Fermi Estimation for Kids: How Many Fit in There?

Guessing and estimating feel like the same thing to a child. They are not, and a jar of dried pasta settles the argument in forty minutes. Have everyone write down a wild guess, sealed. Then do it properly: count how many pieces fit across the bottom, work out how many layers deep the jar goes, multiply. Then tip the jar out and count the lot. The guesses are usually wrong by a factor of five. The step-by-step estimate is usually within 30%. Fermi estimation for kids is the skill of getting an answer within a factor of two about something you cannot count, and it is learnable in one afternoon.

Key Takeaways

• You’ll make three numbers for the same question: a sealed guess, a structured estimate built from measurements, and the true count from tipping the jar out.
• It teaches order-of-magnitude reasoning and why rounding everything hard still gives a usable answer.
• Best from 9, when multiplying two-digit numbers and dividing are solid enough not to swamp the thinking.
• The sealed guess is the part people skip and the part that makes the lesson land. Without it there is nothing to beat.
• Then apply the method to something uncountable, like the leaves on a tree, where you write the chain of assumptions down one line at a time.

What You’ll Need

Item	How much	Where to get it	If you don’t have it
A straight-sided glass jar	1, around 500 ml	Jam jar, pickle jar	A plastic tub, as long as the sides do not taper
Small identical objects	Enough to fill the jar: dried penne, popcorn kernels or identical building bricks	Kitchen cupboard	Marbles, buttons, or 2 cm paper squares
Ruler	1, with millimetres	School supplies	A tape measure
Kitchen scale	1, optional but useful	Kitchen	Skip the weighing variation
Tray or large plate	1	Kitchen	A tea towel, for sorting into groups of ten
Paper and pencil	3 sheets	Anywhere	The estimate chain needs one line per step

Time: setup 5 min, estimating 15 min, counting 20 min, second problem 15 min Cost: nothing, and the pasta goes back in the cupboard Mess level: medium, because something will roll off the table

Do the Experiment: Fermi Estimation for Kids, Step by Step

Step 1 – Everyone writes a sealed guess

Hold up the filled jar. Everybody writes a number on a scrap, folds it, and puts it in the middle of the table. No discussion, no looking at anyone else’s. Adults play too, badly, which is part of the point.

Step 2 – Count one layer across the bottom

Empty the jar. Place pieces across the bottom in a single layer and count them. Do not estimate this: count. For a jam jar and penne you will get something between 12 and 25. Write it on line one of the estimate sheet.

Step 3 – Work out how many layers fit

Stack ten pieces on the table and measure the height of the stack with the ruler, then divide by ten to get the height of one. Measure the inside height of the jar. Divide. Write both numbers and the division on lines two and three. Stacking ten and dividing is the step that makes this accurate: measuring one piece of pasta to the millimetre is hopeless, and measuring ten is easy.

Step 4 – Multiply, then round hard

Multiply pieces per layer by number of layers. Then round the answer to one significant figure and write the rounded version in large handwriting. If you got 287, write 300. The rounding is not laziness. It is a claim about how much precision the method deserves.

Step 5 – Tip it out and count it properly

Pour everything onto the tray and sort into groups of ten, then count the groups. For 300 pieces this takes about twelve minutes with two people, and it is the step that gives the activity its authority. Do not skip it and do not estimate it.

Step 6 – Compare all three numbers, in order

Open the sealed guesses. Write the three results side by side: guess, estimate, true count. Then compute how far out each one was, as a multiple rather than a difference: “my guess was 3 times too small” says more than “my guess was 200 out”.

Step 7 – Now estimate something you cannot count

Pick a target that cannot be counted: the leaves on a tree in your street, the blades of grass in one square metre of lawn, the number of hairs on a head. Build the chain out loud, one assumption per line, measuring what you can. For a tree: count leaves on one small twig, count twigs on one branch, count branches. Three measurements and two multiplications.

What you should see: Sealed guesses from children scatter enormously, often by a factor of ten between family members, and are usually too low for a jar and too high for a room. The structured estimate typically lands within 20% to 40% of the true count, and almost always within a factor of two. Expect the layer count to be the accurate part and the layer height to be the sloppy part, because pasta does not stack neatly. The tree estimate will produce a number in the thousands and nobody will be able to check it, which is exactly the situation the method exists for.

What’s Actually Going On

A Fermi problem is a question you answer by breaking it into pieces you can estimate, multiplying through, and accepting an answer good to within a factor of two or so. Jan Ärlebäck’s 2009 paper in The Mathematics Enthusiast defines them for school use and makes the key point: the value is in the modelling, not the arithmetic. The name comes from the physicist Enrico Fermi, who was known for demanding this kind of back-of-envelope estimate before anyone started a real calculation, and the piano-tuners-in-Chicago problem is traditionally attributed to him.

