Contact
12 programs
Explore the site

Zoho Sheet NORMINV: Normal Distribution Percentiles and Cutoffs

Statistical Intermediate Zoho Sheet

Find normal-distribution percentile cutoffs in Zoho Sheet with official syntax, checked examples, round-trip tests and limitations.

Verification status: Official documentation referenced; example outcomes independently reasoned or arithmetically checked where stated. Not executed in Zoho Sheet. How formula examples are checked

Quick answer: Use =NORMINV(0.95;100;15) to find the value below which 95% of an assumed normal distribution with mean 100 and standard deviation 15 lies. The independently calculated expected cutoff is 124.672804404. NORMINV returns a value in the units of the original distribution, not a probability or a density. The formula and argument order come from Zoho’s official NORMINV help (checked 2026-10-12). These expected outputs were not executed in Zoho Sheet.

NORMINV is an inverse cumulative normal-distribution function, sometimes called a quantile or percent-point function. Supply a cumulative probability and a normal model; it returns the corresponding threshold. To go in the other direction—from a threshold to a cumulative probability—use NORMDIST with TRUE as its fourth argument. Do not use the NORMDIST density mode for this inverse relationship.

Official Zoho syntax, argument definitions and required inputs

The Zoho Sheet English help page documents the following signature (checked 2026-10-12):

NORMINV(p; μ; σ)
Order Argument Required? Meaning Example
1 p Yes Cumulative probability, expressed as a fraction between 0 and 1. 0.95, 95% or A2
2 μ (mu) Yes Mean of the normal distribution in measurement units. 100 or $E$2
3 σ (sigma) Yes Standard deviation of the assumed normal model, in the same units as μ. 15 or $E$3

Input interpretation: 0.95 is 95%; 95 is not a valid probability input. Zoho’s reference supplies three mandatory arguments and no optional default. This is not a percentile ranking among observed samples; it is a mathematical model cutoff. Zoho’s English help uses semicolons between formula arguments. Your actual localized worksheet separator, imported-file representation and numerical settings were not natively tested.

Valid model: A mathematically ordinary normal distribution needs σ > 0, and a finite quantile requires 0 < p < 1. Zoho explicitly lists #NUM! for out-of-range p and negative σ, but its displayed wording does not separately confirm the behavior for σ exactly zero. See the boundary-case section before relying on an untested spreadsheet result.

Why the inverse works

If Φ is the standard-normal cumulative distribution function, then the normal distribution with mean μ and positive standard deviation σ satisfies:

NORMINV(p; μ; σ) = μ + σ × Φ⁻¹(p)

This equation is a mathematical identity, not a claim that Zoho uses a particular internal algorithm. The inverse depends on the probability being cumulative from the left. Thus p=0.95 is the 95th percentile (5% modeled above it); a 95% central interval requires two tail quantiles, not just a single 95th-percentile cutoff.

Worked example 1 — a 95th-percentile threshold

Suppose your distribution model for an output measurement is mean 100 units, standard deviation 15 units. Create these cells:

Cell Label (column A) Input (column B)
A1/B1 Setting / Value (headers)
A2/B2 Cumulative probability 0.95
A3/B3 Mean, μ 100
A4/B4 Standard deviation, σ 15
  1. Enter those three numbers into B2:B4. Choose a blank output cell such as B6.
  2. In B6, enter =NORMINV(B2;B3;B4). Independent expected result: 124.672804404. At this cutoff, the normal model places 95% of values to the left and 5% above.
  3. Check the logic in B7 with =NORMDIST(B6;B3;B4;TRUE). Independent expected result: approximately 0.95. Because numerical inverse calculations can have floating-point error, compare to 0.95 to appropriate precision, not character-for-character equality.
  4. Verify that entering 0.5 into B2 changes the cutoff to 100, the mean. Changing B2 to 0.05 yields approximately 75.327195596. These two values are symmetric around 100 under a normal model.
  5. As a basic input check, do not replace p=0.95 with p=95. The latter is outside the allowed probability interval.

