Unit-2
Risk Estimation
09/13/2025 MBA(Analytics)-02, Second Semester 1
Overview
• Risk Analysis vs Risk Estimation
09/13/2025 MBA(Analytics)-02, Second Semester 2
Risk Analysis
• A multi-step process that aims to understand
and manage potential risks.
• Involves identifying, analyzing, and evaluating
risks to determine their potential impact and
likelihood.
• Helps organizations make informed decisions
about risk mitigation and management.
• Risk analysis is the initial stage of risk
management, which examines the value of
assets to the organization, identifies
vulnerabilities to those assets, and assesses how
sensitive each asset is to those risks.
09/13/2025 MBA(Analytics)-02, Second Semester 3
Risk Estimation
• A specific step within the risk analysis process.
• Focuses on quantifying the likelihood and impact of
identified risks.
• Can involve using various methods and tools, such as
probability-impact matrices, risk scoring, and parametric
estimates.
• The goal of risk estimation is to produce measures of the
health, safety, and environmental risks that are being
assessed.
• Risk estimates can be based on absolute or relative risk
estimates.
– Absolute risk estimates require the predetermination of a
damage state or consequence level of interest.
– Relative risk estimates show which events happen relatively
more often than others.
09/13/2025 MBA(Analytics)-02, Second Semester 4
How to compute deviations from
expected outcomes of model by testing
it with multiple inputs?
• To assess how well the model's predictions
align with actual outcomes when tested with
multiple inputs.
• Compute the residuals and evaluate various
performance metrics.
09/13/2025 MBA(Analytics)-02, Second Semester 5
• Prepare Dataset:
Collect a dataset containing input variables and their
corresponding observed (actual) outcomes.
• Generate Predictions:
Apply a model to the input variables to obtain predicted
outcomes.
• Calculate Residuals:
For each data point, compute the residual, which is the
difference between the observed outcome (y) and the
predicted outcome (y’):
Residual=y−y’
Residuals indicate the deviation of the model's predictions
from the actual observations.
09/13/2025 MBA(Analytics)-02, Second Semester 6
Evaluation of Performance Metrics
• Mean Absolute Error (MAE):
– Calculates the average of the absolute residuals, reflecting the average
magnitude of errors.
MAE= 1/n∑ |y-y’|
• Mean Squared Error (MSE):
– Computes the average of squared residuals, giving more weight to larger errors.
MSE=1/n∑ (y-y’)2
• Root Mean Squared Error (RMSE):
– The square root of MSE, providing error magnitude in the original units.
RMSE=√MSE
• R-Squared(R2):
Indicates the proportion of variance in the observed outcomes explained
by the model:
R2-1-∑(y-y’)2/∑(y-y’’)2
09/13/2025 MBA(Analytics)-02, Second Semester 7
Techniques to compute deviations from expected
outcomes by testing with multiple inputs
• Various statistical and simulation techniques can be used to
compute deviations from expected outcomes when testing
with multiple inputs.
– Standard Deviation
– Monte Carlo Simulations
– Sensitivity Analysis
09/13/2025 MBA(Analytics)-02, Second Semester 8
Standard Deviation
• Purpose: Measures the dispersion or spread of data points around
the mean, indicating how much individual outcomes deviate from
the average.
• How it works:
– Calculate the mean of the dataset.
– Determine the deviation of each data point from the mean.
– Square each deviation and sum them.
– Divide the sum of squared deviations by (n-1) for a sample or n for a
population (where n is the number of observations or data points).
– Take the square root of the result to obtain the standard deviation.
• Interpretation:
– A larger standard deviation implies greater variability and a wider range of
deviations from the mean, while a smaller standard deviation indicates
more consistency and tighter clustering of outcomes.
09/13/2025 MBA(Analytics)-02, Second Semester 9
Example
• Problem Statement: Testing the lifespan of 10 light bulbs and record their lifespans
(in hours): 500, 510, 520, 530, 540, 550, 560, 570, 580, 590.
• Calculate the mean: (500+510+...+590)/10 = 545 hours.
• Calculate the deviations from the mean: For each bulb, subtract the mean (545)
from its lifespan. For example, the first bulb: 500 - 545 = -45 hours. Repeat for all
bulbs.
• Square the deviations: (-45)^2 = 2025, (-35)^2 = 1225, etc.
• Calculate the variance: Sum the squared deviations (2025 + 1225 + ... + 25) and
divide by the number of bulbs (10). The result is the variance.
• Calculate the standard deviation: Take the square root of the variance.
• Interpreting the results:
– A larger standard deviation indicates that the lifespans of the bulbs vary
more widely, with some bulbs lasting significantly longer or shorter than
the average.
– A smaller standard deviation indicates that the lifespans are more clustered
around the average, suggesting better consistency in the bulb production.
