Poisson distribution
Definition
Poisson distribution
The Poisson distribution is a discrete probability model that predicts how often a rare, independent event happens in a fixed window of time or space. Named for French mathematician Siméon Denis Poisson in 1837, it powers staffing models, queue forecasts, and risk pricing across modern business — anywhere arrivals look random but average out.
Key takeaways
- Poisson distribution uses a single parameter, lambda (λ), the average number of events per interval.
- It only fits events that are independent, rare, and counted in whole numbers within a fixed window.
- Outsourcing teams rely on it for call-center staffing, claims pricing, fraud monitoring, and capacity planning.
The math sits behind everyday decisions you probably never label as statistics: how many agents to roster for the 9pm shift, how many spare servers to provision, how much to charge for a cyber-insurance policy. Get the average right, and Poisson tells you the odds of any specific count.
How it works
Poisson distribution calculates the probability of exactly k events in an interval when those events occur at a known average rate λ and are statistically independent. The formula is P(k events) = (λ^k × e^−λ) / k!, where e ≈ 2.71828. Mean and variance both equal λ — a defining feature that makes the model easy to validate against real data.
You need four conditions to apply Poisson cleanly: events are countable in whole numbers, two events cannot happen at the exact same instant, the rate λ is constant across the interval, and each event is independent of the next. Call arrivals, equipment failures, and insurance claims usually satisfy all four.
When λ grows large (above roughly 30), the Poisson curve flattens and starts to mirror a normal distribution, which is why analysts often switch models for high-volume systems. For tiny λ values, the distribution is sharply skewed toward zero.
| λ (avg per hour) | P(0 events) | P(1 event) | P(5 events) |
|---|---|---|---|
| 1 | 36.8% | 36.8% | 0.3% |
| 3 | 5.0% | 14.9% | 10.1% |
| 5 | 0.7% | 3.4% | 17.5% |
| 10 | 0.005% | 0.05% | 3.8% |
Source: standard Poisson PMF calculation, λ values typical of small-to-mid call queues.
Examples
Real-world Poisson use is everywhere once you start looking. Each example below uses dated, published figures so you can sanity-check the model against actual numbers.
Call-center staffing in the Philippines. Manila-based BPOs running US retail accounts often see 240 inbound calls per hour during evening peaks. That gives λ = 4 calls per minute. A workforce planner uses Poisson to size the headcount needed to keep abandon rates under 5%, then layers in Erlang-C — itself a Poisson derivative — for service-level math. The US Bureau of Labor Statistics tracked roughly 2.6 million customer-service jobs in 2023, many of them sized this way (BLS Occupational Employment).
Insurance claim pricing. Lloyd’s of London syndicates model cyber-attack frequency using Poisson because attacks against a given firm are rare and roughly independent. A 2023 Lloyd’s report estimated a major cloud-services outage could trigger USD 4.5 billion in claims globally, a number built on Poisson frequency layered with severity distributions (Lloyd’s systemic risk scenarios).
E-commerce checkout failures. Shopify reported handling 967,000 orders per minute at peak during Black Friday 2023. Engineering teams model checkout-error spikes, the rare and independent kind, with Poisson to set autoscaling thresholds and on-call rotations.
Hospital ER arrivals. A 200-bed regional hospital averaging 12 ambulance arrivals per hour uses Poisson to staff triage. The CDC logged 139.8 million US emergency-department visits in 2021, a workload that demands tight queue modelling.
Related terms
These sibling concepts come up constantly in outsourcing capacity planning and analytics work. Most workforce-management dashboards lean on at least three of them.
- Binomial distribution models a fixed number of yes/no trials; Poisson is its limiting form when trials are many and success is rare.
- Normal distribution approximates Poisson when λ is large, simplifying staffing math.
- Standard deviation measures spread; for Poisson, it equals √λ.
- Probability is the broader umbrella concept Poisson sits inside.
- Forecasting is the discipline that turns Poisson outputs into staffing and inventory plans.
- Predictive analytics extends Poisson with machine-learning features.
- Queuing theory layers service-time models on top of Poisson arrivals.
FAQ
When should I use Poisson instead of binomial distribution?
Use Poisson when the number of trials is very large, the success probability is very small, and you only know the average rate. Use binomial when you have a clear, fixed trial count with a known per-trial probability.
What is lambda (λ) in Poisson distribution?
Lambda is the average number of events in your chosen interval. If a call center averages 60 calls per hour, λ = 60 for an hourly model or λ = 1 for a per-minute model. Both views are valid as long as the rate stays constant.
Can Poisson distribution have decimal answers?
The input λ can be any positive decimal (3.7 calls per minute is fine), but the output k must be a whole number. Probabilities apply to “exactly 4 events,” never “exactly 4.3 events.”
How accurate is Poisson for real business forecasting?
It is accurate when the four assumptions hold: independence, constant rate, rarity, and whole-number counts. Real systems often violate the constant-rate rule across the day, so analysts segment the day into smaller windows where λ stays stable.
Is Poisson distribution still used in modern AI workflows?
Yes. Recommendation engines, anomaly-detection models, and ad-bidding systems use Poisson layers to model event counts before deeper neural-network steps take over.
Want help applying these models to your own staffing or capacity plan? Explore the vetted analytics and BPO partners in the Outsource Accelerator directory to find a team that does this every day.
Independent
