Exponential Distribution Calculator
What Is the Exponential Distribution?
The exponential distribution is a continuous probability distribution that models the time between events in a Poisson process — situations where events occur at a constant average rate independently of one another. A single parameter λ (the rate) fully describes the distribution.
Common applications include the time until a machine breaks down, the interval between customer arrivals, the duration of a phone call, and the time between network requests. The exponential distribution is the only continuous distribution with the memoryless property: knowing how long you have already waited does not change the probability of how much longer you must wait.
Quick definition: X ~ Exp(λ) means X is a non-negative continuous random variable with rate parameter λ > 0, mean 1/λ, and cumulative distribution function F(x) = 1 − e−λx for x ≥ 0. F(x) gives the probability that the waiting time does not exceed x.
Critical: Rate vs. Scale Parameterization
Two conventions exist for the exponential distribution. Confusing them is the most common source of incorrect calculations. Check your software's documentation before computing any probability.
| Parameterization | Symbol | Meaning | CDF | Mean | |
|---|---|---|---|---|---|
| Rate (most common) | λ | Events per unit time | λe−λx | 1 − e−λx | 1/λ |
| Scale | θ = 1/λ | Mean waiting time | (1/θ)e−x/θ | 1 − e−x/θ | θ |
Software note: R uses rate (rexp, dexp, pexp); Python's scipy.stats.expon uses scale = 1/λ by default. Excel's EXPON.DIST uses the rate λ. SAS and SPSS use scale. Always verify which convention applies before entering your parameter.
Exponential Distribution CDF Table
The table below gives F(x) = P(X ≤ x) = 1 − e−λx for five common rate values. All values use the rate parameterization. Click any cell to load it into the calculator. Switch tabs to view survival probabilities S(x) = P(X > x) = e−λx.
All values computed from F(x) = 1 − e−λx (rate parameterization). Values rounded to 4 decimal places. For P(X > x), subtract from 1.0000 or switch to the Survival tab. x = time or value; λ = rate parameter (events per unit).
How to Use the Exponential Distribution Table
Unlike a standard normal or t-distribution table, an exponential CDF table is a convenience tool rather than a necessity. The CDF has a closed-form expression that you can compute directly. The table becomes useful when you need quick lookups for common parameter values or want to verify calculator results.
Step 1 — Identify λ
λ is the rate at which events occur. If the problem states an average time between events (the mean μ), convert using λ = 1/μ. Example: if the average inter-arrival time is 20 minutes, λ = 1/20 = 0.05 per minute, or λ = 3 per hour after converting units.
Step 2 — Confirm the Parameterization
Verify that you have λ (rate), not θ (scale). The table above uses λ. If your source gives θ, compute λ = 1/θ. This single check prevents the most frequent error in exponential distribution problems.
Step 3 — Identify x and Check Units
x must be in the same units as 1/λ. If λ = 2 failures per hour and you want a probability for 45 minutes, convert to x = 0.75 hours before looking up or calculating.
Step 4 — Read or Compute F(x)
Locate the column for your λ and the row for your x. Read the CDF value F(x) = P(X ≤ x). If your exact λ or x is not in the table, use the formula directly or the calculator above.
Step 5 — Derive the Probability You Need
P(X > x) = 1 − F(x) = e−λx
P(a < X ≤ b) = F(b) − F(a) = e−λa − e−λb
P(X = x) = 0 for any specific x (continuous distribution)
Step 6 — Interpret the Result
State the probability in terms of the original problem. "The probability that the next customer arrives within 3 minutes is 0.7769" (for λ = 0.5 per minute, x = 3). Always include units and direction in your interpretation.
Exponential Distribution Formulas
All formulas below use the rate parameterization with λ > 0. For the scale parameterization, substitute θ = 1/λ.
Probability Density Function (PDF)
The PDF gives the relative likelihood of x — not a probability itself. For a continuous random variable, P(X = x) = 0. Probabilities require integration over an interval, which is what the CDF provides.
