Estimation: Tables to the Rescue
One of my biggest fears and nightmares in elementary school was solving word problems. So you can just imagine the fear and dread the Google estimation questions conjured up in me. There were flashbacks to childhood nightmares.
Then one day I met an amazing PM in the mock interview community named Leili Rasouli. Leili demystified the estimation question and taught me a crucial trick, the use of tables to keep me come and provide clarity to my answers. I have tweaked the methodology a bit as I coach others but the core of the approach is 100% Leili’s.
Logic Question
Most people, like me, when I first tackled this question type, forget this is a Fermi question. Which means it is more about logic than math. Your interviewer might make you feel awkward about the wrong number, but the hiring committee cares more about the logic approach than the math.
The Key Rules
Ask clarifying questions before diving in.
Come up with two high-level approaches before presenting a solution.
Pick one approach to show the interviewer. (Hide the other for later.)
Have NO MORE than 3 key variables
A 4th variable is allowed occasionally
Determine your equation before you start presenting your approach
Put your equation in the first row of a table
Present all your work in the table
Delivery Notes
Pause to gather your thoughts after asking clarifying questions
Don’t start talking until you know your key variables and base equation
Be ready to break down one of your key variables into High, Medium, and Low groupings to sense-check your final answer.
Don’t just verbally deliver your equation (unless it is a phone screen)
Don’t write your equation in a line of text
Practice creating a quick grid in Google Docs until you can do it in your sleep
The Secret Sauce: The Table/Grid
Show your work in a table/grid for clarity of thought
Unless you are a math genius, following someone’s equation without seeing it is difficult
There is a reason math classes still use blackboards
For Example
Prompt: Estimate the size of the market for women’s rainboots sold in Seattle per year.
Step 1: Clarify.
By women, you mean adult women, not including female children.
By rainboots, is it okay with you if I presume footwear specifically for rain, not all waterproof boots?
By market, I assume you want the dollar amount sold each year. Is that correct? Or did just want the number of boots sold?
Step 2: Think of two ways of approaching this problem.
Option 1: Percent of female adult population * # of boots per year * avg $ for boot
Option 2: Assume $ spent on shoes each year in Seattle *% rainboots * % women
This would require an assumption of # of shoes each person buys per year
Step 3: Tell the interviewer you plan to demonstrate the high-level equation first, then sense-check yourself by breaking down one or two of the variables. Make it clear you want to quickly show your logic.
Step 4: Present three key variables to the interviewer.
# of pop
# per boots
$ per boot
Step 5: Present basic equation
# of pop * # per boots *$ per boot
Step 6: Explain your logic for number you want to use for each vairable in your high-level equation
i.e. You believe the population of Seattle is roughly 700K to 1M. You will use 1M as a quick estimate unless the interviewer objects.
i.e. Females make up 50% of the population but 10-15% of those would be considered female children
So that gets us 1M*.40
But maybe only 25% of population would buy a rainboot in a given year
So we have about 100,000 women eligible.
Then we will say those who buy will buy once per year
Then we say roughly $25 per boot (maybe this is under estimating it might be closer to $35 but for this first equation, we will keep it simple assuming many will go for super cheap options).
This leaves us with
Pop: 100K
# boots per year: 1
$ per boot: $25
Step 7: Present the equation in grid with variables below each label.
Build basic table labeling the table with the variables.
Show the basic equation in row 2, coming up with a high-level number quickly
Step 8: Tell the interviewer next you would like to sense-check yourself.
Step 9: Establish your low, medium and high groupings
i.e. Frequency of purchase, Avg Price per pair of Boots, or both.
Define Low, medium and high definitions
Explain logic for adjustments made by each group.
i.e. More expensive boots wouldn’t be replaced as often
i.e. Fewer people would invest in high-priced boots, but those that did wouldn’t buy boots as often.
etc.
Step 10: Present basic setup
Step 11: If they want you to keep going, do the math
Step 12: Verbally compare your overall estimate to your adjusted assumptions
Step 13: Be ready to be asked for another approach. If asked, present that total shoes sold approach that you keep hidden in your notes, off camera.
Other Tips & Tricks
Build a cheat sheet of commonly used facts (Think: population statistics, size of common file types, etc.)
Get comfortable with scientific notation (Think: 10^9)
Keep it Simple: Don’t try to overcomplicate things.
Remember: It is more about logic than math.
Google Estimation Analytics Prompts
