DEV Community: Enock Kipngetich

How Statistics Can Be Used To Drive Business Decisions

Enock Kipngetich — Tue, 03 Feb 2026 06:49:49 +0000

A Practical Case Study Using Retail Sales Data

1. Introduction

In today’s data-driven world, businesses collect large volumes of data but often struggle to turn it into actionable decisions. While advanced machine learning models are popular, statistics remains the foundation of effective business analysis. Statistical methods help organizations understand performance, manage uncertainty, test assumptions, and make informed decisions with confidence.

2. Problem Statement

The retail company operates both online and physical stores across multiple regions. Management faces three key challenges:

Understanding revenue performance
What does “typical” revenue look like?
How stable are sales over time?
Evaluating reliability of insights

Are observed patterns meaningful or due to randomness?
Is the data representative of the entire business?
Assessing marketing effectiveness

Does running a marketing campaign increase average revenue per transaction?
Is the observed increase statistically and practically meaningful?

3. Statistical Methods Used

This project applied several core statistical techniques, each serving a specific business purpose.

3.1 Descriptive Statistics

Descriptive statistics were used to summarize and understand revenue data.

Central Tendency

Mean
Median
Mode

These measures helped identify what a “typical” revenue value looks like.

Dispersion

Range
Variance
Standard deviation

These metrics measured how much revenue varies over time and across transactions.

3.2 Distribution Shape Analysis

Revenue data was visualized using histograms to assess:
Skewness (direction of asymmetry)
Kurtosis (presence of extreme values)

Understanding distribution shape helps determine:

Which summary statistics are reliable
Whether standard statistical tests can be applied

3.3 Data Visualization

Several visual tools were used:

Line charts to analyze revenue trends over time
Bar charts to compare store types
Box plots to compare regions
Scatter plots to explore the relationship between marketing spend and revenue
Why this matters:
Visualizations reveal patterns that raw numbers often hide and help stakeholders understand insights quickly.

3.4 Sampling and Bias Analysis

The project examined:

The difference between population and sample
The impact of sampling bias, especially when only urban stores are included

A stratified random sampling approach was recommended to improve representativeness.

3.5 Law of Large Numbers (LLN)

The Law of Large Numbers was demonstrated by showing how sample means stabilize as sample size increases.
Business lesson:
Decisions based on small datasets are risky and may lead to misleading conclusions.

3.6 Central Limit Theorem (CLT)

The CLT was used to show that:
Even when revenue is skewed,
The distribution of sample means becomes approximately normal with sufficient sample size

This justified the use of parametric statistical tests.

3.7 Hypothesis Testing (t-test)

A one-tailed t-test was conducted to evaluate whether marketing campaigns increased average revenue.
Null hypothesis (H₀): No difference in average revenue
Alternative hypothesis (H₁): Campaign revenue is higher
The test used:
95% confidence level
α = 0.05

3.8 Effect Size and Power

Beyond statistical significance, the project calculated Cohen’s d to measure the magnitude of the campaign’s effect.
Power considerations were discussed to explain why some real effects may not appear statistically significant.

4. Key Findings

The statistical analysis produced several important findings:

Revenue is positively skewed
Extreme high-revenue transactions exist
The median is a better measure of typical revenue than the mean
Sales variability is high
High standard deviation indicates unstable revenue
External factors (campaigns, seasonality) strongly influence performance
Online stores perform strongly

Online channels contribute significantly to total revenue
They show scalability advantages

Marketing campaigns increase revenue
The t-test showed a statistically significant increase in average revenue
Cohen’s d indicated a medium practical effect
Sampling matters
Urban-only samples lead to biased conclusions
Representative sampling improves decision reliability

5. Business Implications

The findings translate directly into business actions:

Better Performance Measurement

Use median revenue in reports to avoid distortion from extreme values

Improved Forecasting

High variability suggests the need for better demand planning and revenue smoothing strategies

Evidence-Based Marketing Decisions

Marketing campaigns should be continued and optimized, not evaluated based on intuition alone

Smarter Data Collection

Representative sampling ensures insights reflect the entire business
Larger samples increase confidence in decisions

Balanced Decision-Making

Statistical significance should be combined with effect size and business context

6. Why This Matters for Data Learners

This case study highlights an important lesson for beginners and intermediate data learners:

You don’t need machine learning to deliver value.
Strong statistical thinking is often enough.
By mastering:
Descriptive statistics
Sampling concepts
Probability theory
Hypothesis testing

You can solve real business problems and communicate insights clearly to stakeholders.