Here is the part that is almost never explained, and it is the actual mathematics. Why does a chain of five crude guesses not produce a catastrophically wrong answer? Because the errors multiply rather than add, and independent multiplicative errors partially cancel. Suppose you make five estimates and each one might be wrong by a factor of two, randomly too high or too low. The worst case is 2⁵, a factor of 32. But the worst case needs all five errors to point the same way, which is unlikely. Typically some are high and some are low, the overestimates divide out the underestimates, and the product lands within a factor of two or three. Estimation is reliable because of cancellation, not because the individual guesses are good.

That has a practical consequence worth telling a child: it is better to break a problem into more steps, each guessed roughly, than into fewer steps guessed carefully. More steps means more chances to cancel. It also means one step you are badly wrong about, by a factor of a hundred, will wreck the whole thing, so the useful discipline is to look at each line and ask “could this one be out by a lot?”

This connects to something measurable about children’s thinking. Robert Siegler and Julie Booth’s work on numerical estimation shows that children’s sense of where numbers sit on a mental number line develops gradually and becomes more accurate with age, shifting from a compressed, logarithmic-looking pattern to a linear one. Booth and Siegler later found that the accuracy of those magnitude representations was related to how well children learned arithmetic. Estimating is not a party trick that sits beside maths; it is connected to the machinery that makes arithmetic make sense.

One honest caution about the method. Fermi estimation gives you an order of magnitude, not an answer. It will tell you whether a claim is roughly plausible, which is an enormously useful thing in a world full of numbers somebody made up. It will not tell you how many beans are in the jar to the nearest bean, and pretending otherwise is the main way this gets taught badly.

How to Explain It to Your Kid

Ages 5–8: how many fit across the bottom?

Do not say estimate. Fill one layer, count it together, and then ask: “If we stacked another layer on top, how many would that be?” Count the layers physically by filling the jar one layer at a time, saying the running total out loud. At this age the win is the idea that you can find a big number by counting a small group and then counting the groups. That is multiplication arriving through the hands.

Ages 9–12: build the chain on paper

Insist on one assumption per line, numbered, with units. Then ask the question that trains judgement: “which line are you least sure about?” They will usually point at the right one. Then redo just that line more carefully and see how much the final answer moves. Sometimes a lot, sometimes almost nothing, and learning to tell the difference is the whole skill.

Ages 13+: bound it from both sides

Have them produce a low estimate and a high estimate deliberately, using the smallest and largest plausible value on every line. The two answers bracket the truth. Then set the real investigation: how many litres of water does your household use in a year? Build it from people, showers, minutes and flow rate measured with a bucket and a stopwatch, then check it against the water bill. A Fermi estimate you can audit against a real number is the best version of this exercise there is.

The question to ask: “Somebody on the internet says there are a million pigeons in our city. Without looking anything up, how would you work out whether that is roughly right or ridiculous?”

Why Teaching This Early Pays Off

Estimation is the most broadly useful mathematical skill that school barely teaches. A person who can do this spots a nonsense statistic in a news story, catches their own calculator slip, sizes up whether a plan is feasible, and asks better questions in a meeting. Engineers do it constantly, usually before touching a computer.

The specific payoff from this activity is calibration: knowing how wrong you usually are. A child who has compared a sealed guess against a structured estimate against a real count has one data point on their own reliability, and that is the start of intellectual honesty about numbers. The Framework for K-12 Science Education names using mathematics and computational thinking as a core practice, and Fermi problems are the cheapest serious form of it.

The evidence on estimation is better than most hands-on claims, though still not a promise about grades. Siegler and Booth documented how estimation accuracy develops in young children; Booth and Siegler linked the accuracy of number-magnitude representations to arithmetic learning. That is a real association, not evidence that one jar of pasta changes anything. What one afternoon does is give the word “estimate” a procedure attached to it instead of being a polite word for guessing.

Before You Start: What Parents Should Know

Best age	9 and up for the full chain. From 5 for the layer-counting version in the Ages 5–8 section. The jump at 9 is the arithmetic: if multiplying 18 by 14 costs all their attention, there is none left for the reasoning.
Supervision	Light, but you are the record-keeper. The estimate chain must be written down or the comparison at step 6 is impossible.
Real hazards	Small parts. Dried pasta, beans, marbles and kernels are a choking risk for children under 3, so do this on a table out of reach and sweep up afterwards. Nothing here is edible mid-activity, and raw dried kidney beans in particular should never be eaten; use pasta or building bricks and the question does not arise.
Setup	A table big enough for the jar, the tray and the sorting. Counting 300 items needs more surface than people expect.
Frustration point	Step 5, the real count. Twelve minutes of sorting into tens is genuinely boring. Split the tray in half, one person each, and announce a target time.
Don’t	Don’t skip the sealed guess, and don’t let anyone revise their guess after seeing the jar emptied. The comparison is the entire lesson.

