How to Make a Friendship Quiz People Actually Finish

A friendship quiz is a strange object. It is a test the person taking it never agreed to sit, the marking is public, and the stakes are entirely social. By rights it should not work at all. It works enormously well, and the difference between a quiz that gets finished and one abandoned at question four is almost entirely structural rather than about how clever the questions are.

Just want to take one? The Friendship Calculator builds and scores a quiz about you in a couple of minutes, the Friendship Challenge is the version you send head-to-head, and Friend Mirror flips it so your friends describe you instead. The rest of this is for writing your own.

What follows is the structure: how long, what to ask, what order, and the part most people get wrong, which is that the best question in a friendship quiz is usually the most boring one in it.

Ten questions, and be suspicious of eleven

The instinct when writing a quiz is that more questions produce a more accurate result. In practice questions eleven through twenty tell you very little the first ten did not, and each one costs you some share of the people who would otherwise have finished.

The best available evidence on this is not about quizzes — it is about surveys, and the transfer is an inference rather than a measurement, so it is worth stating what the research actually found. SurveyMonkey analysed 100,000 surveys — 2,000 at each length from one to fifty questions — and found the relationship between length and drop-off is not linear. The sharpest increases in drop-off happen across the first fifteen questions. After about fifteen, each additional question costs less than the ones before it did.

The uncomfortable implication for quiz writers: a friendship quiz never escapes the expensive part of the curve. A fifty-question survey at least gets to the flat section where extra questions are nearly free. A ten-question quiz spends its entire length in the steepest stretch, so every single question has to earn its place.

Caveat worth keeping: that data is about surveys, which people take out of obligation. A quiz about a friend is voluntary and more interesting, so real drop-off is probably gentler. The direction of the effect is well supported; the exact numbers are not transferable.

Specific beats interesting

The most common mistake is writing questions that are enjoyable to read rather than questions that test attention.

"What is my dream holiday destination" is a pleasant question and a bad one. The answer is guessable from a general sense of the person, so getting it right demonstrates nothing. "What do I order every single time we go to that place" sounds worse and works far better — it is small, repeated, observable, and only someone who has actually been paying attention knows it.

There is a pattern in the right-hand column. Each one replaces a preference with a behaviour, and behaviours are observable in a way preferences are not. You can be told what someone likes; you have to have been present to know what they do.

The best question in a friendship quiz is usually one where the correct answer is completely boring and only a real friend knows it.

Six question types that reliably work

If the abstract advice is hard to apply, these are the six shapes that consistently produce a spread of scores. Most good quizzes are built almost entirely from them.

The standing order

What you get, every time, at the place you always go. Coffee, curry, drink, sandwich. It is the single best question type available because it requires no insight whatsoever — only presence. Someone who has been there knows it; someone who has not, cannot deduce it.

The thing you always say

A verbal tic, a phrase you reach for when annoyed, the way you answer the phone. People are almost entirely unaware of their own and everyone around them can quote it back. It reliably produces the biggest reaction of any question in a quiz, because the answer is simultaneously obvious to everyone else and news to you.

The dislike

More distinctive than a like, and much less often discussed. Everybody in a friendship group knows what everyone else enjoys — it comes up constantly. Dislikes surface far more rarely, so knowing one is genuine evidence of attention rather than of exposure.

The behaviour under pressure

What you do when running late, losing an argument, or waiting for bad news. This is the type that separates close friends from friendly acquaintances, because it requires having been present for something that was not fun.

The specific past detail

A first job, the town you grew up in, what you studied. These feel like easy questions and are frequently not — they get mentioned once, early in a friendship, and never again. Anyone who has known you five years either absorbed it in month one or never did.

The thing people get wrong about you

The hardest type and the one that produces the best conversation afterwards, because the answer is contestable in a way the others are not. Somebody will disagree with your own answer about yourself, and that argument is usually more interesting than the quiz.

Building a quiz quickly: take one question from each of the six types, add four more of whichever produced the best material, and you have ten without having to think about balance at all.

The difficulty curve

Question one decides completion, and it does so before the person has invested anything at all.

If the opener is hard, a reasonable share of people conclude they are going to do badly at this and close the tab — the quiz has told them the outcome before they had any reason to persist. If the opener is easy, the same person is now three points up and has a stake in finishing.

