Building numbers you can defend

In a PM estimation interview, you won't know how much Americans pay for pet insurance, and nobody expects you to. What you can do is show where every number came from, so that when someone questions one, you can walk them back to something you do know.

Take the prompt: "How much do US pet owners pay for pet insurance each year?" A premium is what a customer pays the insurer for coverage, so the equation is:

Yearly premiums = dogs and cats × share insured × yearly premium per insured pet

Build each number from a fact you trust

Population is a safe place to begin: about 340 million people (US Census Bureau, 2024). From there, using my own assumptions:

• About 2.5 people per home. 340 ÷ 2.5 = 136, which I round to 130 million homes.

• 40% of homes have a dog, averaging 1.5 dogs. 130 million × 40% = 52 million, call it 50 million. 50 million × 1.5 = 75 million dogs.

• 30% of homes have a cat, averaging 2 cats. 130 million × 30% = 39 million, call it 40 million. 40 million × 2 = 80 million cats.

That adds up to 155 million, and I'll use 150 million.

A recalled number has one defense: "I read it somewhere." A derived number has a chain of steps, so if someone objects to 40%, you change that link and the rest stands.

Cracking the PM Interview recommends having a few round figures ready before you walk in, and this chain shows why: all of it starts from one number I could say without thinking.

Keep your guesses where you can see them

A guess hardens fast. By the time 150 million reaches the last line of the math, it sounds like data. So I put a small g beside every number I choose: 2.5, 40%, 1.5, 30% and 2 in the chain above, then 4% and $700 as they come. The 340 million gets no mark. When I later decide which inputs to test, every g is a candidate, with the pet count standing in for the five behind it. Give the reason as you write each one, because a listener can accept a number they'd have set differently if they can follow why you chose it.

A range belongs beside each guess too: "I'd be surprised if the real figure fell outside this." Keep it narrow where you have reasons and wide where you are guessing. For a true guess, think in multiples, such as half to double.

Come at a hard guess from several sides

The share insured has no clean chain behind it, so try several rough angles and see whether they land near each other. Your own daily life is fair evidence, with one caution: the people around you are not the country.

Among the pets of people I know, say 1 in 10 has a policy. My friends probably earn more than average, so 10% is a ceiling. I've also heard of enough policies that it can't be as rare as 1 pet in 100, or 1%, which gives a floor.

Now a small model. All of these inputs are my assumptions. Pets mostly get insured when young and ages are spread evenly, so one year's new pets are the total divided by the lifespan:

• Dogs live about 12 years. 75 million ÷ 12 ≈ 6 million new dogs a year.

• 15% of them get a policy. 6 million × 15% = 0.9 million new policies a year.

• A policy lasts about 5 years. 0.9 million × 5 = 4.5 million insured dogs.

• One insured cat for every four insured dogs. 4.5 million ÷ 4 ≈ 1 million insured cats.

That gives about 5.5 million insured pets, and 5.5 million ÷ 150 million ≈ 3.7%. My 4% stays a guess that these angles make plausible, because "15% get a policy" is largely the same unknown in a new form. The model doesn't earn a narrower range, so half to double my 4%, or 2% to 8%, stands, inside the floor and ceiling.

Make the arithmetic easy to say

Use numbers you can multiply in your head, since your guesses are too rough for exact digits to help. Rounding some up and others down lets errors partly cancel, but check which way you leaned. Three of the four rounds in my pet count went down. Unrounded, the chain gives 136 million × 40% × 1.5 = 81.6 million dogs and 136 million × 30% × 2 = 81.6 million cats, or 163.2 million in all. Using 150 million gave up about 8% (150 ÷ 163.2 ≈ 0.92), a fair price as long as you know the estimate leans low.

For large numbers, separate the small part from the zeros, multiply the small parts, then add up the zeros.

• 150 million × 4% = 6 million insured pets.

• 6,000,000 × 700: 6 × 7 = 42. Six zeros plus two zeros is eight zeros. 42 followed by eight zeros is $4,200,000,000, or $4.2 billion.

Say each line as you write it, in an order someone else could follow, with the unit beside every number: 6 million pets, $700 per pet per year. Leave "a month" off the $60 dog plan, and a few lines later it can pass for a yearly price: 6 million × $60 = $360 million, a twelfth of what 6 million × $720 gives.

For growth questions, learn the Rule of 72: something growing at a steady x% a year doubles in about 72 ÷ x years. If an interviewer says the market grows 20% a year, 72 ÷ 20 = 3.6 years, so $4.2 billion reaches $8.4 billion in a little under four years. As a check, 1.2 × 1.2 × 1.2 × 1.2 ≈ 2.07. The rule runs short as growth climbs: at 100% a year it says 0.72 years, when the true answer is 1. If 72 divides badly, swap in a nearby number that divides cleanly: at 7% a year, 70 ÷ 7 = 10 years.

Ask or assume

Sort each missing number by where it lives. Suppose the prompt were "should our pet insurer add a cheaper plan?"

• Only the company has it. Our current policy count sits in a dashboard. A PM in that seat would simply look it up, so ask, with your fallback attached: "I'd like our policy count. If you'd rather not share it, I'll assume half a million and flag it."

• You could read it beforehand. For the company you're meeting, that means its customer numbers and recent results, where it publishes them. Read those before you go in.

• You can't reach it from the room. The US dog count is published somewhere, but not where you're sitting, and nobody is likely to know how many owners would trade coverage for a lower price. Build either one from what you trust, as the chain above does, and label the result as an assumption.

Find the assumption that moves the result most

Change one assumption at a time, keeping the others at their middle values. This is called a sensitivity check. The starting point is 150 million × 4% × $700 = $4.2 billion.

The share insured swings the answer more than the other two combined. The pet count still moves it by about 27% ($1.12 billion ÷ $4.2 billion), so it matters, but the remaining time goes to the share.

One caution: in a pure multiplication, a 10% change in any input moves the answer by 10%, so these gaps only mirror how wide I set each range. What the table really shows is which input I'm least sure of, with a dollar figure on that doubt.

For the full range, I'll vary the share and the premium together and hold pets fixed at 150 million:

• Low: 150 million × 2% = 3 million. 3 million × $500 = $1.5 billion.

• High: 150 million × 8% = 12 million. 12 million × $900 = $10.8 billion.

My answer: about $4 billion a year, somewhere between $1.5 billion and about $11 billion, with the share insured as the input to check first.

How it compares

You won't have the following figures in the room. When you practice, though, look up the true figure afterward and trace the miss back through the structure and each guess.

The North American Pet Health Insurance Association reported about $5.7 billion in US premiums for 2025, on about 7 million insured pets. The American Veterinary Medical Association counted about 164 million dogs and cats in 2025. That works out to about 4.3% insured (7 ÷ 164) and about $810 per insured pet ($5.7 billion ÷ 7 million).

So the estimate came in about a quarter low and inside its range. All three inputs ran low, so no single guess explains the miss. The share insured landed close, and the largest miss was the premium, where $700 was low by about $110.

For the vocabulary around these methods, see market sizing and estimation cases. To practice, try Fermi Problem: Sizing the Pothole Count on Mainland American Roads or Estimate the Truck Fleet for a Grocery Service With a 24-Hour Delivery Promise.