Skip to content
Tutorials Center
My learning

Topics

Lesson 4 of 38 · Getting Started

Poisson Arrivals

Understand the Poisson process: counts per interval, exponential gaps, merging and splitting streams. Check each fact with a quick simulation.

Most basic queueing models assume that customers arrive according to a Poisson process. It is the standard model of "completely random" arrivals: events that happen independently, at a steady average rate, with no memory of the past. Phone calls, web requests from many unrelated users and radioactive decays are all well described by it.

This lesson states the key facts and verifies each with code.

Two equivalent views

A Poisson process with rate λ can be described in either of two ways.

Counting view. The number of arrivals in an interval of length t has a Poisson distribution with mean λt:

P(N(t) = k) = e^(-λt) (λt)^k / k!

Counts in disjoint intervals are independent.

Gap view. The times between consecutive arrivals are independent and exponentially distributed with mean 1/λ:

P(gap > x) = e^(-λx)

These are the same process seen from two angles. If you generate exponential gaps and count how many land in each unit of time, the counts are Poisson.

Let us generate gaps and count arrivals per minute. A Poisson distribution has the property that its mean and variance are both equal to λt. For λ = 3 per minute and t = 1, both should be near 3.

Python Gaps to counts

Press Run (or Ctrl + Enter) to see the result here.

Program input · what you'd type while the program runs

Type the answers before you press Run, one per line. Each time the program reads a line (with input()), it gets the next line from this box.

The simulated frequencies track the formula closely. Notice that even with an average of 3 per minute, some minutes see 0 arrivals and some see 7 or more. That is the randomness that causes queues.

Three useful properties

  1. Memorylessness. The time until the next arrival does not depend on how long you have already waited. The process "forgets".
  2. Superposition (merging). If you merge independent Poisson streams with rates λ1 and λ2, the result is Poisson with rate λ1 + λ2. Two checkout doors feeding one hallway give one Poisson stream.
  3. Thinning (splitting). If each arrival is independently sent to branch A with probability p, branch A is Poisson with rate pλ, and branch B is Poisson with rate (1-p)λ, and the two branches are independent.
flowchart LR
  A["Stream 1<br/>rate λ1"] --> M["Merge"]
  B["Stream 2<br/>rate λ2"] --> M
  M --> C["Poisson<br/>rate λ1 + λ2"]

Merging independent Poisson streams adds their rates.

Python Merging and splitting

Press Run (or Ctrl + Enter) to see the result here.

Program input · what you'd type while the program runs

Type the answers before you press Run, one per line. Each time the program reads a line (with input()), it gets the next line from this box.

When Poisson is a good model

The model fits well when many independent customers each rarely arrive, such as thousands of users hitting a site. It fits poorly when arrivals are scheduled (appointments), come in batches (a bus unloading) or depend on the queue (people avoid long lines). Always ask whether the assumption is reasonable before trusting a formula that relies on it.

Check your understanding

Quick check · 4 questions

of 4 answered You scored / 4 · perfect!

  1. 1. Arrivals are Poisson with rate 4 per hour. What is the mean gap between arrivals?

    The mean gap is 1/lambda = 1/4 hour = 15 minutes.
  2. 2. Two independent Poisson streams have rates 2 and 5 per minute. What is the merged stream?

    Superposition of independent Poisson streams adds their rates: 2 + 5 = 7.
  3. 3. For a Poisson count with mean 3, what is the variance?

    A Poisson random variable has variance equal to its mean.
  4. 4. Which situation is a poor fit for Poisson arrivals?

    Scheduled appointments are regular, not memoryless random arrivals.

Summary

  • A Poisson process has Poisson counts (mean λt) and exponential gaps (mean 1/λ).
  • It is memoryless: the past does not change the wait for the next arrival.
  • Merging adds rates; random splitting scales rates and keeps independence.
  • It models many independent sources well, but not schedules or batches.

Next: Exponential Service Times