Cumulative Distribution Function (CDF)
F(x) = P(X ≤ x) is the area under the PDF from 0 to x. As x → ∞, F(x) → 1. At x = 0, F(0) = 0. At x = 1/λ (the mean), F(1/λ) = 1 − e−1 ≈ 0.6321.
Survival Function
S(x) = 1 − F(x). In reliability engineering, this is the probability a component survives beyond time x. In queueing theory, it gives the probability that service or waiting exceeds x.
Key Properties
Variance = 1/λ²
Std Dev = 1/λ
Median = ln(2)/λ ≈ 0.6931/λ
Mode = 0
Skewness = 2 (always)
Exponential CDF vs PDF
Students often confuse the PDF and CDF. The table below makes the distinction explicit.
| Feature | PDF — f(x) | CDF — F(x) |
|---|---|---|
| Full name | Probability density function | Cumulative distribution function |
| Formula | λe−λx | 1 − e−λx |
| Output range | 0 to ∞ (can exceed 1) | 0 to 1 (always a probability) |
| Represents a probability directly? | No | Yes |
| At x = 0 (λ = 1) | f(0) = 1 | F(0) = 0 |
| Shape | Decreasing from λ at x=0 | Increasing from 0, approaching 1 |
| Typical use | Density, likelihood, fitting | Probability lookup, quantiles |
Worked Examples
Each example below follows the same six-step process: identify λ, confirm units, state x, choose the correct probability formula, calculate, and interpret.
Example 1 — Machine Reliability
Scenario: A machine has an average failure rate of λ = 0.5 failures per hour. What is the probability that it fails within the first 2 hours?
| Step | Action | Result |
|---|---|---|
| 1 | Identify λ | λ = 0.5 failures/hour |
| 2 | Identify x | x = 2 hours (same units as 1/λ) |
| 3 | State the probability | P(X ≤ 2) = F(2) |
| 4 | Apply formula | F(2) = 1 − e−0.5×2 = 1 − e−1 |
| 5 | Calculate | 1 − 0.3679 = 0.6321 |
Interpretation: There is a 63.2% probability the machine fails within the first 2 hours. The mean time to failure is 1/0.5 = 2 hours, so this result makes sense: by the mean, roughly 63% of machines have already failed.
Example 2 — Customer Arrivals
Scenario: Customers arrive at a service counter at an average rate of 3 per hour. What is the probability that the next customer arrives more than 30 minutes (0.5 hours) after the last one?
Given: λ = 3 per hour, x = 0.5 hours. We need P(X > 0.5) = S(0.5).
Interpretation: There is a 22.3% probability that the next customer takes more than 30 minutes to arrive. The mean inter-arrival time is 1/3 hour = 20 minutes.
Example 3 — Interval Probability
Scenario: Server requests arrive at rate λ = 2 per second. Find the probability that the next request arrives between 0.5 and 1.5 seconds.
We need: P(0.5 < X ≤ 1.5) = F(1.5) − F(0.5)
F(0.5) = 1 − e−2×0.5 = 1 − e−1 ≈ 0.6321
P(0.5 < X ≤ 1.5) = 0.9502 − 0.6321 = 0.3181
Interpretation: There is a 31.8% probability the next request arrives between 0.5 and 1.5 seconds. The mean inter-request time is 1/2 = 0.5 seconds.
Example 4 — Unit Conversion
Scenario: A hospital receives emergency calls at an average rate of 4 per hour. What is the probability that the next call arrives within 10 minutes?
Unit conversion first: λ = 4/hour. Convert x to hours: x = 10 minutes = 10/60 = 1/6 hour ≈ 0.1667 hours.
Interpretation: There is approximately a 48.7% probability that the next emergency call arrives within 10 minutes. The mean waiting time between calls is 1/4 hour = 15 minutes.
Example 5 — Memoryless Property
Scenario: A light bulb lasts an exponentially distributed time with λ = 0.1 failures per hour. Given that it has already lasted 100 hours, what is the probability it lasts at least 50 more hours?