Verification: These are independently calculated results based on normal-distribution mathematics, not actual Zoho formula executions. The specific inverse behavior at invalid inputs has not been tested in a Zoho account.

Worked example 2 — build a percentile table

You need thresholds for several service levels or design bands. Enter mean 100 into E2 and σ 15 into E3. Create a new table:

Row A: Cumulative probability Formula to enter in B2 and copy down Independently expected cutoff
2 0.05 =NORMINV(A2;$E$2;$E$3) 75.327195596
3 0.10 Fill the B2 formula down 80.776726517
4 0.50 Fill the B2 formula down 100.000000000
5 0.90 Fill the B2 formula down 119.223273483
6 0.95 Fill the B2 formula down 124.672804404

Here A2:A6 hold the probabilities; enter the formula in B2 and copy down to B6. The dollar signs on $E$2 and $E$3 stop the mean and standard deviation from sliding when filling the formula. Using the same settings in each row makes the percentage-to-threshold relationship easy to audit.

These quantiles are model predictions, not observed 5th/10th/50th/90th/95th percentile values in a dataset. For an empirical percentile of your actual records, consult a percentile function only after separately checking its data-inclusion and interpolation rules; it solves a different problem.

Advanced application — calculate a central 90% interval

Suppose a designer wants a central interval containing 90% of a normal-model population rather than a one-sided 90th percentile. Each tail must contain 5% of probability. In a separate pair of cells:

Lower endpoint: =NORMINV(0.05;100;15)
Upper endpoint: =NORMINV(0.95;100;15)

Expected endpoints are 75.327195596 and 124.672804404. The normal-model interval is [75.3272, 124.6728], with 5% below and 5% above. It would be wrong to present [NORMINV(0.1;100;15), NORMINV(0.9;100;15)] as central 90%; that interval actually has 10% in each tail, or 80% in the center.

To validate the interval’s logic, take the two tail cumulative probabilities using the complementary NORMDIST guide:

=NORMDIST(124.672804404;100;15;TRUE)
 -NORMDIST(75.327195596;100;15;TRUE)

Independent mathematical expectation: approximately 0.90 (more precisely, near 0.900000000 when using unrounded underlying quantiles). Small deviations are expected if you round the endpoint inputs to nine decimals. This two-function design is handy for auditable service-level thresholds, measurement tolerances, or distribution illustrations. It is not a confidence interval for an estimated mean. A confidence interval for an unknown parameter needs a different statistical procedure and assumptions.

Advanced application — score business-risk thresholds

A logistics forecast models delivery duration as normal with mean 48 hours and standard deviation 6 hours. To find a time target that a modeled 90% of deliveries meet, use:

=NORMINV(0.90;48;6)

Expected result: 55.689309393 hours. If the contract requires that 95% of deliveries meet a target, use =NORMINV(0.95;48;6) for a mathematical expectation of 57.869121762 hours. Do not round the thresholds downward without considering the operational consequences. These are planning estimates conditional on a normal model; actual deliveries can have skew, holidays, weather disruptions, or correlations, so verify the model against field data before treating the results as promises.

The reverse question—What proportion of modeled deliveries take no more than 54 hours?—belongs to NORMDIST(54;48;6;TRUE), with independent expected result 0.841344746069. Knowing which direction you are calculating helps avoid mixing up percentiles and probabilities.

