A percentage is only part of the story
A post says that 80% of viewers preferred one video. That sounds decisive, but 80% of what? If five people responded to a poll, the result describes four responses. If thousands responded, it describes a much larger group. Either way, you still need to know who saw the poll and who chose to answer. The percentage tells you a share; the denominator tells you what that share is a share of.
The denominator is the total used to calculate a fraction. In a claim that 40% of posts received comments, it might be all posts on an account, posts published in a particular month, or only posts selected for a report. Those choices can produce different impressions without changing the arithmetic. When a number feels striking, pause before interpreting it and ask: What was counted in the total?
Look for the group behind the number
Online claims often name a broad group while measuring a narrower one. A headline might say that readers dislike a new feature, while the figure comes from people who answered a poll posted beneath an article about that feature. Those respondents are real readers, but they are not necessarily representative of everyone who uses it. People with strong opinions may be more likely to respond, and people who never encountered the poll cannot be counted.
Check whether the claim concerns people, responses, posts, views or purchases. These are not interchangeable denominators. One person can generate several views, and one enthusiastic commenter can leave multiple replies. Also look for exclusions: Did the calculation include private accounts, deleted posts, repeat visitors or only those who completed a survey? If the total is unclear, a careful reading is not that the claim must be false. It is that its reach is uncertain.
Separate relative change from absolute change
Risk claims can be especially confusing because a large relative change may describe a small absolute change. Imagine a hypothetical setting in which two out of 1,000 people experience an outcome. If a second group has four out of 1,000, the rate has doubled: a 100% relative increase. The absolute difference is two additional people per 1,000, or 0.2 percentage points. Both descriptions are mathematically correct, but they answer different questions.
To make sense of a risk claim, look for the starting rate, the new rate and the population each rate describes. A claim that something cuts risk in half is incomplete if it does not say whether the original risk was common or rare. It also matters whether the groups being compared are similar and whether the time period is the same. A comparison over one week should not be casually read as a prediction over a lifetime.
Watch for totals that quietly change
A chart can make two percentages appear comparable even when their denominators differ. Suppose one creator says 20% of followers saved a post and another says 20% of viewers saved one. The first figure uses followers as its total; the second uses viewers. Neither figure alone establishes which post inspired more saving. To compare them, you would need a common basis, along with a clear account of how saves and views were measured.
The same issue appears in before-and-after claims. An account may report that a format went from 10% to 15% of its posts, then call that a five-percentage-point increase. That is different from a 5% relative increase: compared with the starting share, it is a 50% increase. Ask whether a change is expressed in percentage points or as a percentage of the original figure. Then check whether the underlying total changed too. A rising share can occur even when the number in that category stays the same, if the overall total falls.
Use a quick denominator check before sharing
You do not need to reconstruct every calculation you encounter. Start with four questions: What is the number out of? How many people or items does that total represent? Who or what was left out? What is the comparison group? For a risk claim, add two more: What was the starting rate, and over what time period was the outcome measured? If a post does not provide those details, look at the material it links to, if any, before repeating its conclusion.
If the denominator remains unavailable, adjust your language to match what you know. Instead of saying that most users prefer a feature, you might say that most respondents to a particular poll preferred it. Instead of repeating that a risk doubled without context, note that the starting rate was not given. This habit does not make every number useless. It makes the boundary between a measured result and a broader claim easier to see.