By the memoryless property: P(X > 150 | X > 100) = P(X > 50) = e−λ×50
Interpretation: There is only a 0.67% probability of lasting at least 50 more hours. Critically, the fact that it already survived 100 hours is irrelevant — the calculation uses x = 50, not x = 150. This is the memoryless property in action.
The Memoryless Property
The memoryless property is the defining characteristic of the exponential distribution among all continuous distributions. Formally:
In practical terms: if you have waited s time units and the event has not yet occurred, the remaining waiting time has the same distribution as if you were starting fresh from time 0. Past waiting time carries no information about future waiting time.
Where It Applies
Electronic components (when failure is due to random external shock rather than wear), radioactive decay, queue service times in Markovian models, and inter-arrival times in Poisson processes.
Where It Does Not Apply
Human lifetimes, mechanical wear, and anything with an increasing failure rate over time. For those situations, the Weibull distribution is more appropriate.
Discrete Counterpart
The geometric distribution is the discrete analog of the exponential distribution and also has the memoryless property. Just as the exponential models waiting time in continuous time, the geometric counts the number of trials until the first success.
The Exponential Distribution and Poisson Process
The connection between the exponential distribution and the Poisson process is one of the most useful relationships in probability. The two distributions share the same rate parameter λ, but they describe different quantities:
| Distribution | Type | What it models | Parameter | Mean |
|---|---|---|---|---|
| Poisson(λ) | Discrete | Number of events in a fixed time interval | λ (events/time) | λ |
| Exponential(λ) | Continuous | Time between consecutive events | λ (events/time) | 1/λ |
Key relationship: If calls arrive according to a Poisson process at rate λ = 5 per hour, then the number of calls in any 1-hour window follows Poisson(5), and the time between consecutive calls follows Exp(5) with a mean of 1/5 = 12 minutes. The Poisson counts events; the exponential measures the gaps between them.
Exponential vs Related Distributions
Understanding how the exponential distribution relates to similar distributions helps in selecting the right model for a given application.
vs Geometric Distribution
The geometric distribution is the discrete counterpart: it models the number of Bernoulli trials until the first success, while the exponential models continuous waiting time. Both are memoryless. The exponential is the continuous-time limit of a geometric distribution as the trial length shrinks to zero.
vs Gamma Distribution
The exponential is a special case of the gamma distribution with shape parameter k = 1. A Gamma(k, λ) distribution models the waiting time until the k-th event in a Poisson process. When k = 1, it reduces to Exp(λ).
vs Weibull Distribution
The Weibull generalizes the exponential by adding a shape parameter. When the shape parameter equals 1, the Weibull reduces to the exponential. Shape > 1 gives an increasing hazard rate (aging); shape < 1 gives a decreasing hazard rate (infant mortality). Use Weibull when memorylessness is not a reasonable assumption.
vs Normal Distribution
The normal distribution is symmetric and defined on (−∞, +∞). The exponential is right-skewed and defined on [0, ∞). The two are fundamentally different in shape. For large samples of exponential data, the sample mean approaches normality through the Central Limit Theorem, but individual observations are not normally distributed.
Common Mistakes to Avoid
Each of the following errors appears regularly in homework, exams, and applied analyses. Recognizing them before they occur saves significant time.
Confusing rate and scale
If the mean waiting time is 5 minutes, then θ = 5 (scale) and λ = 1/5 = 0.2 (rate). Plugging θ = 5 into a formula that expects λ = 0.2 gives a completely wrong answer. Always identify which parameterization a formula or software function requires.
Unit mismatch
λ = 3 per hour and x = 20 minutes cannot be used together directly. Convert x to 20/60 = 1/3 hour first, then compute F(1/3) = 1 − e−3×(1/3) = 1 − e−1. Mixing units is one of the most common exam errors.
Computing P(X = x) for a continuous variable
For any continuous distribution, P(X = x) = 0 exactly. The probability that a wait time equals exactly 3.000... hours is zero. Only intervals have non-zero probability. Use the CDF: P(X ≤ x) rather than the PDF value f(x).
