Approach for solving Guesstimates

Click here to read the previous part of the article

Think about the question being asked, and ask clarifying questions

When you begin with the guesstimation exercise, you must be very clear about what exactly the interviewer is trying to ask you. It is important that you do not assume anything initially.

A lot of times, interviewers will deliberately leave out important pieces of information while framing their question to see whether or not the candidate smart enough to ask for that information.

So be a little pompous about your intelligence!

For instance, at one of my interviews, I was asked to estimate the number of people commuting by the metro in an hour. The questions that immediately sprung to my mind on being asked this question were:

(a) Is this a peak-hour that we’re talking about, or a non-peak hour
(b) Is this a weekday that we’re looking at, or a weekend, and
(c) Is this a typical day, or some special occasion like a festival, etc.

Without having the answers to these questions, coming up with a reasonable approach would’ve been difficult.

Break it down

In your mind, imagine a tree. That’s exactly what your approach to a guesstimation question should look like.

It should have one main trunk, with many branches, and sub-branches, and so on. Your task, then, is to identify these different ‘branches’.

 Let’s say the interviewer asks you to estimate the number of shampoo bottles sold by XYZ brand in India last month.

1. Identify the trunk 
Once you’re sure that you clearly understand the question, the next thing to do is to identify the main ‘trunk’ in this case. To do this, you need to identify all the people who could possibly be consumers of shampoo bottles. A good proxy for this would be the population of the country. So you’ve found your trunk!  In fact, in 90% of guesstimation questions, population is the main trunk.

Now you can start solving the question by assuming that the population is a random number like 100 or 1000, and then making further calculations accordingly. Using such numbers makes the calculations easier… you can extrapolate your findings on the actual population value later.

2. Identify the Branches and Sub-branches
Try to dis-aggregate as much as possible by adding more and more variables to your answerRemember… The interviewer is also trying to test you on the basis of your ability to identify variables.
In our example, for instance, to start with, you might want to classify the population into 3 sections -- rich, middle-class, and poor. Let’s suppose that you know beforehand that brand XYZ is quite popular, and of average price, and therefore attracts rich and middle-class consumers. Poor people generally use sachets ... let’s say you decide to assume that.

Make sure that you speak out your assumptions loudly, so that the interviewer also knows what assumptions you are making…don’t just mumble or assume things in your head.
Also, keep a pen and paper handy, and keep jotting down your assumptions.

First branch identified -- Income Group
So you now have two branches—one who can afford shampoo bottles, and the other who can’t.

You may take a rough cut-off limit of poverty for estimating the % of poor people in the population ... Remember, the  interviewer may be testing you on your general knowledge, but if you know the concept of poverty line and other economics fundas ,no harm in showing off a little!

Next, you may want to partition the "People who can afford" branch on the basis of whether those people stay in rural areas, or in urban areas.

Sub-branch identified -- Geography—Rural/Urban

Again, you can assume some distribution statistics or if you have an idea about the rural-urban distribution in the country, you can use that.

There could be a multitude of other factors like the gender distribution, age distribution, etc. that could be used to branch this problem out further, and to arrive at the most disaggregated level, beyond which one cannot think of more ways of breaking down the data.

3. Go the extra mile

Now, it is time to make a Big Bang ... while you have used all the normal weapons till now, it is now time to bring your "Brahmastra" into the picture!

Try to add a little humour to the equation.

You can say that the age-wise baldness index in males could also be considered to narrow down the population J

I’m sure you get the idea by now. What is required is to logically disaggregate the problem as much as possible. Then, using some basic statistics related to the city/country, common sense and prior business knowledge (if possible), make the calculations.

Now once you are done branching the problem out, calculate the shampoo usage for each sub-branch or branch (can call it a node too), and then add these numbers to arrive at the final aggregated value.

4. What numbers should be used for breaking down the data?

We know that shampoo consumption would be higher amongst females than amongst males, so all we need to do is, make a reasonable assumption about the female/male ratio in the country to disaggregate the data.

Also, age- wise usage would be a factor; teenagers are likely to be more conscious about their looks than, say, the old people. So we would need to bucket the population on the basis of the age-distribution, and arrive at the usage pattern accordingly.

Remember…even if you feel that your assumptions are not very accurate, there’s no need to worry. In real life projects, you don't assume, you get actual data to work on. 

If you are interested, here are a few facts and figures:

Some of the statistics that are commonly used while solving guesstimates, and you might want to browse through before going in for an interview are given below:
India statistics
Population:  1.29 bn (2015 updated)
GDP: $2.3 tn (2015 updated)
Age distribution (2014 updated): 
0-14 years:  28.5%
15-24 years: 18.1%
25-64 years: 47.6%
65 years and over: 5.7%
Gender distribution (2014 updated): 108 males per 100 females
Rural-urban distribution: 
Urban: 31.3%
Rural: 68.7%
% of population below poverty line: 
As per World Bank figures: 23.6% (2011 updated)
As per national definition of poverty: 21.9% (2011 updated)

There are many other statistics that are frequently used in solving guesstimates, and depending upon your ability to remember numbers, you could learn them as well. Besides, if you’ve been residing at a particular place for a long time, or if you belong to some place other than where you’re currently staying, you might be asked guesstimates related to that place. Again, depending on your ability to remember numbers, you could learn/not learn them.

In addition to the approach described above, there are a few more things related to guesstimates that are very important from the interview point of view and you should, therefore, keep these in mind:

• When making calculations, round off figures where possible to make calculations simpler. For instance, if the % of people living in urban areas is 31.3%, and you round it off to 30%, it will make the calculations much simpler for you.

• At every calculation step, mentally check if the figure you’ve arrived at seems reasonable. Say you were estimating the number of metro commuters in an hour, and you got a figure the same size as the population of Delhi. You know in your head that something is wrong with this figure, and should revise it by figuring out what factors you might have missed out in your calculations.

• There might arise situations where you have absolutely no idea about some statistic that you need to use for solving your question. In that case, think of the best proxy for the factor at hand. As long as you can make reasonable assumptions while breaking down the problem, and can make the interviewer aware of the caveat (that the true estimate would depend upon the actual value of the factor whose value you have no idea about, and have therefore assumed based on some logic), it should be okay.

• There can be multiple approaches to arriving at the final answer, and it is for you to decide which one you find the most logical.

• Most importantly, it is never possible to get an absolutely accurate figure for the parameter you’re estimating unless, of course, you somehow know the value beforehand, because calculations will *always* involve assumptions, rounding-off, etc. The interviewer too is not looking for an accurate figure, but is only checking your approach. So when solving a guesstimate, it is important to make logical assumptions based on sound reasoning, but at the same time, keep calculations simple.

I believe that by following the steps described in the article, one can grasp how to approach guesstimation questions.

In my next article, I’ll be dealing with some specific guesstimation questions to give you guys more clarity on how guesstimates really work, so keep reading!

Enjoy reading our other articles and stay tuned with ....

Kindly do provide your feedback in the 'Comments' Section and share as much as possible.