Probability Essentials for Analytics Interviews
After Descriptive Statistics for Analysts: Key Concepts Explained, the next question is how likely an event is and which probability model fits the data. Probability essentials matter in analytics interviews because they connect formulas such as joint probability, conditional probability and Bayes' Theorem with business situations like loan approval, email clicks, UPI transactions and Swiggy orders. This toolkit gives you the core formulas first, then shows when to use the most common probability distributions.
- Joint probability: P(A B) = P(A) × P(B|A), probability of both A and B occurring.
- Conditional probability: P(A|B) = P(A B) / P(B), probability of A given B has occurred.
- Bayes' Theorem: P(A|B) = P(B|A) × P(A) / P(B), used to update belief (prior) with new evidence to get posterior probability.
- Normal distribution is used for continuous data, natural phenomena and measurement errors.
- Binomial distribution is used for Yes/No outcomes and fixed n trials.
- Poisson, Uniform, Exponential and Log-Normal distributions map to count of rare events, equal probability over a range, time between events and skewed positive data.
The Big Picture: Formulas First, Distributions Next
Probability essentials have two parts. The first part is reasoning with events using joint probability, conditional probability and Bayes' Theorem. The second part is choosing the right distribution based on the type of data, the parameters available and the business example.
Core Probability Formulas
The three formulas below are the base toolkit for probability questions. They help you express the probability of two events occurring, the probability of one event after another has occurred and how belief changes after new evidence.
P(A|B) = P(B|A) × P(A) / P(B). Update belief (prior) with new evidence to get posterior probability.
Probability Distributions and When to Use Them
Distribution choice depends on when to use it, its parameters and the business situation. The table below gives the interview-ready mapping across Normal, Binomial, Poisson, Uniform, Exponential and Log-Normal distributions.
In this table, μ stands for mean, σ is used for std dev, n means trials, p means success prob, λ means average rate or rate, and a and b represent min and max. These parameters help match the probability model to the business example.
How to Choose the Right Distribution
Start by identifying the data pattern. Continuous data, natural phenomena and measurement errors point to Normal. Yes/No outcomes with fixed n trials point to Binomial. Count of rare events in fixed time or space points to Poisson.
Equal probability over a range points to Uniform. Time between events and waiting times point to Exponential, such as time between consecutive Swiggy orders in a city. Skewed positive data and multiplicative effects point to Log-Normal, such as stock prices, income distribution and startup valuations.
Conclusion
Probability essentials combine event formulas with distribution selection. Use joint probability, conditional probability and Bayes' Theorem to reason about events, then choose the distribution that matches the data pattern, parameters and business example.
The common mistake is treating all business data as Normal. Despite being called normal distribution, most real-world Indian business data such as income, website traffic, viral content shares and startup valuations follows a power-law or log-normal distribution - NOT a normal distribution!