Skip to main content

Data refreshed daily · DR by Ahrefs

Sign in

Percentiles and Leaderboard Statistics Made Plain

Statistics · 10 min read ·

Ranks, percentiles, medians and averages tell different stories. A plain guide to reading leaderboard numbers and avoiding common traps.

Illustration: A distribution chart on light grey with a skewed histogram, a median line in teal, a mean line in amber and shaded quartile bands labelled

A leaderboard is a list in order. First, second, third. It looks simple, and in a sense it is. But the numbers around a leaderboard, the percentiles, the averages, the changes, can be read many ways, and some of those ways mislead.

This guide explains the common statistics you meet on leaderboards and dashboards in plain words, shows where each is useful or misleading, and offers rules for reporting your own position honestly. No special mathematics is needed, only attention.

Rank: the simplest number

A rank is a position in an ordered list. First is the highest by whatever measure the list uses.

What a rank tells you: where you stand relative to others in this list, this week.

What it leaves out:

  • How far apart the entries are. First and second may be a hair apart, while second and third are far apart.
  • How many entries there are. Fifth of ten is not the same as fifth of five hundred.
  • What was measured. Votes, visits, growth, quality?
  • Who was included. A rank among a chosen few is not a rank among everything.

So a rank should always travel with its list: "fifth of forty in this category this week by community votes".

Percentiles: position, scaled

A percentile puts a rank on a common scale. The value at the 80th percentile is the one below which 80 per cent of observations fall. If you sit at the 80th percentile, you did better than about four in five of the group.

Percentiles make comparisons easier across lists of different sizes. Being fifth of ten and fifth of five hundred are very different, and the percentile shows it.

Useful things to know:

Quartiles split the group into four equal parts. The first quartile is the 25th percentile, the second is the median and the third is the 75th. "Top quartile" means the best quarter.

Deciles split it into ten parts. "Top decile" means the best tenth.

Percentiles are about order, not size. Two entries in the 90th percentile may differ greatly in raw score.

Small groups make jumpy percentiles. With twenty entries, a one-place change moves your percentile by five points. Always show the count.

Mean, median and mode

Three kinds of "typical".

The mean, or average, is the sum divided by the number of values. It uses every value, which is its strength and its weakness: one very large number pulls it up.

The median is the middle value when numbers are sorted. Half are above, half below. It resists extremes.

The mode is the most common value.

Most leaderboard data are skewed. A few entries have very large numbers and most have small ones. In such distributions, the mean is often much higher than the median, and quoting the mean makes a typical entry look better than it is.

For example, if most products in a category receive a few dozen votes in a week and one receives thousands, the average is far above what the typical product gets. The median tells you what a typical entry achieved.

When reporting a typical result, prefer the median, and say so.

Skew and the long tail

Many real distributions follow a pattern in which a small share of entries account for most of the total. The Pareto principle, often summarised as eighty per cent of effects coming from twenty per cent of causes, is a rough description of this kind of imbalance. It is not a law, and the real proportions vary, but the shape is common: a few giants and a long tail of small ones.

What this means for you:

  • Do not compare yourself with the top few unless you have their resources. Compare with the median and with similar entries.
  • Expect most entries to be in the tail. Being in the tail is normal.
  • Growth in the tail is real growth. A move from the 30th to the 50th percentile is meaningful even if you are far from first.

Change over time

Leaderboards change weekly. Reading change takes care.

Rank change is the number of places moved. It depends on how many entries sit near you. Moving up three places in a crowded middle is a small change in score. Moving up three at the top may be large.

Score change is the difference in the underlying measure. It is usually more informative.

Percentage change compares to a base. It exaggerates small bases: from two to four is a hundred per cent, and still two.

Percentile change shows movement relative to the group, scaled.

Present change with a sign and an arrow, never colour alone: up with a plus, down with a minus, a neutral mark for none. Include the period and the dates.

Sample size

Every statistic depends on how much data stands behind it. With a handful of entries or a few votes, results are fragile.