09/13/2025 MBA(Analytics)-02, Second Semester 10
Monte Carlo Simulation
• Purpose: Simulates a range of possible outcomes by repeatedly running a model with
random input values, allowing for analysis of the potential impact of uncertainty.
• How it works:
– Identify variables with inherent uncertainty and assign probability distributions to
them.
– Use random numbers within these distributions to generate different input
values.
– Run the model with these inputs and record the results.
– Repeat this process many times (e.g., thousands or millions of times).
– Analyze the distribution of the results to understand the range of possible
outcomes, including potential deviations from expected values.
• Interpretation:
– The simulation provides a probability distribution of potential outcomes, showing
the likelihood of different deviations from the expected value.
09/13/2025 MBA(Analytics)-02, Second Semester 11
Example
• A Monte Carlo simulation can estimate deviations from
expected outcomes by repeatedly running a model with
different, randomly sampled inputs to represent the
uncertainty of various factors.
• By analyzing the distribution of the simulated outputs, we can
gain insight into the potential range of results and identify the
most likely outcome, as well as understand the probability of
exceeding or falling short of specific targets.
09/13/2025 MBA(Analytics)-02, Second Semester 12
Scenario
• Imagine you're planning to launch a new product and want to
estimate its potential sales revenue. Instead of making a single,
fixed sales forecast, you can use a Monte Carlo simulation to
explore a range of possible outcomes, considering various factors
like market demand, pricing, and marketing effectiveness.
1. Identify Input Variables:
– Identify the key factors influencing the sales revenue, such as:
Average monthly demand (units sold)
– Average selling price per unit
– Marketing spend
– Customer acquisition cost
– Probability of a customer making a purchase
– Customer lifetime value
09/13/2025 MBA(Analytics)-02, Second Semester 13
2. Define Distributions:
– For each input variable, define a probability distribution that
represents the uncertainty in its value.
For example: Average monthly demand could be modelled as a normal
distribution with a mean based on your market research and a
standard deviation reflecting the potential range of variation.
– Customer acquisition cost might be modelled as a triangular
distribution with a minimum, maximum, and most likely value.
3. Run the Simulation:
– Use a random number generator to sample values for each input
variable from its defined distribution. This process is repeated
many times (thousands or millions of iterations) to create a large
sample of possible outcomes.
– For each iteration, calculate the predicted sales revenue based
on the sampled input values. This could involve a simple formula
or a complex model depending on the situation.
09/13/2025 MBA(Analytics)-02, Second Semester 14
4. Analyze the Results:
– Analyze the distribution of the simulated sales revenue
outcomes.
– Calculate the average or expected sales revenue, the standard
deviation (which measures the variability), and other relevant
statistics.
– Determine the probability of achieving specific revenue targets
or the probability of exceeding or falling short of certain
thresholds.
– Visualize the results using histograms, charts, or other
appropriate methods to illustrate the range of possible
outcomes and identify trends.
09/13/2025 MBA(Analytics)-02, Second Semester 15
Sensitivity Analysis
• Purpose: Investigates how changes in input variables affect the
output or outcome of a model.
• How it works:
– Identify key input variables that are believed to have a
significant impact on the outcome.
– Vary these variables individually or in combination while
keeping other variables constant.
– Observe how the output changes in response to these
variations.
• Interpretation:
– This helps determine which input variables are most
influential and how much they impact the deviation from the
expected outcome.
09/13/2025 MBA(Analytics)-02, Second Semester 16
Sensitivity Analysis
• Sensitivity analysis uses multiple inputs to assess how variations in
specific variables impact an expected outcome, revealing the
model's sensitivity to those changes.
• A business is analyzing the impact of price increases on sales by
varying price points and observing the effect on demand.
09/13/2025 MBA(Analytics)-02, Second Semester 17
Financial Modelling
Scenario:
– A company is deciding whether to invest in a new project.
Model:
– The Net Present Value (NPV) calculation is used to model the
project's potential profitability.
Sensitivity Analysis:
– The company explores how the NPV changes when key inputs like
interest rates, discount rates, project cost, or projected sales are
varied.
– For instance, "What if interest rates increase by 1%, 2%, or 5%? How
does this impact the NPV?"
Outcome:
– The analysis reveals which input variables have the most significant
impact on the NPV, helping the company to understand the project's
risk and potential return under various scenarios.
09/13/2025 MBA(Analytics)-02, Second Semester 18
Sales Prediction
Scenario:
– A business wants to understand how increased marketing spending
affects sales revenue.
Model:
– A regression model that predicts sales based on marketing expenditure,
customer demographics, and other factors.
Sensitivity Analysis:
– The model is tested by varying the marketing spending variable to see
how it affects sales predictions.
– For example, "What happens to predicted sales if marketing spending
increases by 10%, 20%, or 30%?"
Outcome:
– The analysis reveals the relationship between marketing spending and
sales, allowing the business to determine the optimal spending level to
achieve desired sales targets.
09/13/2025 MBA(Analytics)-02, Second Semester 19