Forgetting the complement for P(X > x)
When asked for "the probability that the waiting time exceeds x," the answer is P(X > x) = e−λx, not F(x). The survival function is already simple — there is no need to compute 1 minus anything. Just apply the formula directly.
Misapplying the memoryless property
Given that X > s, the conditional probability P(X > s + t | X > s) equals P(X > t) — using t, not s + t. The remaining time restarts from zero. A common mistake is computing e−λ(s+t) instead of e−λt.
Using a negative x value
The exponential distribution is defined for x ≥ 0 only. F(x) = 0 and f(x) = 0 for all x < 0. If you get a negative x from your problem setup, check your units or problem statement.
Software Implementation
Different software packages use different parameterizations. The examples below compute F(2) = P(X ≤ 2) for an exponential distribution with mean = 1 (λ = 1, θ = 1).
Excel / Google Sheets
Syntax: EXPON.DIST(x, lambda, cumulative). Set cumulative = TRUE for CDF F(x); FALSE for PDF f(x). The second argument is λ (rate), not θ (scale).
Python — SciPy
expon.cdf(2, scale=1)
scipy.stats.expon uses the scale parameter θ = 1/λ. For λ = 2, set scale = 1/2 = 0.5. The loc parameter shifts the distribution; leave it at 0 for a standard exponential.
R
R's pexp() uses the rate parameter λ directly. For the survival function, use lower.tail = FALSE: pexp(2, rate = 1, lower.tail = FALSE). rexp(n, rate) generates random samples.
Python — Direct Formula
x, lam = 2, 1
cdf = 1 - math.exp(-lam * x)
You can always compute the CDF directly from the closed-form formula. This avoids any ambiguity about parameterization conventions in statistical packages.
Entity and Formula Glossary
The table below defines every term and symbol used on this page.
| Term / Symbol | Definition | Formula / Notation | Related Concept |
|---|---|---|---|
| λ | Rate parameter; events per unit time | λ > 0 | Poisson process rate |
| θ | Scale parameter; mean waiting time | θ = 1/λ | Gamma distribution scale |
| F(x) | Cumulative distribution function (CDF) | 1 − e−λx | P(X ≤ x) |
| f(x) | Probability density function (PDF) | λe−λx | Not a probability directly |
| S(x) | Survival function; reliability function | e−λx | P(X > x) = 1 − F(x) |
| Mean | Expected value of X | E[X] = 1/λ = θ | Average waiting time |
| Variance | Spread of the distribution | Var(X) = 1/λ² | SD = 1/λ (equals mean) |
| Median | Value at which F(x) = 0.5 | ln(2)/λ ≈ 0.6931/λ | Always less than mean |
| Hazard Rate | Instantaneous failure rate given survival to x | h(x) = λ (constant) | Memoryless property |
| Poisson Process | Process where events occur at constant rate λ | N(t) ~ Poisson(λt) | Inter-arrival time ~ Exp(λ) |
Frequently Asked Questions
Is there an exponential distribution table like a z-table?
Unlike the normal distribution, the exponential CDF has a closed-form expression: F(x) = 1 − e−λx. Printed tables exist, but they are less commonly used in practice because you can calculate exact values with a basic calculator. This page provides a lookup table for common λ values along with an interactive calculator.
Why does F(1/λ) ≈ 0.6321 regardless of λ?
At x = 1/λ (the mean), F(1/λ) = 1 − e−1 ≈ 0.6321 for any value of λ. This is because the product λ × (1/λ) = 1, so the formula reduces to 1 − e−1 every time. In other words, roughly 63.2% of observations fall below the mean for any exponential distribution.
How do I find a quantile or percentile of the exponential distribution?
Set F(x) = p and solve for x. Since F(x) = 1 − e−λx, solving gives x = −ln(1 − p)/λ. For the 90th percentile with λ = 2: x = −ln(0.10)/2 = 2.303/2 ≈ 1.151. In R: qexp(0.9, rate = 2). In Python: expon.ppf(0.9, scale = 1/2).
