Introduction to the Exponential Distribution

The exponential distribution models the time between independent events. Let's take a look at its parameter, PDF, CDF, and real-world applications.

The exponential distribution is one of the most important probability distributions in applied mathematics and simulation. It models the time between independent events that occur at a constant average rate. Examples: how long until the next customer arrives, how long until a machine fails, or how long until a radioactive atom decays.

If you’re building simulations, this distribution will be your workhorse. It’s vital that we understand this distribution and how it works.

The Parameter: Lambda

The exponential distribution has a single parameter: the rate parameter $\lambda$ (lambda), where $\lambda > 0$.

Lambda represents the average number of events per unit time, often called the tick in simulations. From this single parameter, we derive everything else:

Notice that the mean and standard deviation are equal: $\frac{1}{\lambda}$. This is a distinctive feature of the exponential distribution.

Example: If customers arrive at a coffee shop at an average rate of $\lambda = 3$ per hour, the expected time between arrivals is $\frac{1}{3}$ hour, or 20 minutes.

Probability Density Function (PDF)

The PDF gives the probability of the random variable having a specific value. For the exponential distribution:

$f(x) = \lambda e^{-\lambda x} \quad \text{for } x \geq 0$

The PDF is $0$ for $x < 0$. Negative time does not make sense.

Key observations:

• The curve starts at $f(0) = \lambda$ and decreases monotonically.
• Higher $\lambda$ values produce steeper curves concentrated near zero.
• Lower $\lambda$ values produce flatter curves that spread further out.
• All exponential distributions have a long tail that goes to infinity.

Use the slider below to see how $\lambda$ shapes the PDF:

Cumulative Distribution Function (CDF)

The CDF gives the probability that the random variable is less than or equal to a given value. For the exponential distribution:

$F(x) = P(X \leq x) = 1 - e^{-\lambda x} \quad \text{for } x \geq 0$

Key observations:

• $F(0) = 0$. The probability of a zero-length wait is zero.
• The CDF approaches $1$ as $x$ goes to $\infty$. The event will eventually happen if you wait long enough.
• Higher $\lambda$ values cause the CDF to rise more quickly. Events happen sooner.

Use the slider below to see how $\lambda$ shapes the CDF:

Reading the CDF: Pick a point on the x-axis (a time $x$) and read the corresponding y-value. That’s the probability the event has occurred by time $x$. For example, with $\lambda = 0.5$, the probability of the event occurring within 2 time units is $F(2) = 1 - e^{-0.5 \cdot 2} = 1 - e^{-1} \approx 0.632$.

The Survival Function

A natural companion to the CDF is the survival function (also called the reliability function):

$S(x) = P(X > x) = 1 - F(x) = e^{-\lambda x}$

Where the CDF answers “what’s the probability the event has happened by time $x$?”, the survival function answers “what’s the probability the event has not happened by time $x$?” This perspective is especially useful in reliability engineering and medical statistics.

Real-World Examples

The exponential distribution appears whenever events occur independently at a roughly constant rate:

• Customer arrivals. A call center receives calls at a rate of $\lambda = 12$ per hour. The time between consecutive calls follows $\text{Exp}(12)$, with an expected gap of 5 minutes.

• Equipment failure. A server has a failure rate of $\lambda = 0.001$ per hour (one failure per 1,000 hours on average). The time until the next failure follows $\text{Exp}(0.001)$.

• Radioactive decay. A radioactive isotope decays at rate $\lambda$. The time until the next decay event is exponentially distributed.

• Wait times. Patients arrive at a client at a rate of $\lambda = 12$ per hour. The time between arrivals follows $\text{Exp}(12)$, with a mean of 5 minutes.

In each case, the key assumption is that events are independent and occur at a constant average rate. When this assumption holds, the exponential distribution is the right model. When it doesn’t — say, if failure rates increase with equipment age — a different distribution is needed.