Interactive module guide
Poisson Distribution Calculator & Guide
The Poisson distribution models a count of events observed over a stated amount of time, space, distance, or other exposure when events occur independently at a stable average rate.
Its single parameter λ is both the expected count and the variance for that exposure.
Use it for event counts with no fixed maximum; do not confuse it with the waiting time between events or with data whose variance is far larger than the mean.
Free · No sign-up · Calculations stay in your browser
Is this the right module for my question?
Use it when
- A Poisson model is plausible for arrivals, incidents, defects, or emissions counted over a clearly defined exposure. The counting window and the definition of one event must remain consistent. If exposure doubles while the rate remains stable, the expected count should double as well.
- Choose it when individual events are comparatively rare at a fine scale, simultaneous events are negligible in that scale, and occurrence in one small interval does not change occurrence in another. These conditions describe a process, not merely a histogram of nonnegative integers.
- Use the exact PMF and CDF for questions about a particular count, at most a threshold, or at least a threshold. The distribution is especially useful when zero is a meaningful possible observation and there is no pre-specified number of trials.
Choose another method when
- Do not use a basic Poisson model for proportions, fixed-n successes, waiting durations, or counts with a structural upper bound. A binomial distribution handles a fixed number of opportunities, while an exponential or gamma model may describe elapsed time.
- Avoid it when the event rate changes across the window, events arrive in bursts, units differ in unmodeled risk, or one event changes the chance of another. Such mechanisms commonly produce overdispersion, underdispersion, or excess zeros relative to Poisson variance.
Interactive tool
The calculator loads as you approach this section so the guide remains fast on mobile connections.
How to read the result
Read one PMF bar as the probability of exactly that many events in the declared exposure. Read the CDF step at k as the probability of k or fewer. Comparing results across different window lengths without converting λ to the matching exposure is meaningless.
An unusually high count under the model may reflect a genuine rate change, clustering, an exposure error, or a poor model. The probability itself does not identify which explanation is correct and should not be interpreted as the probability that the process is “out of control.”
Inputs and parameter meaning
Enter λ as the expected number of events in the exact exposure represented by one observation. A rate of 3 calls per hour corresponds to λ=3 for a one-hour count, λ=1.5 for a half-hour count, and λ=24 for an eight-hour count only if the rate is stable across those periods.
The random variable X takes whole-number values 0,1,2,… without a finite upper limit. The PMF gives exact count probabilities. The CDF at k gives P(X≤k), and P(X≥k)=1−P(X≤k−1). Be explicit about the exposure and endpoint whenever a probability is reported.
Because both mean and variance equal λ in the ideal model, λ controls center and spread simultaneously. Larger λ moves the typical count upward, broadens the distribution in absolute units, and makes its standardized shape more symmetric.
Formula and available parameters
Density or probability mass
Mean
Variance
- Rate parameter (λ): interactive range 0.1 to 20; default 3.
What the method is doing
The mass P(X=k)=λ^k exp(−λ)/k! balances the growing number of ways events can occur with the exponential probability of no event. The formula assumes a homogeneous rate over the chosen exposure. Aggregating periods with different rates can create a mixture that is not Poisson even if each period is conditionally Poisson.
For large λ the distribution becomes more symmetric and may be approximated by a normal distribution with mean λ and variance λ. A continuity correction can improve finite-count interval estimates. Exact Poisson calculations remain preferable when convenient, especially for small means or tail decisions.
Worked example: support tickets during one hour
A support desk receives an average of λ=2.5 urgent tickets per hour during a stable overnight shift. An analyst wants the modeled probability of receiving no more than one urgent ticket in the next hour, assuming the historical homogeneous Poisson process is an acceptable description.
- 1Define X as the number of urgent tickets in a one-hour window and set λ=2.5. The event “no more than one” is P(X≤1), so add P(X=0) and P(X=1) or read the CDF at one.
- 2The two masses are exp(−2.5) and 2.5exp(−2.5). Their sum is 3.5exp(−2.5), approximately 0.287. The PMF plot shows why counts around two or three are more common under this rate.
- 3If the staffing decision covers a two-hour interval, do not reuse λ=2.5. Under a stable rate the corresponding mean is λ=5, producing a different count distribution and a much smaller probability of at most one ticket.
Interpretation
Under the stated process assumptions, one or zero urgent tickets occur in roughly 29% of one-hour windows. This is a model-based frequency, not a guarantee for the next hour. Staffing should also consider ticket duration, time-of-day variation, burst behavior, and the cost of unusually busy periods.
Common mistakes
- Using a rate without matching its unit to the observation window is a fundamental error. Always multiply a per-unit rate by the actual exposure before entering λ.
- Mean close to variance is not proof of a Poisson mechanism, and unequal sample mean and variance are not automatically decisive in a small dataset. Diagnose the process and quantify sampling uncertainty.
- Do not use the Poisson count CDF to answer a waiting-time question. The exponential distribution describes the time to the next event only under the corresponding Poisson-process assumptions.
Limits and independent validation
The tool represents the basic one-parameter model and does not fit covariates, offsets, zero inflation, time-varying rates, or random effects. Those features often matter for operational and epidemiological counts.
A normal approximation can look visually close near the center while missing a tail threshold. Evaluate the exact probability relevant to the decision rather than relying on the overlap of central curves.
Before using the result
- Check event definitions, exposure units, rate stability, and whether repeated windows are independent. Plot counts against time and compare variation across comparable exposures before adopting a homogeneous rate.
- Verify a result by summing exact masses or using an independent Poisson CDF. For complements, calculate through k−1 when the upper event includes k, and retain adequate numerical precision.
Related modules and resources
See every option in the Probability Distribution Explorer or review the DistriScope methodology. Educational information; last reviewed 2026-08-06. Verify consequential calculations independently.