Most people hear "interquartile range" and their eyes glaze over. But here's a question that actually matters: if someone tells you the middle of your data sits in a certain spread, how much of the data are they even talking about?
Not obvious, but once you see it — you'll see it everywhere Simple, but easy to overlook..
Turns out, the answer is cleaner than you'd think — and it's one of those things that's obvious once you see it, but easy to mess up if you're guessing. The interquartile range represents the middle 50% of a dataset. That said, not 25%. In practice, not 75%. Fifty.
This is where a lot of people lose the thread.
And yet, I can't count how many blog posts and "quick stats" threads get this backwards Worth knowing..
What Is the Interquartile Range
Look, the interquartile range — you'll see it written as IQR — is just a way to measure spread. But not the spread of everything. Only the spread of the stuff in the middle.
Here's the thing: when you line up all your data points from smallest to largest, you can chop that line into four equal chunks. Those chunks are called quartiles. Even so, the first quartile, or Q1, is the cutoff where the lowest 25% ends. Here's the thing — each chunk holds 25% of your data. The third quartile, Q3, is where the lowest 75% ends — meaning the top 25% starts after it The details matter here..
The interquartile range is the distance between Q1 and Q3. So if Q1 is 10 and Q3 is 20, your IQR is 10. But the number itself isn't the percentage. The percentage of data the IQR covers is the data that lives between Q1 and Q3.
Breaking Down the Four Quarters
Picture a street with 100 houses, numbered in order of size.
- The first 25 houses are the bottom quartile.
- Houses 26 through 50 are the lower-middle.
- Houses 51 through 75 are the upper-middle.
- The last 25 are the top quartile.
The interquartile range looks at houses 26 through 75. That's 50 houses. Half the street.
Why It's Called "Inter"
The "inter" part means between. Consider this: 25 plus 25. In real terms, not the inner quartile, not a quarter of a quartile. And since each quartile boundary marks off 25%, the space between the first and third is two of those blocks stuck together. Between Q1 and Q3. Between the quartiles. Middle 50%.
Why It Matters
Why does this matter? Because most people skip it and then trust the wrong number.
If you're looking at income data, test scores, or response times on a server, the average can lie. In real terms, a few extreme values pull it sideways. But if you think the IQR is only 25% of the data, you'll underestimate how much of the picture you're seeing. The IQR tells you where the typical case actually sits. If you think it's 75%, you'll overestimate and include the outliers you were trying to dodge.
In practice, knowing that the IQR is the middle 50% changes how you read a box plot. The whiskers are the rest, excluding the true extremes. That's why it's the IQR. That box? Half your data is in that box. Real talk — once that clicks, charts stop being decoration and start being useful Not complicated — just consistent. That alone is useful..
And here's a place people get burned: reporting. Someone writes "our median load time is stable, IQR shows low variance" and a manager hears "most users.But it deliberately leaves out the slowest 25% and fastest 25%. And a defensible one, sometimes. Now, " They're right — the IQR does cover most (half) of users. If your worst customers are in that top tail, ignoring them because "that's outside the IQR" is a choice. But know what you're excluding Surprisingly effective..
Most guides skip this. Don't The details matter here..
How It Works
So how do you actually find the percentage — or rather, confirm it — and use the IQR properly? Let's walk through it.
Step One: Order the Data
You can't find quartiles in a pile. Sounds basic. A dataset of 2, 9, 4 becomes 2, 4, 9. Always. Sort the values low to high. You'd be surprised how often this gets skipped in messy spreadsheets.
Step Two: Find the Median
The median is the middle value. It splits the data into two halves. This is Q2, though we don't always call it that. If you have an even count, it's the average of the two middle numbers Nothing fancy..
Step Three: Split Again for Q1 and Q3
Take the lower half (below the median) and find its median. That's Q1. Take the upper half and find its median. That's Q3.