Can the exponential distribution be used when the hazard rate increases over time?
No. The exponential distribution has a constant hazard rate h(x) = λ, which means the risk of the event at any moment does not depend on how long the process has been running. For systems that age (increasing hazard) or improve with use (decreasing hazard), the Weibull distribution is more appropriate.
What is the relationship between the exponential and Erlang distributions?
The Erlang(k, λ) distribution is the sum of k independent Exp(λ) random variables. It models the waiting time until the k-th event in a Poisson process. When k = 1, Erlang reduces to Exponential. The Erlang distribution is a special case of the gamma distribution with an integer shape parameter.
How do I test whether data follows an exponential distribution?
Common methods include the Kolmogorov-Smirnov test, the exponential Q-Q plot (plotting order statistics against theoretical exponential quantiles), or checking whether the mean approximately equals the standard deviation (a property unique to the exponential distribution). The mean and standard deviation are both 1/λ for exponential data.
What happens to the exponential distribution as λ increases?
As λ increases, events occur more frequently: the mean waiting time 1/λ decreases, the distribution becomes more concentrated near zero, and the PDF decays more steeply. As λ decreases toward zero, the distribution spreads out and the mean grows large. The shape remains the same (right-skewed, mode at zero) regardless of λ.
Can the exponential distribution model waiting times in minutes and hours simultaneously?
Yes, but you must keep units consistent. If λ = 3 per hour, the mean is 20 minutes. If you want a probability in minutes, convert: λ = 3/60 = 0.05 per minute. Then F(x) = 1 − e−0.05x where x is in minutes. The probability is the same either way; only the numerical λ and x change with the unit choice.
References and Further Reading
NIST/SEMATECH e-Handbook of Statistical Methods. Section 1.3.6.7: Exponential Distribution. National Institute of Standards and Technology. itl.nist.gov — Authoritative U.S. government reference for the exponential distribution, including parameter estimation, goodness-of-fit, and applications in reliability.
Casella, G. & Berger, R. L. (2002). Statistical Inference (2nd ed.). Duxbury. Chapter 3 covers the exponential family and includes derivations of the exponential distribution CDF, moments, and memoryless property.
Penn State STAT 414: Probability Theory. Lesson 15: Exponential Distribution. Pennsylvania State University. online.stat.psu.edu — Open-access university course notes covering the derivation, properties, and memoryless property of the exponential distribution.
Ross, S. M. (2019). Introduction to Probability Models (12th ed.). Academic Press. Chapter 5 covers the exponential distribution in the context of continuous-time Markov chains, Poisson processes, and queueing theory.
SciPy Documentation: scipy.stats.expon. docs.scipy.org — Official documentation for the exponential distribution in Python's SciPy library, including the scale parameterization convention used by scipy.stats.
Related Statistical Tables & Resources
Why the Exponential Distribution Is Unique Among Continuous Distributions
Constant Hazard Rate
The hazard rate h(x) = f(x)/S(x) = λ for all x. This is constant — it does not increase or decrease with time. Every other common continuous survival distribution (Weibull, gamma, log-normal) has a non-constant hazard rate. The constant hazard is a direct consequence of the memoryless property.
Mean Equals Standard Deviation
For the exponential distribution, E[X] = SD(X) = 1/λ. The coefficient of variation (CV = SD/mean) equals exactly 1. This property can serve as a quick check: if you estimate the mean and standard deviation from data and they are very different, the exponential may be a poor fit.
The Minimum of Exponentials
If X₁ ~ Exp(λ₁) and X₂ ~ Exp(λ₂) are independent, then min(X₁, X₂) ~ Exp(λ₁ + λ₂). This result generalizes to any number of independent exponential variables. It has direct applications in reliability: if a system fails when any one of k components fails, the system lifetime follows Exp(λ₁ + ⋯ + λₖ).