Estimating a quantity nobody can count: guess, model, assume, approximate

0:00How many one-rupee coins would it take to balance you? Read that again. There is no number in it. Nothing given, nothing to substitute, nothing to solve for. And yet it has an answer. Not because you know something. Because you can build something. An estimate is not a calculation done badly. It is a different object with a structure of its own: a relation, assumptions you say out loud, and an answer reported no further than they reach.

0:34Get that structure right, and two honest people can disagree by a factor of six and both still be right. Get it wrong, and no assumption you could defend will save you. Start somewhere easier. Someone wants to give away sugar matching one child's weight, and flour matching another's. The child on one side of a balance, goods on the other until the beam is level. The question is what it will cost.

1:05The instinct is to start hunting for numbers. Resist it. Reaching for arithmetic before you know what multiplies what is how people go wrong. The first thing to write is not a number at all. It is a sentence about which quantities meet. Here it is. The cost of the sugar is the child's weight in kilograms, times the price of one kilogram. That is the whole of the mathematics in this problem.

1:37Two slots, both empty, and a multiplication between them. It says the answer scales with the weight. And with the price, in exactly the same way. It says nothing about ages, because age is not in it. Everything from here is filling in slots, and that part can differ from person to person. The relation cannot. So: what were we actually told? Two ages. One child is thirteen, the other eleven.

2:10That is it. No weights, no prices. So both slots must be filled by us, in the open. Say the older child weighs forty-five kilograms, and sugar costs seventy rupees a kilogram. Forty-five kilograms at seventy rupees a kilogram is three thousand one hundred and fifty rupees. Say the younger weighs fifty kilograms, and flour costs fifty rupees a kilogram: two thousand five hundred. You probably objected just then. The younger child came out heavier than the older.

2:47You are allowed to. That is what saying it out loud is for — the error is not assuming, it is assuming silently. Notice also that the dearer goods gave the bigger bill, and not because they were dearer. Sugar at seventy was matched to the lighter child, and still cost more than flour at fifty. You cannot read that off either number. Only the relation knows. So here is the routine. Four steps, in order.

3:18One: guess. Say a number before you calculate anything. Two: relate. Write down which quantities multiply. Three: assume. Fill every slot nobody gave you, and say what you filled it with. Four: compute — and then look back at your guess. Guessing is first, and that is not decoration. Here is why. Suppose your arithmetic says five hundred coins. Is that right? If you never guessed, you have nothing to check it against.

3:52It came out of the arithmetic, and arithmetic is happily correct about a wrong relation. But if you had said beforehand — thousands, surely — five hundred arrives and something in you objects. The guess is the first thing that can notice a wrong answer. And a wrong guess costs nothing. Nobody marks it. It is not the answer; it is the alarm. So, back to the coins, properly this time.

4:23Guess first. Hundreds? Thousands? Lakhs? Crores? Now the relation. The number of coins, times the weight of one coin, is the weight of the person. Again, no numbers yet, and the shape is already fixed. It says heavier coins mean fewer of them. It says the answer is a division: the person's weight, one coin at a time. And it tells us which slots nobody has filled. Nothing in that question mentions what a coin weighs — or what you weigh.

4:57That is not a defect in the question. That is the question. So weigh a coin, or assume one. Say a one-rupee coin is four grams, and say the person on the pan weighs forty-five kilograms. Both of those are ours. Neither was given. Everything has to be in one unit, so forty-five kilograms becomes forty-five thousand grams. Forty-five thousand divided by four is eleven thousand two hundred and fifty.

5:31Thousands. Check that against the guess you were holding. But do not write that down. It claims to be right to the last coin, and rests on a mass somebody guessed. Say about eleven thousand — or ten thousand, if you claim only the one digit your assumption gave you. And eleven thousand coins is eleven thousand rupees, against three thousand one hundred and fifty for the sugar. Same weight, different bill.

6:02What about five-rupee coins, or ten-rupee notes? Guess again first. A five-rupee coin is heavier — say six grams. Forty-five thousand divided by six is seven and a half thousand coins, worth thirty-seven thousand five hundred rupees. A note is paper. Say one gram. Forty-five thousand notes, worth four hundred and fifty thousand rupees. Going from coins to notes multiplied the number of objects by four. It multiplied the money by forty.

6:39Same relation, same weight — but the answer in rupees did not follow the answer in objects. This has to be rebuilt, not scaled. The relation does not have to end in money at all. Suppose the younger child, grown up, gives away grain matching their own weight every year. Fifty kilograms of it. How many people does that reach? The number of people, times the grain in one person's share, is the total grain.

7:12Nobody has told us what one share is — two slots empty, and a relation answers exactly one question at a time. So assume it. Say half a kilogram each. Fifty thousand grams divided by five hundred grams is one hundred people, every year. Because the relation is a multiplication, halving the share doubles the count: two hundred people from the same fifty kilograms. That is not a flaw. Every assumption in the relation carries the answer in full.

7:50Now something that looks like a different kind of problem. A group of walkers arrived in town this morning, having covered about four hundred kilometres on foot. When did they set out? No price, no total handed over — it feels like a new technique. It is the coin question again. Distance is walking speed times time — three slots, exactly like the others. We know the product and one factor, and want the other. Only the story has changed.

8:24Assume four kilometres an hour, and four hundred kilometres is a hundred hours of walking. Assume eight hours a day, and that is twelve and a half days — call it thirteen, and do not pretend to the half. One more, worth guessing at first. How many times could a person walk round the world in a lifetime, walking every day? Once? Tens of times? Hundreds? Round the world is forty thousand kilometres — the only number you are given.

8:57Four kilometres an hour for eight hours is thirty-two kilometres in a day. Three hundred and sixty-five days a year, sixty years of walking. That comes to seven hundred thousand, eight hundred kilometres in a life. Divide by forty thousand: seventeen and a half times round. Call it eighteen. Tens — not once, not hundreds. If you guessed tens, your instinct is already working. If everyone assumes their own numbers, how can anybody be wrong?

9:33Try every coin mass a reasonable person might pick: two grams up to twelve, eleven different guesses. Every answer lands between three thousand seven hundred and fifty at one end, and twenty-two thousand five hundred at the other. The two extremes disagree by a factor of six, and all eleven say thousands. Not one can be shown to be wrong, and the disagreement never changed the answer. Now the other kind of mistake. Multiply by the coin's weight instead of dividing.

10:08Sweep the same eleven assumptions and you get ninety thousand up to five hundred and forty thousand — a band that never once touches the right one. Its closest approach is still four times outside. And notice what the guess alone would have done. The cheapest of those wrong answers still reads as thousands. The guess raises the alarm; the sweep is what convicts. That is the difference. A number you chose can be argued with; a relation you got wrong cannot be argued back into place.