Most real situations that involve counting or waiting fit one of a small number of named probability distributions, and picking the wrong one is a common source of wrong answers in practice. This works through four distributions with real numbers, binomial, Poisson, uniform and exponential, and the specific signal in a question that points to each one.
In this guide
- Discrete versus continuous distributions
- Binomial: counting successes over a fixed number of trials
- Poisson: counting events over a fixed interval
- Uniform: every value in a range equally likely
- Exponential: waiting time between events
- Choosing the right distribution
- Common mistakes
- Cheat sheet
- Try it yourself
- FAQ
Discrete versus continuous distributions
A discrete distribution assigns probability to separate, countable values, like 0, 1 or 2 successes. A continuous distribution spreads probability smoothly across a whole range instead.
| Distribution | Type | What it counts or measures |
|---|---|---|
| Binomial | Discrete | Number of successes in a fixed number of trials |
| Poisson | Discrete | Number of events in a fixed interval of time or space |
| Uniform (continuous) | Continuous | Any value within a range, all equally likely |
| Exponential | Continuous | Waiting time until the next event |
Binomial: counting successes over a fixed number of trials
A binomial distribution applies when there is a fixed number of trials, each with only two possible outcomes, a constant success probability, and independent trials. Its mean is n times p, and its variance is n times p times (1 minus p).
| Question (n=10 trials, p=0.4) | Result |
|---|---|
| P(exactly 3 successes) | 0.2150 |
| P(exactly 4 successes) | 0.2508 |
| P(3 or fewer successes) | 0.3823 |
| Variance | 10 x 0.4 x 0.6 = 2.4 |
Drawing cards one at a time without replacement looks binomial at first glance, fixed number of draws, two outcomes per draw, but it is not: removing each card changes what remains, so the success probability is not constant across draws, which breaks a required condition.
Poisson: counting events over a fixed interval
A Poisson distribution models the number of times a rare, independent event occurs within a fixed interval of time or space, calls arriving at a help desk per hour, for instance. A defining feature: its mean and its variance are both simply equal to the rate parameter, usually written as lambda.
| Question (average rate = 3 complaints per day) | Result |
|---|---|
| P(exactly 2 complaints tomorrow) | 0.2240 |
| P(exactly 0 complaints tomorrow) | 0.0498 |
Scaling the rate up to 4 errors per hour on a website, the probability of 2 or fewer errors in the next hour comes out to 0.2381, summing the individual probabilities for 0, 1 and 2 errors.
Binomial approximated by Poisson
When the number of trials is large and the success probability is small, while their product n times p stays moderate, the binomial distribution’s shape converges closely to a Poisson distribution with that same rate. This is why rare-event counts, like defects in a very large batch, are often modelled directly as Poisson rather than binomial.
Uniform: every value in a range equally likely
A continuous uniform distribution gives every value within a given range an equal chance of occurring, a flat, constant probability density across the whole range with no peak or dip anywhere. A random number generator returning any decimal between 0 and 1 with equal likelihood is a realistic example; human height or a light bulb’s lifetime are not, since both cluster around typical values.
| Question (uniform between 2 and 10) | Calculation | Result |
|---|---|---|
| Mean | (2 + 10) / 2 | 6.0 |
| Variance | (10 minus 2) squared / 12 | 5.33 |
| P(strictly between 3 and 7) | (7 minus 3) / (10 minus 2) | 0.50 |
| P(greater than 8) | (10 minus 8) / (10 minus 2) | 0.25 |
Because probability is just proportional to interval width here, the probability of landing in any sub-range is simply that sub-range’s width divided by the total range width, no bell curve or peak to account for.
Exponential: waiting time between events
The exponential distribution models the waiting time until the next event in a process where events occur at a constant average rate, defined only over non-negative values, since a waiting time can never be negative. Its mean is 1 divided by the rate.
| Question | Calculation | Result |
|---|---|---|
| Mean wait, arrivals at 0.5 per minute | 1 / 0.5 | 2.0 minutes |
| P(next arrival within 4 minutes), same rate | 1 minus e^(-0.5 x 4) | 0.8647 |
| P(a part survives past 5 years), failure rate 0.2/year | e^(-0.2 x 5) | 0.3679 |
The exponential distribution’s memoryless property means the probability of waiting an additional amount of time does not depend on how long you have already waited. A component that has already run for a year is, remarkably, no more or less likely to fail in the next hour than a brand-new one.
