First, the key terms.

Random variable
A numerical quantity whose value depends on an uncertain outcome.
Expected value
A probability-weighted average under a specified model.
Independence
A condition where learning one event does not alter another’s probability in the relevant sense.

The idea, explained

A discrete variable takes separate values such as a count of arrivals; a continuous variable models quantities such as duration. A valid probability model assigns nonnegative likelihoods that total one across all possibilities. Expected value is a probability-weighted average, not necessarily a possible single outcome. Different families describe different situations: Bernoulli models a binary result, binomial a count across specified trials, Poisson a count under rate assumptions and normal a symmetric continuous pattern. Selection must follow the process and evidence.

STEP 1

Define outcomes and units

Distinguish customer arrivals per hour from minutes between arrivals. They are related but different variables. Specify the time interval, population and possible bounds before choosing a distribution. A normal model can produce impossible negative values for some quantities.

STEP 2

Match assumptions to the process

Binomial reasoning commonly assumes a fixed trial count, common success probability and independence. Real customers or events may influence each other. A familiar distribution should not be selected only because its spreadsheet function is convenient.

STEP 3

Compare average with variability

Two processes can share an expected value but differ greatly in spread and extreme outcomes. Inspect probabilities of consequential thresholds and the shape of the distribution, not just the mean. Test sensitivity to uncertain distribution parameters.

Follow a worked example.

A fictional daily demand model assigns probability 0.2 to 0 units, 0.5 to 10 units and 0.3 to 20 units. Probabilities total 1. Expected demand is 0 × 0.2 + 10 × 0.5 + 20 × 0.3 = 11 units.

Walk it through, one step at a time

  1. List every modeled outcome and verify that probabilities are valid and sum to one.
  2. Calculate the weighted average, keeping the demand unit consistent. Eleven is an average across repeated modeled days, not one of the listed daily outcomes.
  3. If stock is 10, the model’s shortage probability is 0.3. That threshold question may matter more operationally than the average.

Where people get stuck

A distribution fitted to historical observations may fail when the process, population or measurement changes.

Now try it yourself.

Pause before opening each answer. Explain your reasoning in your own words.

1. Check the core idea

Must the expected value be an outcome that can occur?

Reveal the answer & reasoning

No. A binary variable with equal probabilities has expected value 0.5 even though each realized result is zero or one.

2. Apply what you learned

Two equally likely payoffs are $0 and $100. Is the expected $50 a guaranteed payment?

Reveal the answer & reasoning

No. The expected value is $50, but the modeled individual outcomes remain $0 or $100. The average does not remove risk.