Errors, boundaries and practical troubleshooting

Scenario Evidence or mathematical implication What to do
#NUM! with p outside the unit interval Zoho documents a numerical error when p is not between 0 and 1. Enter a fraction such as 0.95, not 95; do not force out-of-range values into the model.
p exactly 0 or 1 A finite mathematical normal quantile does not exist at either endpoint (tails extend indefinitely). Zoho’s wording does not fully distinguish endpoint inclusion in the help text. Treat these as invalid for finite cutoffs. The exact native Zoho error remains untested; don’t quote Excel as Zoho evidence.
#NUM! with negative σ Zoho specifically documents σ below zero as an error. Correct the sign or input formula. Standard deviation cannot be negative.
σ = 0 Standard normal rescaling assumes positive σ. Zoho’s help mentions σ < 0 but does not separately specify the σ=0 response. Consider the distribution degenerate and stop; native result/error is unknown pending a test.
#VALUE! Zoho lists incorrect data types among possible causes. Check whether "0.95" is text, whether values were imported with mismatched decimal separators, and whether source cells truly contain numbers.
#NAME! Zoho’s general error table associates this with misspelled/unrecognized functions and identifiers. Spell NORMINV exactly. Do not assume NORM.INV is a supported Zoho alias because Excel uses it.
#REF! Zoho documents missing/invalid cell references. Verify the mean, σ and p cells were not deleted or moved.
Unexpected apparent exact equality from an inverse/CDF round trip Floating-point results are approximate; rounding can hide error. Compare within a tolerance suitable for your use case, not as exact text.
Very extreme p, such as 1-1E-15 Extreme tails are numerically sensitive; the Zoho reference does not document tested precision guarantees at those extremes. Benchmark native output against a trustworthy numerical reference before using it in high-stakes analyses.

Zoho’s NORMINV page also lists generic #N/A! explanations, but it does not establish a reproducible NORMINV-specific input that triggers that error. Do not invent one. Zoho NORMINV error section (checked 2026-10-12).

Function choice: NORMINV, NORMDIST and standardized alternatives

  • NORMINV turns a cumulative probability p into a threshold x for a normal model with specified μ and σ.
  • NORMDIST with TRUE turns a threshold x into a cumulative probability p. Its FALSE mode gives a density instead, so it is not the inverse pairing.
  • NORMSINV works with the standard normal distribution, fixed mean 0 and standard deviation 1; its outputs are z-scores. Zoho NORMSINV (checked 2026-10-12).
  • NORMSDIST returns standard-normal cumulative probability for a given z-score. Zoho NORMSDIST (checked 2026-10-12).

Use these distinctions to avoid accidentally interpreting a standardized z-score as a quantity in hours, dollars or centimeters.

Documented Excel and Google Sheets differences

Product Documented naming and syntax What can safely be concluded
Zoho Sheet NORMINV(p; μ; σ) Current official English Zoho Help shows these three parameters and semicolon separators; it does not certify all locale and platform variants.
Microsoft Excel Legacy NORMINV(probability,mean,standard_dev); current Microsoft guidance recommends NORM.INV Microsoft’s legacy help specifies p ≤ 0 or p ≥ 1 and σ ≤ 0 as errors, a more explicit boundary statement than Zoho’s article. Excel’s support/precision profile is not evidence of native Zoho behavior.
Google Sheets NORMINV(x, mean, standard_deviation); its function list also lists NORM.INV as a name referring to NORMINV Google’s English reference uses commas; do not assume Zoho accepts NORM.INV or uses Google’s parsing rules.

Sources checked 2026-10-12: Microsoft NORMINV, Google NORMINV, Google function list. These links verify documented naming and boundary rules for those products only; do not infer software equivalence in rare numerical cases.

The AVERAGE guide explains a mean from data. See NORMDIST for cumulative and density calculations. The direct official references for Zoho NORMDIST, Zoho NORMSINV, and Zoho NORMSDIST were checked 2026-10-12.

Evidence, platform and validation status

Verified documentation: The official Zoho NORMINV function reference was directly read on 2026-10-12, as were related NORMDIST/NORMSINV/NORMSDIST references and separate Microsoft/Google documentation for comparisons. The Zoho What’s New page was consulted for release context but provided no verified first-supported version, account-tier restriction, mobile/offline calculation guarantee, or cross-edition parity statement for this specific function. None is asserted here. Independent calculations: All expected normal quantiles, reverse CDF checks and interval endpoints were evaluated mathematically with Python/SciPy; they are not screenshots or actual app results. Native Zoho Sheet executions: 0.