7. Conclusion

Statistics plays a critical role in transforming raw data into informed business decisions. Through this retail sales case study, we demonstrated how statistical methods help businesses understand performance, reduce uncertainty, test assumptions, and evaluate strategies objectively.

When applied correctly, statistics provides not only answers, but confidence that decisions are based on evidence rather than guesswork.

Ridge vs Lasso Regression

Enock Kipngetich — Thu, 22 Jan 2026 05:40:31 +0000

1. Ordinary Least Squares (OLS)

method used in linear regression to estimate model parameters by minimizing the sum of squared errors between actual and predicted values.
Loss=∑](𝑦−𝑦^)2Loss=∑(y−y^)2

Why OLS can overfit:

.Works poorly with many features
.Sensitive to noise
. Performs badly when features are correlated
. Fits training data too closely

2. Regularization

reduces overfitting by adding a penalty term to the loss function.

Why it helps:

Controls model complexity
Reduces large coefficients
Improves performance on unseen data

3. Ridge Regression (L2 Regularization)

Loss Function:
Loss=∑(𝑦−𝑦^)2+𝜆∑𝛽2Loss=∑(y−y^)2+λ∑β2

Key Points:

Shrinks coefficients toward zero
Reduces variance
Keeps all features
Works well with multicollinearity
Does not perform feature selection

4. Lasso Regression ( Regularization)

Loss Function:
Loss=∑(𝑦−𝑦^)2+𝜆∑∣𝛽∣Loss=∑(y−y^)2+λ∑∣β∣

Key Points:

Can reduce coefficients to exactly zero
Performs automatic feature selection
Produces simpler, more interpretable models

5. Ridge vs Lasso Comparison

Ridge: Uses L2 penalty, shrinks coefficients, keeps all features, good when all variables matter.
Lasso: Uses L1 penalty, sets some coefficients to zero, performs feature selection, good when few variables matter.

6. House Price Prediction Example

Features:
House size
Bedrooms
Distance to city
Nearby schools
Noisy variables

If all features affect price → Ridge Regression

Keeps all features
Reduces overfitting
Handles correlated variables

If only few features matter → Lasso Regression

Removes irrelevant features
Improves interpretability
Reduces noise

7. Model Evaluation

Detecting Overfitting:
High training accuracy + low test accuracy → Overfitting
Similar train & test error → Good model

Role of Residuals:

Show prediction errors
Help detect patterns and outliers
Random residuals = good model

Summary

OLS -Simple but overfits
Ridge -Shrinks coefficients
Lasso -Selects features
Regularization -Improves generalization

HOW TO CONNECT POSTGRES SQL TO POWER BI USING LOCAL POSTGRES AND AIVEN

Enock Kipngetich — Sun, 23 Nov 2025 13:33:17 +0000

Connect to Local PostgreSQL from Power BI

Open Power BI Desktop
Click Get Data
Choose PostgreSQL database
Enter: Server: localhost:port Database: postgres

Enter your PostgreSQL username and password

Click Connect

Once connected, you can load or transform your data.

Connecting Power BI to PostgreSQL Hosted on Aiven

Aiven

Step 1 — Create a PostgreSQL Service in Aiven

In the Aiven console:

Click Create Service
Select PostgreSQL
Choose cloud vendor, region, and plan
Deploy the service

Aiven provides:

Host
Port
Database name
Username
Password

Host: pg-12345.aivencloud.com
Port: 26257
User: avnadmin
Database: defaultdb

Test Aiven Connection via Terminal

Step 5 — Connect Power BI to Aiven PostgreSQL

Open Power BI Desktop
Select Get Data → PostgreSQL database
Enter:

Server: pg-12345.aivencloud.com
Database: defaultdb

Enter Aiven username & password
Click Connect

Power BI will establish a secure SSL connection to Aiven.

# EXCEL

Enock Kipngetich — Tue, 14 Oct 2025 19:08:53 +0000

WHAT IS EXCEL

analysis tool used to clean data

# DATABASE

Enock Kipngetich — Tue, 14 Oct 2025 07:40:12 +0000

TYPES OF DATABASES

relational database.
non relational database.

