Start with one question: median or mean?

The single most common way a true number misleads is the word “average.” The mean adds everything up and divides; the median is the middle person. For anything money-related — income, savings, net worth — a small number of very large values pulls the mean far above where most people actually are.

Imagine five households with $10,000, $20,000, $30,000, $40,000, and $2 million. The mean is $420,000. The median is $30,000. Both are “the average,” and only one of them describes a life anyone in that group is living. Whenever a headline says “the average American has X,” your first question should be which one was used — and if the article doesn't say, assume the flattering one. On this site we default to medians for exactly this reason, and we say so when a source only reports means.

A number is a snapshot of a distribution

Even a median compresses an enormous range into one point. Real populations are spread out — widely. For almost any life measure (savings, sleep, social time, exercise), the honest picture is a curve: a large middle, long tails, and far more variation than public conversation implies. When you see one number, ask what the 25th and 75th percentiles look like. If a source can't tell you, it is describing a point, not a population — and you can't locate yourself inside a point.

Check the denominator

“Millennials who invest in index funds report higher satisfaction.” Which millennials? All of them, or the subset who had money to invest, answered a brokerage's survey, and finished it? Every statistic has a denominator — the group actually measured — and it is usually narrower than the headline implies. Government surveys with large representative samples (the census, labor statistics, national health surveys) have broad denominators. A poll run by a company with something to sell usually doesn't. The number can be honestly computed and still describe a group you don't belong to.

Survivorship and selection: who you're actually comparing yourself to

The lives you see are a filtered sample. People post promotions, not layoffs; engagement photos, not quiet Tuesdays. Businesses that failed don't write “how I did it” threads. This is survivorship bias, and it doesn't just distort social media — it distorts your intuition about what's normal, because your intuition is trained on what's visible. When your gut says “everyone is doing better than me,” it is usually reporting on a curated sample, not a population. The research on this gap is one of the most consistent things in the library — see why you feel behind even when you're not and what comparison research actually shows.

Base rates: rare things are rare, common struggles are common

Before reacting to any claim, ask what share of people it applies to. Extraordinary outcomes get coverage precisely because they are exceptional — that's the definition of news. Meanwhile, the ordinary struggles almost nobody photographs (thin savings, interrupted sleep, friendships that take work) show up in population data at rates that would genuinely surprise most people. A useful habit: whenever a story makes you feel unusual, go find the base rate. It is usually far less lonely than the feeling suggests — actual savings by age is a good place to see this in action.

One study is a hint, not a fact

Single studies — especially small, surprising ones — are where headlines come from and where replication failures live. Findings worth building a picture on tend to be replicated across samples, hold up in meta-analyses, or come from large representative surveys. When research on a question is genuinely mixed, the honest summary is “mixed,” not whichever side is more shareable. That's why pages here hedge (“roughly,” “about,” “in most studies”) and say plainly when evidence is thin. Precision that the underlying research can't support isn't rigor — it's decoration.

Self-reported numbers drift

People round their height up, their weight down, their gym visits up, and their screen time down — not from dishonesty, but because memory is reconstructive and questions carry social pressure. Numbers measured directly (time-use diaries, device logs, administrative records) routinely disagree with what people report about themselves. When two sources conflict, check whether one of them asked people and the other watched people. The watched number is usually closer.

A checklist for any statistic you meet

  • Who measured it? A statistical agency, a research team, or someone selling something?
  • Median or mean? If money is involved and it doesn't say, be suspicious.
  • Who is in the denominator? Everyone, or a narrow, self-selected slice?
  • When? Numbers about money and technology age quickly; a decade-old figure is a different world.
  • Is it measured or self-reported? And in which direction would people shade the truth?
  • Is it one study or a pattern? Surprising single results deserve patience, not panic.
  • What's the spread? If you only got one number, you haven't seen the distribution yet.

How we hold ourselves to this

Every page on this site names its sources — government data, peer-reviewed research, and established surveys only — and every number links back to where it came from. You can read the methodology, browse the full source list, see the headline figures collected on By the Numbers, or check where the data comes from. If a number here ever fails the checklist above, that's a bug — tell us.