When It Doesn’t Work

• The estimate is wildly out, by more than a factor of three. Almost always the layer height. Pasta and kernels do not stack in flat layers, so the real number of layers is lower than the ruler suggests. Measure a stack of ten pieces inside the jar rather than on the table.
• The jar tapers. A jar with sloping shoulders breaks the layers-times-layer-count model, and the estimate comes out too high. Use a straight-sided jar, or stop filling at the shoulder and measure to there.
• The real count disagrees with itself. Two people counted the same group of ten. Sort into tens first, then count only the groups, and move each counted group physically to one side.
• Everyone’s sealed guess is close. It happens, especially with adults who have done this before. Use a harder container next: a shoebox, or the number of grains of rice in a cup.
• The tree estimate goes nowhere. Counting leaves on a twig is harder than it sounds on a dense tree. Pick a sparse branch, or switch to blades of grass in a 10 cm square, which is countable in three minutes.

Take It Further

Audit an estimate against a real bill. Estimate your household’s annual water use from first principles: number of people, showers per week, minutes per shower, litres per minute measured by timing how long a bucket takes to fill. Then find the figure on the water bill. Agreement within a factor of two is a success, and disagreement is more interesting because you can find out which line was wrong.

Weigh instead of counting. Count out exactly 50 pieces of pasta, weigh them, weigh the whole jarful, and divide. Compare this estimate with the layers method and with the true count. Weighing usually wins, and asking why is a good conversation about which measurements are easy to make precisely.

Run a calibration game over a week. Each day, one person poses an estimation question and everyone writes a range they are 90% confident contains the answer. At the end of the week, count how often the true answer fell inside the range. If it is nine times out of ten, the ranges are honest. Most people are far too confident, and seeing that in your own numbers is the lesson.

What to Watch For Over the Next 3 Months

• Week 4: They challenge a number out loud, usually one from a game or an advert, with something like “that can’t be right”.
• Month 2: The phrase “about how many” starts replacing “how many”, and they begin breaking questions into steps unprompted.
• Month 3: The question worth encouraging is one about something genuinely unknowable, like how many words they have said in their life. Do not answer it. Build the chain together.

Frequently Asked Questions

What is the difference between a guess and an estimate?

A guess is one number produced with no structure. An estimate is built from measurable or knowable parts, multiplied together, with the assumptions written down so each can be checked and corrected. The difference shows up in the comparison at step 6, where the structured estimate reliably beats the guess.

Why is it fine to round everything so roughly?

Because the errors are multiplicative and partly independent, so overestimates and underestimates cancel. Five rounded steps, each possibly wrong by a factor of two, usually combine to an answer within a factor of two or three rather than a factor of 32. Precision on any single line is wasted effort.

What age does this start working?

About nine, set by the arithmetic rather than the idea. If multiplying a two-digit number by a two-digit number takes all of a child’s working memory, there is nothing left for the reasoning. Younger children get real value from the layer-counting version.

How accurate should we expect to be?

Within a factor of two for a well-built chain, and often within 30% for the jar. If your estimate is out by more than a factor of three, one line is badly wrong rather than all of them being slightly wrong, and finding that line is the most useful part of the exercise.

Is this actually useful or is it a maths game?

It is what engineers and scientists do before committing to a real calculation, because it catches impossible plans and wrong units early. For a child, the everyday version is noticing that a number somebody stated confidently is off by a factor of a hundred.

---

About the author

Ricky Flores is the founder of HiWave Makers and an electrical engineer with 15+ years of experience building consumer technology at Apple, Samsung, and Texas Instruments. He writes about how kids learn to build, think, and create in a tech-saturated world. Read more at hiwavemakers.com.

---

Sources

1. Ärlebäck, J. B. (2009). “On the use of realistic Fermi problems for introducing mathematical modelling in school.” The Mathematics Enthusiast, 6(3), 331–364. https://doi.org/10.54870/1551-3440.1157
2. Siegler, R. S., & Booth, J. L. (2004). “Development of Numerical Estimation in Young Children.” Child Development, 75(2), 428–444. https://doi.org/10.1111/j.1467-8624.2004.00684.x
3. Booth, J. L., & Siegler, R. S. (2008). “Numerical Magnitude Representations Influence Arithmetic Learning.” Child Development, 79(4), 1016–1031. https://doi.org/10.1111/j.1467-8624.2008.01173.x
4. National Research Council. (2012). A Framework for K-12 Science Education. National Academies Press. https://doi.org/10.17226/13165
5. Chi, M. T. H., & Wylie, R. (2014). “The ICAP Framework: Linking Cognitive Engagement to Active Learning Outcomes.” Educational Psychologist, 49(4), 219–243. https://doi.org/10.1080/00461520.2014.965823
6. NRICH, University of Cambridge. “Exploring Wild and Wonderful Number Patterns” and related number investigations. https://nrich.maths.org/problems/exploring-wild-wonderful-number-patterns

Related reading: what a million actually looks like, measuring pi with string, and rolling 1,000 dice.