  • Show the count. "Eleven votes" is different from "eleven hundred".
  • Be cautious about percentages. A conversion rate from twenty visits means little.
  • Wait for more data before drawing conclusions.
  • Look at ranges, not just points, where you can.

The temptation is to treat any number from a dashboard as precise. Small numbers are noisy.

Bias in who appears

Leaderboards show who is on the list. Those who are absent are invisible.

Survivorship bias is the error of concentrating on the entries that made it through a selection process and overlooking those that did not. A list of "top products of the year" shows survivors. It tells you little about how many tried and failed, or what they did differently.

Another bias: lists that are open to anyone to enter are filled by those who chose to enter, often the more active. Comparing yourself with them is fair, but they are not a random sample of the market.

What is being measured

The most important question about any leaderboard statistic is what the underlying measure is.

  • Community votes measure the opinions of those who voted.
  • Visits measure attention, not quality.
  • Growth measures change, favouring small bases.
  • Domain Rating measures link-based strength.
  • Verified data measures what has been checked.

Each is a different lens. Combining them into a single score hides which lens is doing the work. On this site, paid visibility, community votes and verified facts are kept separate and labelled, and votes can never be bought, so a rank by votes means something specific.

How to report your own position honestly

If you want to share your result, give the reader what they need to judge it.

  1. The measure: "by community votes".
  2. The group: "in the productivity category".
  3. The size: "of forty entries".
  4. The period and date: "this week, ending on this date".
  5. The position: "ninth, in the top quartile".
  6. The change, if relevant: "up four places".

An honest sentence might be: "Ninth of forty in productivity by community votes this week, in the top quartile, up four places." It is longer than "Top ten product", and it is true, specific and checkable.

Avoid:

  • Dropping the group: "Top ten" of what?
  • Using a record as a typical result.
  • Rounding up.
  • Using a rank from a different period.
  • Implying that a rank by one measure is a verdict on quality.

Advertising rules in many countries expect claims that readers will take as objective to be backed by evidence, and "top" claims are no exception.

Reading charts of distributions

A few chart types appear on dashboards.

Histogram. Bars show how many entries fall in each range. The shape reveals skew.

Box plot. A box covers the middle half of the data, with a line at the median and whiskers to the extremes, with outliers as dots. It shows the spread at a glance.

Ranked bar chart. Entries in order, with bars for the measure. Start the axis at zero.

Scatter plot. Two measures against each other, one dot per entry.

Look for the median and the spread before you look for your own dot.

A quick checklist for any leaderboard number

  • What is being measured?
  • Who is in the group?
  • How many entries?
  • What period and date?
  • Is it a rank, a percentile or a raw score?
  • Is the typical value a mean or a median?
  • How big are the gaps between places?
  • Is the sample large enough?
  • Is the change shown with a sign and dates?

Asking these takes a minute and prevents most misreadings.

On this site

The leaderboards page shows how entries are ranked, with votes, paid visibility and verified facts kept separate, and the launches page lists what is launching today, this week and soon. The categories page lets you compare within a group, and the developers page describes the free API and MCP server for pulling the underlying numbers into your own sheet.

The short version

A rank is a position in a list, and it needs its list. Percentiles scale position across groups of different sizes, the median is a better typical value than the mean for skewed data, change should be shown with a sign, an arrow and dates, and small samples are noisy. Remember survivorship and what is actually measured, and report your position with the measure, group, size, period and date. Numbers are honest when their context travels with them.

Questions and answers

What is a percentile?
The value below which a given percentage of observations fall. If you are in the 80th percentile, you scored higher than about 80 per cent of the group.
Why use the median instead of the average?
The median is the middle value and resists extreme numbers. The mean can be pulled by a few very large or very small values.
Does a higher rank always mean a better product?
No. A rank depends on what is measured, who is in the group and how scores are calculated. It describes position, not quality.
How should I report my position?
State the measure, the group, the period and the date, for example 'in the top quarter of its category this week by community votes'.
Can small groups give misleading percentiles?
Yes. With few entries, one place can swing a percentile greatly. Show the count as well.

Sources

Related guides

Ask us