in their wheelhouse*.

So, while a person might (or might not!) be comfortable calculating the tip at a restaurant, or “guess-timating” how much you’ll pay at the gas tank for filling up your mechanical steed, it’s not guaranteed that you’ll have similar speed, precision, or overall comfort – even on problems with identical structure – in the middle of a case, even if you’re fluent with numbers and nerves don’t significantly affect your performance.

*Implications for Mental Math Practice*

What’s the takeaway? You need to *practice the way you’ll play* – once you’ve gotten down the basics, this means training *in-context*, on business-specific mental math problems. I’ll finish this “round” (okay, pun intended this time) of discussion by naming a few “themed” examples.

Practicing Percentages

First, you must be an expert at converting percentages to “raw” values. What I mean by this is that, given one number and a percentage figure, you must be able to rapidly, and accurately, perform a mental multiplication problem to tell me what the value of interest is. This is needed in the reverse directions as well – converting ratios or proportions to percentages, or solving for some original amount given a share and its percentage – but problem of breaking down figures into smaller components appears often, in many business contexts. Furthermore, it is easier to compare the relative sizes of business segments or effects using numbers than percentages!

Some examples of “use cases” are:

- Computing the profit
*margin*for a given product line in a company - Breaking a market into segments based on the share of individual players
- Figuring out required upfront costs to attain a certain return in investment (ROI)
- Segmenting costs into categories like “labor, equipment, land, and inventory,” etc.
- Calculating “utilization,” as when assessing time available to actually treat patients
- Estimate the number of infections, to see if it exceeds 5% of all people in a country

Finance Applications

Another example from Benjamin goes like this. Your bank offers you a certain % savings rate for your new account: how long will it take your money to double? It is not terribly difficult to derive an *approximate*algorithm that works, and that algorithm is known as the “Rule of 70.”

There are harder (and slower) ways to compute the time it takes a principal investment to grow, but usually these involve inverting an exponential expression, using logarithms. I don’t know about you, but I can’t fathom trying to round and multiple against “logs” in my head, so I would just think:

70 divided by ( interest rate) = # of years to double.

So, a 3% interest rate should take about 23.33 years (it’s actually about 23.45 – off by a month and a half, only!) and a 7% rate should take about 10 years (10.25 – also about 3 months off) according the Rule of 70. Of course, if the interest rate is given not annually but monthly, it still works – but the answers would be 23 and 10 *months* (off by 3 days or 1 week, respectively). Get this: the same idea also applies if you want the money to *triple* – that’s just the Rule of *110*!

What about a harder problem? Benjamin offers us the following *amortization* example^{*}. Pretend you borrow money to buy a house – say, $200,000 – and the bank charges 6% interest annually, “compounded monthly.” In plain language, this means that they plan to charge you

(6% divided by 12 months/year) = 0.5% interest,

each month, on whatever amount of money you haven’t yet paid back. If the bank gives you 30 years to pay off the loan (and you plan to use all the time, rather than trying to pay it off faster), can you figure out the amount of money you’ll need to pay the bank each month?

First of all, you know you’ll need to make (30 years x 12 months/year) = 360 payments … this you can (hopefully) calculate exactly, without too much trouble! If there were no interest, it would simply be a matter of paying

$200,000 divided by 360 “monthly” payments in all,

which would amount to (20,000 divided by 36) after knocking off the extra zeros; best of all, we can also divide by the dividend and divisor by 4 to reframe the problem as (5,000 divided by 9), and then divide both again, by 5, to reveal a much cleaner-looking problem, (1000 divided 5/9). If you’ve done your due diligence to memorize the pattern for fractions containing the number “9” as the denominator, you’ll know that 5/9 = 0.555…. moving the decimal says this is $555.

What about the *interest*? In the very first month, it will be ($200,000 x 0.005) = $1000, and it will never be higher since the multiplicand above ($200,000) goes down every time you make a payment, while the rate stays the same! Thus, as a ballpark estimate, you might say a monthly payment for this case will amount to $1,555 (or less). If we had *rounded *the original division problem to (20,000 divided by 40) for expedience, we’d have estimated it at $1500!

## VERDICT: Mental Math Practice Should Begin with Rounding

No matter what you do to rehearse your mental math skills, I’d suggest beginning with a firm grounding (I know, enough with the puns) in establishing your rounding capabilities. Arthur Benjamin himself performs addition and subtraction from left to right (instead of the commonly taught right-to-left procedure, which is predominantly a pencil-and-paper algorithm) in order to preserve information about the *largest orders of magnitude *in a problem above all else; indeed, this is the very *essence* of what rounding does for you, and that’s how it allows you to prioritize.

Feel free to take the most common mental math “advice” I hear in consulting interview-prep communities: avoid ever hesitating, or grabbing your phone-calculator, to solve arithmetic problems that come your way. “They say” that you should be actively practicing with every bill, receipt, and baking recipe you encounter – and while there’s nothing *wrong* with this, I do see two counterpoints as particularly persuasive. First, I know a lot of MBB consultants – juniors as well as a few partners – and somehow, all of them seem to have managed to ensure that their inner human calculators don’t dominate their livelihoods… while none of them have lost basic numeracy abilities. Second, without the right *context* your preparation will never be as robust.

*Let us know in the comments what you would like to see next – more examples of mental math breakdowns? More mental math strategies, from FirmsConsulting and beyond? Or more about learning theory, and how to train your skills in the environment, social, and psychology context that will help you transfer your latent abilities into case interview success? Until next time!*

*This article was contributed by Joe, a valued member of the FIRMSconsulting community.*