Normal Distribution Calculator: Area, Cut-Offs & Two Groups

A filling line puts an average of 355 mL into each can, with a standard deviation of 2.5 mL. The label says 350 mL. If fill volumes follow a normal distribution, 2.28% of cans come out under the label, about 1 in 44. The average sits 5 mL above the label, yet on a line making 100,000 cans a day that is about 2,275 short cans.

This calculator answers normal-distribution questions in your own units. Type the mean, the standard deviation and a value, and it gives the area below, above, between or outside your values, with the shaded curve and every step. Other tabs find the value for a given percentage (cut-offs like the top 10%), work with the average or total of several values, and compare two groups by how much their curves overlap. It is set to the cans.

Normal Distribution Calculator

Type values in their own units (grams, cm, points). For a continuous variable, below and “at or below” give the same area.

Normal distribution –
The curve (green = your area)
Step by step

    The Area Under the Curve Is the Answer

    A normal distribution is fixed by two numbers. The mean μ puts the peak in place; the standard deviation σ sets how wide the bell is. Every probability is an area under that curve, and the whole area is 1.

    For the cans, three questions and three areas:

    QuestionAreaAnswer
    Under 350 mLto the left of 3500.0228 (2.28%)
    Between 352 and 358 mLbetween the two values0.7699 (76.99%)
    Over 360 mLto the right of 3600.0228 (2.28%)

    The first and last answers match because 350 and 360 sit the same distance, 5 mL or two standard deviations, on either side of the mean. The curve is symmetric, so equal distances cut off equal tails.

    Behind each answer is one conversion, z = (x − μ) ÷ σ, which turns your value into a count of standard deviations. For 350 mL, z = (350 − 355) ÷ 2.5 = −2, and the area to the left of z = −2 is 0.0228. The calculator does this conversion for you, shows the z-score, and labels the curve in millilitres rather than z. For scores you want to compare across different tests, the z-score calculator works from the z side instead.

    Between and Outside Two Values

    “Between” is the left area at the upper value minus the left area at the lower value. For 352 to 358 mL, z runs from −1.2 to +1.2:

    P(352 < X < 358) = 0.8849 − 0.1151 = 0.7699.

    “Outside” is the two tails added together: 1 − 0.7699 = 0.2301 here. Use it for tolerance questions, where both too much and too little is a problem.

    For a continuous measurement, “below 350” and “at or below 350” give the same area, because a single exact value has no width. That is different from counts. With whole numbers, “fewer than 15” and “at most 15” differ by the chance of exactly 15, as the binomial distribution calculator shows when it approximates counts with this curve.

    From a Probability to a Count

    A small percentage turns into a large number once you multiply by how many items there are. The expected count is simply N × p. For 100,000 cans a day from the line above:

    Fill underzShareCans per day
    350 mL−2.02.28%about 2,275
    349 mL−2.40.82%about 820
    348 mL−2.80.26%about 256
    347 mL−3.20.069%about 69

    Each extra millilitre below the label cuts the count to roughly a third, because the tail of the bell thins out faster and faster. That is why a target set a little higher, or a spread made a little tighter, removes most of the problem: the cans that disappear are the ones in the thin end of the tail. The calculator adds a “1 in N” figure next to small results to make this kind of reading quick.

    Density Is Not a Probability

    The height of the curve at a point is called the density. At 355 mL it is 0.1596. That number is easy to misread as “a 16% chance of exactly 355 mL”, and it isn’t one.

    The density is a probability per unit of x: per millilitre here. For a measurement that can take any value, the chance of hitting exactly 355.000… mL is zero. What has a real probability is a range. A fill that reads “355 mL” on a gauge that rounds to the nearest millilitre lies between 354.5 and 355.5, and that range holds 0.1585, which is close to the density × 1 mL width because the curve is nearly flat over that short stretch.

    Densities can also exceed 1. A curve with σ = 0.2 has a peak density of 1.995, which would be an impossible probability but is a perfectly good density. The calculator shows the curve height in its own box, labelled as a density, so it isn’t confused with the area.

    From a Percentage Back to a Value

    The Find a value tab runs the calculation backwards: you give a percentage and get the value that cuts it off.

    • Bottom 1% of cans: 355 − 2.3263 × 2.5 = 349.18 mL. One can in a hundred holds less.
    • Top 5%: 355 + 1.6449 × 2.5 = 359.11 mL.
    • Middle 95%: 355 ± 1.96 × 2.5 = 350.10 to 359.90 mL.

    The same step answers a design question. Suppose no more than 0.5% of cans may fall under 350 mL. The z with 0.5% below it is −2.5758, so either the mean must rise to 350 + 2.5758 × 2.5 = 356.44 mL, or, keeping the mean at 355, the spread must shrink to 5 ÷ 2.5758 = 1.94 mL. The first costs extra liquid in every can; the second costs better equipment. The normal curve puts a number on both choices.

    Averages and Totals Behave Differently From Single Values

    The Average or total tab handles questions about several values at once. Two rules drive it:

    • The average of n independent values has the same mean but a standard deviation of σ ÷ √n, called the standard error.
    • The total of n values has mean n × μ and standard deviation σ × √n.

    Totals spread out with √n, not n, because highs and lows partly cancel.

    Six-packs. One can falls under 350 mL 2.28% of the time. The average of a six-pack has a standard error of 2.5 ÷ √6 = 1.021 mL, so the chance that the six cans average under 352 mL is only 0.16%, and the chance they average under 350 mL is about 1 in 2 million. A buyer checking six-packs would almost never see what a buyer checking single cans sees once every 44 cans.