The curve that works: one easy, then a gentle climb, with the two hardest questions at positions eight and nine rather than at the very end. Finishing on something mid-difficulty means people end feeling competent, and people who end feeling competent are the ones who share the result — which is how the quiz reaches anyone beyond the person who wrote it.

Wrong answers have to be plausible

The part almost everyone under-thinks. Four options where three are obviously jokes is not a multiple-choice question, it is a formality with decoration, and a quiz made of those produces a score of ninety per cent for everybody — which measures nothing and is boring to receive.

Every wrong option should be something you can imagine being true of you. If you are asked what you order and the options are your actual order plus three things you would never touch, anybody who has met you once gets it right.

Three options is usually enough. Four is fine. Beyond that you are mostly generating filler, and filler options are where implausible ones creep in.

A quick way to generate good wrong answers: take the true answer and shift it one step rather than inventing something unrelated. If the real answer is that you always order the same curry, the wrong options should be three other things from that menu, not a salad. Near-misses force people to rely on actual recall; distant options let them rely on a general sense of who you are.

The test for a wrong option: if somebody who knows you a bit would be able to eliminate it without thinking, it is not doing any work. Every option should cost the person taking it at least a second.

You are a bad judge of what people know about you

The last structural problem is that the quiz author is the worst-placed person to guess which facts about them are visible. People systematically overestimate how well their preferences are known and underestimate how well their habits are.

The reason is straightforward enough: you experience your preferences constantly from the inside and your habits are invisible to you precisely because they are automatic. Everyone else has exactly the opposite view of you — they never see the preference and they see the habit every single time.

The practical fix is mechanical rather than introspective. Draft twenty questions, then cut every one where you can already predict what everybody will answer. What survives is the set where you genuinely do not know how people will do, and that is the only set that produces a spread of scores.

A ten-question template

If you want to skip the theory, this shape works. Fill in the blanks with your own specifics and you have a usable quiz in about ten minutes.

1. Easy opener. Something almost everyone gets — a visible, frequently-mentioned favourite.
2. A habit. What I always order, always wear, always say.
3. A small preference with a plausible alternative. Tea or coffee, and how I take it.
4. Something about my past. Where I grew up, what I studied, a first job.
5. A dislike. Dislikes are more distinctive than likes and much less often discussed.
6. A behaviour under pressure. What I do when I am late, stressed, or losing an argument.
7. Something only a close friend has seen. A private routine, a superstition, a specific fear.
8. Hard one. A detail I have mentioned exactly once.
9. Hard one. Something people usually get wrong about me.
10. Mid-difficulty finisher. Warm, gettable, ends the quiz on a right answer.

Question seven is the one that makes a quiz feel personal rather than generic, and question nine is the one that produces the best arguments afterwards. If you only put effort into two of them, make it those.

Who you send it to changes what it measures

A detail that gets ignored: the same ten questions produce completely different results depending on the audience, and it is worth deciding which one you actually want before writing anything.

Sent to three close friends, a quiz is a measurement. The scores will be high, the differences between them will be small, and the interesting output is the ranking — which of the three knows you best, a question that has probably never been settled.

Sent to thirty people including colleagues and people you met twice, it is not a measurement at all, it is entertainment. Most scores will be low, the ranking is meaningless, and the enjoyable part is the outliers: the person who barely knows you and somehow scored eight.

The mistake is writing for the first audience and sending to the second. Ten questions about small private habits sent to thirty acquaintances produces thirty scores of two out of ten, and a result nobody wants to share. If the list is wide, pitch the difficulty at someone who knows you moderately well, and put the genuinely obscure questions at eight and nine where they function as a tiebreak rather than as a wall.

The result screen is the product

It is tempting to treat the score as an output — the thing that comes out at the end after the real work. It is closer to being the point. The quiz is a mechanism for generating something shareable; the questions are how you get there.

A result that travels has three properties. A number, so it can be compared against someone else's. A label, so it can be quoted without a screenshot. And a way to challenge somebody, so the thing propagates instead of terminating with whoever took it.

Drop any one of the three and the quiz dies in the group chat it was sent to. That is not a small loss — most of the enjoyment in a friendship quiz happens in the argument afterwards, and without a shareable result there is nothing to argue about.

This is the part that is genuinely tedious to build by hand, which is the main argument for not building it by hand: the Friendship Calculator handles the scoring, the result card and the challenge link, and the party games list has the other formats worth running when a group is already together in one room.