Guesstimates

Introduction

Guesstimates are a good way to learn a large portion of the consulting toolkit. Best of all, when you get started with guesstimates, you can cut out a lot of the communication aspects, and focus on structuring and problem solving. That's why it's important to keep practicing your communication and structuring alongside this. Because in a little while you'll have to bring all of it together, as a part of case solving.
In a guesstimate, the interviewer will ask you to "guesstimate" a certain numeric value. It's a guesstimate, because you will not have enough information to arrive at the "correct" answer, and so you'll have to make a few clever guesses. But it is largely a structured estimate, where you break the problem down to the point where the guess work has minimal impact on the final number, and your guesses are sharp and close to accurate. So let's break it down and have a look at what the elements of a guesstimate are:

Clarifying Questions

As usual, when we receive the question, we need to resist the urge to get into answering it. The first thing that we need to do is figure out what the question means. In order to this across any case (or guesstimate) there are a few standard questions that I find very useful. Some of these are not as relevant or useful for most guesstimates, but it's good to get acquainted with the full set of questions from now on:

- Objective -

Always make sure that you are clear on what the requirement is, by paraphrasing/synthesising and asking the interviewer if you have understood the ask correctly. Here you might ask whether you have to arrive at a number, just outline an approach, whether they would be sharing data (no longer a guesstimate) or if you have to make assumptions etc. You would also definitely want to clarify beyond these details whether the meat of the question has been correctly captured.

- Mentioned terms -

Guesstimates are usually simple one sentence questions. Even when they're not, by the time you're done clarifying the objective you should have reduced it to one sentence. In this sentence you now need to pick up each term which is either unknown to you, or ambiguous in this context, and clarify what is meant by the term. Keep in mind that a lot of terms that you understand well will still be ambiguous in this context, and should be clarified. For example, in a certain question you might want to clarify whether morning means midnight to noon, or sunrise to noon.

- Working model -

This is usually less relevant in guesstimates. But you might need to better understand a business model or working model before you can make reasonable assumptions and structures. For example, if you have a guesstimate on the number of people streaming on a certain social media platform right now, you might not have used that social media platform extensively. You might want to clarify how the platform works, so that you don't make unfair assumptions.

- Industry/space -

This too is not often required for guesstimates. But much like the context of a working model, it might be relevant to understand what competition exists, or how social media works overall, to make fair assumptions in the example earlier.

- Other question words -

It's always useful to think through the 7 question words (what, when, where, who, why, which, how). Once your objective is clear, the most relevant ones to ask are usually when, where and who as filters to clarify what assumptions will be fair and simplifying for the guesstimate.

Structuring

Once we're clear on what we're guesstimating, we need to put in place a structure to guesstimate the same. Given that we're arriving at a number, it's easy enough to see that we have to use the numeric formula method of structuring. While we know that we need to structure the problem in the format of a formula, there is still a lot of scope for playing around with the structure selected. Two of the levers available are:

- Type of formula use -

For example the number of trees in India can be calculated as the sum of trees in urban areas, rural areas and areas in between. Or it can be calculated as the average tree density across the country multiplied by the area of the country.

- Granularity/grouping of formula -

For example the number of trees in India can be calculated as the sum of trees in urban areas, rural areas and areas in between. Or it can be calculated as the sum of trees in metros, tier 2 towns, tier 3 towns, villages and areas in between.

A few different ideas for formula type and granularity might come to mind when you start thinking about the problem. In order to select the best suited method, you need to quickly evaluate them on 3 criteria:

- Determining factor -

We could sum up the number of trees across each state or region, instead of urban vs rural. But we look at which of these formulae captures the factor which determines the number of trees more accurately. The number of trees might be determined largely by the population density and the kind of activity in the area. Therefore urban vs rural vs areas in between best captures the determining factor behind these numbers.

- Simplicity -

Rather than selecting a super accurate, complex formula, you should use one which is simple, can be calculated relatively quickly and can easily be explained to someone. In some cases, if your formula is large and chunky, you can club multiple terms together, to keep it simple, and break it down further subsequently, if needed.

- Benchmarks -

Very importantly you want the terms of the formula to be easy to directly benchmark and compare to numbers and facts which are common knowledge (or at least you know). If you cannot (often the case) directly benchmark the first level of the structure, you should be able to structure each term in the first level to give second level terms which are easy to benchmark. In guesstimates, the goal is often to keep things simple, and therefore not have to go beyond 2 levels. This is not a fixed rule of course, and you would get good structuring practise by going deeper at times.
The formula/structure which captures the determining factor, is simple to write out, and doesn't necessarily go too many levels deep before you have reasonable proxies to arrive at an estimate, is the best one to use. Aside from the best structure to use, if you have another structure which uses a different type of formula (not just different grouping) and is also fairly good to use, set that formula aside for your sanity check.
Keep in mind that you need not figure out what the whole structure will look like across all levels right at the beginning. You will soon build an instinct that allows you to judge just from the first level of a few potential structures, which one you want to use. With the first level in hand you will then begin to navigate through the structure in a depth-first manner, detailing it out, till each branch has reached a level where you can guess the values of the nodes. Navigating the structure should become more clear through an example.

Guessing

For each final node, you'll need to take a smart guess as to what the value could be. Guessing follows the same basic principles of being quick and simple, but also accurate. For that we base our guesses on proxies or benchmarks. A good proxy is:

- Factual -

Based on a well known/accepted fact.

- Comparable -

Similar to the situation at hand, and therefore a good benchmark.

- Creative -

Because a guesstimate is meant to be a little difficult, there will often be a bunch of nodes where you don't have ready at hand information which is easily comparable. Therefore a little creativity comes into play in finding parallels that can be drawn and cleverly using the limited facts that one would know to estimate all sorts of unknown values.
In guessing there will be assumptions that you will make. The important step is to justify your assumption either with facts or with reasonable intuitive thinking. After each guess you should be aware, and call out whether that guess is likely to be accurate enough, or whether it is something that you might want to revisit after a sanity check.
It's important to note that round guesses will prove more useful than precise guesses which are awkward to do mental math with.

Calculate

Your structure should look like a mathematical tree. Each node is represented by a formula involving each of it's child nodes. You should therefore be able to work your way up the structure from the guessed end node values. Ideally you should brush up on your mental math or doing quick arithmetic on paper. Figure out which numbers are quick and easy to run your basic arithmetic operations with. Then ensure that in the guessing phase you are appropriately rounding numbers in that manner. While calculating you may have the option of rounding up or down. Try to balance out the number of times and the degree to which you round in each direction. This helps maintain accuracy without slowing down the process of calculation.
Importantly, keep writing down the arrived at intermediately calculated values alongside your structure. When you revisit some assumptions and adjust your calculations, this will help save time.

Sanity Check

Now that you have arrived at an overall value, you might or might not be able to judge whether it is fair, or whether some of your assumptions need a little adjustment. If time permits, you would therefore want to do a quick (much faster than the original guesstimate) sanity check on the number. In a sanity check, you pick up your second choice formula (only if the type is different, and not just the grouping). Once again, perform a guesstimate using this, but simplify the structuring process significantly, and reuse any assumptions that you're confident of. This way you should quickly be able to arrive at another value which is either in the same ballpark, or significantly higher or lower. From here you take a call on whether you want to revise some assumptions up or down, and by how much. This remain an intuitive exercise, not one where you try to exactly match the 2 numbers.

Example

A sample guesstimate has been worked out in this ppt file.

ppt file.

Alternately this is a video version of the same guesstimate.

video