    An elevator rated for 1,000 kg. If adult weights in a building average 80 kg with a standard deviation of 15 kg, twelve riders weigh 960 kg on average, with a standard deviation of 15 × √12 = 51.96 kg. The chance that a full car of 12 goes over 1,000 kg is 22.1%. For a single rider to exceed 1/12 of the limit (83.3 kg) is far more likely, at 41.2%; the total is steadier than any one person. With 13 riders the average total is already 1,040 kg and the chance of exceeding the rating jumps to 77.0%.

    These rules are exact when the values themselves are normal. When the population is skewed, the average still drifts toward a normal shape as n grows, which is the central limit theorem. How large n has to be depends on how skewed the population is; a fixed number like 30 is a rough guide, not a guarantee. The theorem is about averages and totals. It never makes the individual values normal.

    Changing Units Leaves the Probabilities Alone

    Converting a measurement to other units rescales the mean and the standard deviation together, so every z-score and every area stays the same. In US fluid ounces (1 fl oz = 29.5735 mL) the cans average 12.004 fl oz with an SD of 0.0845 fl oz, the label becomes 11.835 fl oz, and the share under it is still 2.28%.

    Shifting all values by a constant moves the mean but not the SD. Temperature shows both at once: °F = 1.8 × °C + 32, so a mean converts with the full formula while an SD is only multiplied by 1.8. Adding 32 to a standard deviation is a slip that shifts every later answer.

    Two Groups: Overlap and Who Comes Out Ahead

    The Two groups tab compares two normal distributions. Suppose a second filling line, B, averages 353 mL with a standard deviation of 1.0 mL, against line A’s 355 mL and 2.5 mL.

    • P(a random can from A holds more than one from B): 0.771. The difference A − B has mean 2 and standard deviation √(2.5² + 1.0²) = 2.693.
    • Cohen’s d: 2 ÷ √((2.5² + 1.0²) ÷ 2) = 1.05, a large difference in standard-deviation units.
    • Overlap: 45.3% of the area under one curve also lies under the other. The calculator finds the points where the two curves cross and adds up the lower curve on each stretch. Because B’s curve is narrower and taller, the curves cross twice, once on each side of B’s peak.

    Now the twist. Line B has the lower average, yet only 0.13% of its cans fall under 350 mL, against line A’s 2.28%: about 17 times fewer short cans. B’s tight spread matters more than A’s higher mean. Comparing means alone would pick the wrong line.

    Is Your Data Normal Enough?

    Every answer above assumes the bell shape. Real data can fail it in three ways:

    • Skew. Waiting times, incomes and house prices have long right tails. A normal curve fitted to them puts probability on negative values and understates large ones. When μ − 3σ falls below zero for something that can’t be negative, the calculator says so in a tip.
    • Heavy tails. Some data are bell-shaped in the middle but produce extreme values more often than the curve predicts. The normal model gives 3.2 × 10⁻⁵ for being 4 SDs above the mean and 2.9 × 10⁻⁷ for 5 SDs. Figures that small depend entirely on the shape far out in the tail, so treat them as rough unless the data have been checked that far.
    • Bounds and lumps. Test scores capped at 100, or values rounded to a few levels, don’t follow a smooth curve.

    Two quick checks help. Plot a histogram and look for a single symmetric peak. Then compare shares: roughly 68% of values should lie within one standard deviation of the mean and 95% within two, as the standard deviation calculator reports for your data. A normal Q-Q plot, which lines up sorted data against normal cut-offs, shows departures in the tails more clearly than a histogram.

    Sample Estimates Instead of the True Mean and SD

    The calculator treats μ and σ as known. When they come from a sample instead, the estimates carry their own error. With only 10 measurements, the range expected to hold the next single value is wider than μ ± 1.96σ: it uses the t-distribution and an extra term for the uncertain mean, x̄ ± 2.262 × s × √(1 + 1/10) = x̄ ± 2.373 s, about 21% wider. With large samples the difference fades. For intervals around a mean estimated from data, the confidence interval calculator handles the t step.

    Normal Probabilities in Excel and Google Sheets

    Both spreadsheets use the same two functions, with your mean and SD as arguments:

    What you wantFormulaCans result
    Area below x=NORM.DIST(350, 355, 2.5, TRUE)0.02275
    Area above x=1 - NORM.DIST(360, 355, 2.5, TRUE)0.02275
    Area between=NORM.DIST(358, 355, 2.5, TRUE) - NORM.DIST(352, 355, 2.5, TRUE)0.76986
    Curve height (density)=NORM.DIST(355, 355, 2.5, FALSE)0.15958
    Value with 1% below=NORM.INV(0.01, 355, 2.5)349.184

    The last argument of NORM.DIST is the switch that trips people up: TRUE gives an area, FALSE gives the density, which is not a probability. Microsoft’s own example uses x = 42, mean 40 and SD 1.5, giving 0.9087888 with TRUE and 0.10934 with FALSE; the calculator gives the same.

    Five Questions on the Filling Line, With Answers

    Each uses μ = 355 mL and σ = 2.5 mL unless stated. Try them in the calculator first.

    1. What share of cans hold more than 357 mL? z = 0.8, area to the right 0.2119 (21.19%).
    2. What share fall between 350 and 356 mL? 0.6554 − 0.0228 = 0.6327.
    3. Above what volume is the fullest 10%? 355 + 1.2816 × 2.5 = 358.20 mL.
    4. What is the chance that a 24-can case averages under 354 mL? Standard error 2.5 ÷ √24 = 0.510, z = −1.96, 0.0250.
    5. A line with mean 356 and SD 2.0 replaces the old one. How many cans per 10,000 now fall under 350 mL? z = −3, about 13.5 per 10,000 (0.135%), down from 228.

    The answers show the two levers again: moving the mean and tightening the spread both shrink the tail, and averaging many cans shrinks it further still.