Now, methods vary slightly depending on whether you include the median itself in those halves. Here's the thing — different software does it differently. But regardless of the tiny math tweak, Q1 marks the 25% line and Q3 marks the 75% line Simple as that..
Step Four: Subtract
IQR = Q3 − Q1. Plus, that's your spread. Because of that, the values between those two points? Also, they are 50% of your sorted data. Every value from the 26th percentile to the 75th percentile Turns out it matters..
Step Five: Use the Percentage, Not Just the Number
If Q1 is 100 ms and Q3 is 300 ms, IQR is 200 ms. But the takeaway isn't "200.In practice, " It's "the middle half of all requests took between 100 and 300 ms. " That's the middle 50% speaking.
A Quick Example With Real-ish Numbers
Say you have 12 test scores: 48, 52, 55, 61, 63, 67, 70, 72, 78, 81, 85, 90.
Median is between 67 and 70 → 68.Its median (Q1) is between 55 and 61 → 58. Plus, 5. But lower half: 48, 52, 55, 61, 63, 67. Upper half: 70, 72, 78, 81, 85, 90. Here's the thing — its median (Q3) is between 78 and 81 → 79. 5 Not complicated — just consistent..
IQR = 79.Still, 5 cover six of the twelve students. In practice, 5 − 58 = 21. In real terms, 5. Half. The scores from 58 to 79.Middle 50%.
Common Mistakes
Honestly, this is the part most guides get wrong. They confuse the IQR with a single quartile.
Mistake one: saying the IQR is 25%. Still, no. One quartile is 25%. The range between two quartiles is two of them.
Mistake two: thinking IQR = the median. Now, the median is a point. The IQR is a span. Related, but not the same cloth Simple as that..
Mistake three: assuming IQR contains the "average" case in a weighted sense. It contains 50% of cases, evenly. If your data is lumpy, the middle 50% by count might still miss where the mass sits. Worth knowing.
Mistake four: using IQR to mean "the normal range" in a way that implies 95% or something. Consider this: 5. Think about it: 5 and 97. That's a different tool — usually standard deviation or percentiles like 2.The IQR is specifically the 25-to-75 band.
And a quiet one: people calculate Q1 and Q3 with the wrong method and then argue about the percentage. The percentage doesn't change. In real terms, the boundaries might shift by a hair. But it's always the middle half.
Practical Tips
Here's what actually works when you're dealing with this in real life.
Use the IQR to spot outliers the simple way. That's not a law, just a handy fence. On the flip side, 5×IQR or above Q3 + 1. A common rule: anything below Q1 − 1.On the flip side, 5×IQR is an outlier. But notice — those outliers are outside the middle 50%, in the tails.
When you report the IQR, say the percentage. " Say "the middle 50% falls within a 12-point range.On the flip side, don't just say "IQR is 12. " It lands better and it's honest The details matter here..
If you're comparing two groups, compare IQRs, not just averages. One group might have the same median but
a much wider IQR, meaning its middle half is more scattered and less predictable. Two teams can hit the same typical value yet operate with completely different consistency underneath.
For skewed data, lean on the IQR more than the mean. Averages get dragged by long tails; the middle 50% stays grounded in where most of the action actually is. If your latency graph has a few ugly spikes, the mean will scream while the IQR stays calm and useful Worth keeping that in mind..
And if you're automating this, most stats libraries compute Q1 and Q3 for you—but check which quartile method they use. Some interpolate, some don't. The middle-50% idea survives either way, but your boundary numbers might differ slightly, and someone will eventually ask why your report doesn't match theirs.
And yeah — that's actually more nuanced than it sounds.
Conclusion
The IQR is not a fancy number. Think about it: it's a simple, sturdy way to say: "Ignore the extremes for a second—here's where the bulk of my data actually lives, and that band covers exactly half of everything. " Learn to compute it, report it with its percentage, and use it to compare spread without getting fooled by averages. Do that, and you'll read distributions more honestly than most people who wave around standard deviations they don't fully trust.
