The Most Average Website
An average is a sentence with almost all of the information taken out.
10 of them, each with its source, its date, and what it does not prove.
Exhibit 1
The average colour of the universe
#FFF8E7
Averaged across the light of more than 200,000 galaxies, the universe is a slightly beige white. It was announced in 2002 as turquoise. That was wrong: the software had been calibrated incorrectly, and colour specialists spotted it before the astronomers did. The corrected shade was named by public vote — "cosmic latte" beat "cappuccino cosmico".
Glazebrook & Baldry, Johns Hopkins, 2002. The survey paper itself states no colour; this figure comes from their separate colour analysis.
Exhibit 2
The average human body temperature
falling 0.03 °C per birth decade
Across 677,423 measurements spanning 157 years, men born in the early 1800s ran 0.59 °C warmer than men today; women are down 0.32 °C since the 1890s. What makes measurement error an unlikely explanation is that the same decline shows up within the Union Army cohort, not only between cohorts.
What the study does not settle is the famous 37 °C itself. Whether that 1850s figure was accurate, or whether the thermometers of the day simply read high, is in the authors’ own words "remaining unanswered". Bodies have changed. Whether the original number was right is a different question, and this is not the paper that answers it.
Protsiv et al., eLife, 2020.
Exhibit 3
The average reading speed
238 words per minute
English non-fiction, read silently, by adults. Fiction runs faster at 260, because its words are shorter. Aloud, it drops to 183. The number is not the interesting part — the paper’s finding about the number is: these estimates are "lower than the numbers often cited in scientific and popular writings". The average everyone quotes is wrong, and this is the average of 190 studies establishing that.
Brysbaert, Journal of Memory and Language 109, 2019. 190 studies, 18,573 participants.
Exhibit 4
The average day
86,400.000 to 86,400.002 seconds
A day is defined as 86,400 seconds. Since 1972 the average one has run nought to two milliseconds longer. It has been lengthening for a very long time — roughly 2.3 milliseconds per century since the eighth century BCE, as the Moon draws off Earth’s rotation. Then on 29 June 2022 the planet turned in 1.59 milliseconds under 24 hours, the shortest day ever recorded. Averages are like that.
As stated by Wikipedia’s "Earth’s rotation" article, fetched 31 August 2026.
Exhibit 5
The average colour of this page
#AE8D82
Every colour this page can draw with, averaged. It is computed from the stylesheet itself when the page is built, so it describes the real palette rather than a note about the palette written down separately and left to drift. There are 10 of them.
Computed at build time from src/styles/global.css.
Exhibit 6
The average exhibit here
76 words
Counted across the 4 sourced exhibits, not including this one. Including this one would change the answer, which is a small demonstration of why measuring the thing you are part of is harder than it looks.
Computed at build time from this catalogue.
Exhibit 7
The average of every number on this page
48,812.19
Body temperatures, galaxy counts, seconds, participants and years, added together and divided by how many there are. The units do not match and the result means nothing whatsoever. It is, however, an average, and this is a museum of averages.
Computed at build time from every numeral in the sourced exhibits.
Exhibit 8
The average word on this page
4.65 letters
Measured across every word of the sourced exhibits. There is a tempting sentence to write here comparing that to written English generally — and no sourced figure on this page to support it, so it is not written.
Computed at build time from this catalogue.
Exhibit 9
The average letter of the alphabet
the 13½th
Between M and N, and therefore not a letter. This is the only exhibit here that required no measurement, no source and no fetching: it follows from the definition of an average and the number 26. It is also the only one that will still be correct in a million years.
Arithmetic.
Exhibit 10
The average exhibit
number 5.5
There are 10 exhibits in this museum, so the average one is number 5.5, which does not exist. Every exhibit here is either side of it.
Computed at build time from the length of this catalogue.
What is not here
Four averages were researched and not shown. They are listed because a museum that displays only what it could source is a different institution from one that displays everything it found.
- The average human height
- No defensible global figure exists — only country-by-country tables and, in one 2023 study, a range rather than a mean. A per-country number would be sourceable and would also be an average of a subset, on a page about the average of everything.
- The average song tempo
- Every source returns 120 to 130 BPM as a range, from a different corpus each time: top-10,000 streamed, the Million Song Dataset, Billboard, nine million Spotify tracks. The methodology drifts, not just the value, so pinning a date to it would not make it true.
- The average sentence length
- Answers ranged from six words to nearly nine depending entirely on which corpus and which register was counted. There is no number to print.
- The average name
- Depends on a country and a year, and is therefore an average of a subset with an expiry date.
That was an advertisement, positioned and sized entirely conventionally. We considered labelling it "the world's most average ad placement", which is funnier, and then read the policy on labelling advertisements clearly.
Certificate of Average
Enter anything you like. The certificate declares you average regardless, because this page has no way of checking and would not be qualified to judge if it did.
A museum of averages, and what averaging costs
Every exhibit here is an average, and every one of them is a small lie. That is not a complaint about the arithmetic — the arithmetic is fine. It is that an average takes a distribution, which has a shape, and returns a number, which does not. We keep them anyway, because the alternative is 677,423 thermometer readings and no idea what they mean.