Choosing the right distribution
Poisson and exponential are two sides of the same process: if events occur as a Poisson process, the waiting time between consecutive events follows an exponential distribution, and the two share the same underlying rate. The question to ask is whether you are counting how many events happen (Poisson) or measuring how long until the next one (exponential).
| Scenario | Best-fit distribution | Why |
|---|---|---|
| Inspecting 20 items, each 5 percent independently defective | Binomial | A fixed count of trials, each with the same success rate |
| Calls arriving at a call centre in a given hour, known average rate | Poisson | No fixed trial count, just an average rate of arrivals over a window |
| Waiting time between one bus and the next, constant average arrival rate | Exponential | Measuring the gap between events, not how many arrive |
Common mistakes
- Applying binomial to a without-replacement scenario. A constant success probability across trials is required; removing items without replacing them breaks that condition.
- Confusing Poisson (how many) with exponential (how long). Poisson counts occurrences in a fixed window; exponential measures the gap between occurrences. The two are related but answer different questions.
- Assuming a continuous distribution when outcomes are actually countable. If a question asks for a specific count, 0, 1, 2 successes, it needs a discrete distribution, not a continuous one.
- Treating ‘equally likely in a range’ as the same as ‘clustered around an average’. Only a genuinely flat density, like a random number generator, is uniform; most real measurements cluster instead.
- Forgetting the memoryless property applies only to the exponential distribution. Plenty of real waiting times, like a light bulb’s lifetime, are not memoryless, past usage does affect the odds of failing soon.
Cheat sheet
| Question | Answer |
|---|---|
| What does binomial need: fixed or unknown trial count? | A fixed number of trials |
| What is the binomial mean formula? | n times p |
| What is the binomial variance formula? | n times p times (1 minus p) |
| What does Poisson count? | Events in a fixed interval of time or space |
| What is distinctive about Poisson's mean and variance? | Both equal lambda, the rate |
| When does binomial approximate Poisson? | Large n, small p, moderate n times p |
| What is the uniform distribution's mean? | The midpoint of its range |
| What does exponential model? | Waiting time until the next event |
| What is the exponential mean formula? | 1 divided by the rate |
| What links Poisson and exponential? | Same rate: Poisson counts events, exponential measures the gaps between them |
| What is the memoryless property? | Future waiting time does not depend on time already waited |
Try it yourself
Quick check
A fair coin is flipped 6 times. What is the probability of getting exactly 4 heads (4 dp)?
Show answer
0.2344. This is binomial with n equals 6 and p equals 0.5, a fixed number of independent trials with a constant success probability, so plugging 4 successes into the binomial formula gives 0.2344.
Frequently asked questions
What conditions does a binomial distribution require?
A fixed number of trials, each with only two possible outcomes, a constant success probability, and independent trials. Drawing without replacement violates the constant-probability condition.
What does the Poisson distribution model?
The number of times a rare, independent event occurs within a fixed interval of time or space, such as calls arriving at a help desk per hour, based on a known average rate.
What is the difference between a discrete and a continuous distribution?
A discrete distribution assigns probability to separate, countable values. A continuous distribution spreads probability smoothly across a whole range of possible values.
How are the Poisson and exponential distributions related?
If events occur as a Poisson process, the waiting time between consecutive events follows an exponential distribution, and the two distributions share the same underlying rate parameter.
What does the memoryless property of the exponential distribution mean?
The probability of waiting an additional amount of time does not depend on how long you have already waited, so a used component is no more likely to fail soon than a brand-new one.
When is it appropriate to approximate a binomial distribution with a Poisson distribution?
When the number of trials is large and the success probability is small, while their product stays at a moderate value, the binomial distribution’s shape converges closely to a Poisson distribution with that same rate.
Test yourself
Timed questions on Probability Distributions, with an explanation for every answer.