The mean is almost never the right average
There are at least three things called "the average" and they answer different questions. The mean adds everything up and divides. The median is the value in the middle when you line them all up. The mode is whichever value turns up most. For a symmetrical distribution they land in roughly the same place, which is why the distinction feels academic until it very suddenly is not.
The mean is the one everybody reaches for, and it is the one that a single extreme value can drag anywhere it likes. Put a billionaire in a room of a hundred people and the mean wealth in that room is tens of millions; the median is unchanged. Nobody in the room has the mean amount of money. It describes no one present. That is not a failure of arithmetic, it is the mean doing exactly what it is defined to do, and the reason it keeps appearing in headlines is that it is the one that moves.
What "the average person" hides is more than skew, though. It hides the fact that averaging across a population produces a description of nobody in it. The classic demonstration comes from aircraft design: cockpits built to the average pilot's measurements fit almost no actual pilots, because a pilot who is average in height is rarely also average in arm length, and one who is average in both is rarely average in the third thing you measure. Average across enough dimensions and the number of people matching the composite falls to zero. The average is a real number about a person who does not exist.
That is worth holding while reading the rest of this page. The average reading speed of 238 words per minute is a genuine finding about 18,573 people, and there is no particular reason to think it describes you. The average body temperature is falling, and yours is doing whatever it is doing. Every exhibit is true and none of them is about anybody.
Why some of these are about the page itself
Roughly half the exhibits are measurements of this page. The average colour of its own palette, the average length of its own exhibits, the average of every number printed on it. These are the only exhibits that cannot rot, cannot be miscited, and are not aggregated from somebody else's work — they are computed when the page is built, from the page, so they are exactly true at the moment you read them and will still be exactly true if somebody edits the prose around them.
The colour one is the clearest case. It would have been easy to write the hex down and move on, and it would have been wrong the first time anyone adjusted a token. Instead the page reads its own stylesheet during the build and averages what it finds. There is no version of that which can drift.
The cuts are an exhibit
Four averages were researched and are not shown, and they are listed rather than quietly dropped. This matters more than it sounds. A museum that displays only what it could source is a different institution from one that displays everything it found — the first is curating, the second is reporting, and the difference is invisible to the visitor unless the rejections are on the wall.
It also happens that the cuts are more instructive than the survivors. Average height fails because there is no global figure, only a hundred national ones. Average song tempo fails because every source uses a different corpus and returns a range. Average sentence length fails because the answer depends entirely on what you counted as a sentence. In each case the problem is not that the number is hard to find. It is that the question has no single answer, and the confident figures you will find elsewhere are answers to narrower questions with the narrowing left out.
Sourcing, honestly
Every exhibit carries its source on the exhibit rather than in a footnote, because exhibits get quoted one at a time and a citation at the bottom of a page does not survive the crop. Where the sourcing is weaker, it says so: two exhibits rest on an encyclopaedia and name it, rather than citing the measurement technique the encyclopaedia was itself summarising.
One near-miss is worth recording. The average colour of the universe is usually cited to the survey paper that produced the data. That paper does not contain a colour, or a hex value, or any statement about what colour the universe is — it links out to a separate analysis by the same authors. A citation can be perfectly formatted, point at a real paper by the right people in the right year, and still attach to a claim the paper never makes.
About the certificate
The Certificate of Average takes whatever you type and declares you average. It does not compare your answers to anything, because it cannot: there is no sourced figure here for hours slept, and inventing one to make the interaction feel more rigorous would be exactly the failure the rest of the page is about.
What it is instead is an honest joke. The certificate is issued by a website, on no evidence, and says so on its face. That puts it slightly ahead of a number of things people frame.
Why the design is like this
The page is deliberately unremarkable: ordinary type at an ordinary size, an ordinary measure, everything centred, nothing that would catch a designer's eye. What is not average is the contrast and the tap targets — the genuinely average website fails both, and being average is not a defence for being unusable. That gap is arguably the most honest exhibit here, and it is the one we could not bring ourselves to put in a frame.
Questions
Where do these averages come from?
Each exhibit carries its own source and the year it was measured, on the exhibit rather than in a footnote. Two rest on an encyclopaedia and say so; the rest are from the papers themselves.
Why are some exhibits about the page itself?
Because they cannot go out of date. They are computed from the page when it is built, so they are exactly true when you read them and stay true if the prose around them is edited.
Why is the average of every number on the page meaningless?
Because the units do not match. Adding body temperatures to galaxy counts and dividing produces a number with no interpretation, which is included deliberately: it is what averaging looks like when nobody checks whether the things being averaged belong together.
What was left out?
Average height, song tempo, sentence length and the most common name. Each is listed on the page with the reason, because what could not be sourced is more revealing than what could.
Is the mean the right average to use?
Usually not. The mean is the one everybody reaches for and the one most easily dragged around by outliers, which is why the median is often the honest answer and almost never the quoted one.
Is the design of this page deliberately unremarkable?
Yes, in its typography and layout. Not in its contrast or its tap targets: the genuinely average website fails both of those, and being average is not a defence for being unusable.