Relational database

stores data in structured form in rows and columns.

Advantages & Disadvantages

Structured data relationship
Acid compliance

#Difference between star and snowflake schema

Enock Kipngetich — Thu, 09 Oct 2025 06:14:28 +0000

Star schema

Simple database design used in data warehousing where a central fact table is connected directly to dimension tables.

Snow flake schema.

Data warehouse schema that organizes data in a logical and normalized way meaning that data is divided into multiple related tables to reduce redundancy and improve data integrity.

Difference between snowflake schema and star schema.

Star schema has Central fact table connected directly to dimension tables while snow flake has central fact table connected to dimension tables, which are further normalized into multiple related tables.
Star schema dimension tables contain redundant data. Snow flake dimension tables are broken into smaller related tables to eliminate redundancy.
Star fewer joins simple queries. Snow flake more joins complex queries.
Star schema faster query performance due to fewer joins. Snow flake Slower query performance due to more joins.
Star schema consumes more storage space. Snow flake requires less storage space.
Star schema easier to understand and design resembles a star shape. Snow flake more complex to understand resembles a snowflake shape.

Advantages of star schema.

Easy for users and developers to understand.
Faster query execution due to fewer joins.
Efficient for OLAP and data visualization tools.
Ideal for data marts and reporting.

Disadvantages of star schema

Redundant data in dimension tables increases storage cost.
Updates or changes in dimensions can cause data inconsistency.
Not ideal for very large or complex datasets.

Advantages of snow flake schema

Reduced data redundancy.
Better data integrity due to normalization.
Easier maintenance updates to dimension data propagate automatically.

Disadvantages of snow flake schema .

Complex queries with multiple joins slower performance.
Harder for business users to understand.
Slightly more difficult to design and maintain.

Where to use snow flake schema.

You have a large and complex database with many hierarchical relationships.
Data consistency and storage efficiency are more important than query speed.
You need a well-organized structure for detailed analysis and reporting.

Where to use star schema.

Data Warehouses and Data Marts.
Business Intelligence and Analytics Platforms.
Management Dashboards and Reporting Systems.
Ad-Hoc Analysis Environments.

#Raw data security

Enock Kipngetich — Wed, 08 Oct 2025 09:51:46 +0000

https://www.google.com/imgres?q=data%20analysis%20images&imgurl=https%3A%2F%2Fwww.simplilearn.com%2Fice9%2Ffree_resources_article_thumb%2FWhat_is_Data_Analysis.jpg&imgrefurl=https%3A%2F%2Fwww.simplilearn.com%2Fdata-analysis-methods-process-types-article&docid=kDHjxOzq3zlBJM&tbnid=tz18iGnDGgy5eM&vet=12ahUKEwjtrfzQqZSQAxXvW0EAHSusPMoQM3oECBgQAA..i&w=848&h=477&hcb=2&ved=2ahUKEwjtrfzQqZSQAxXvW0EAHSusPMoQM3oECBgQAA

#Row level data security in power bi

Enock Kipngetich — Tue, 07 Oct 2025 11:55:32 +0000

why excel is still relevant in the era of power bi and python

Enock Kipngetich — Sat, 04 Oct 2025 11:38:47 +0000

Enock Kipngetich

Oct 4 '25

#Why excel is still relevant in this era of Power bi $ Python

#beginners #excel

1 min read

#Why excel is still relevant in this era of Power bi $ Python

Enock Kipngetich — Sat, 04 Oct 2025 09:29:10 +0000

Excel

Has been the cornerstone of data analysis, business reporting, and financial modeling.
Accessibility and Familiarity
Easy to learn and use
Most professionals already know the basics. Universally available.
Flexible budgeting to dash board creation.

Limitations of excel.

It struggles with very large datasets.
Manual errors are common when formulas or links break.
Collaboration can be clunky compared to web-based platforms.

Conclusion

Excel is far from obsolete.
It remains the gateway to data analytics a versatile tool that bridges traditional spreadsheet work and modern data science.
In the era of Power BI and Python, Excel’s relevance lies in its adaptability.
Rather than being replaced, it has become the foundation upon which more advanced analytics tools are built.
Organizations value accessibility, collaboration, and flexibility, Excel will continue to be a vital part of the analytics ecosystem.

# Key concepts of a schema

Enock Kipngetich — Tue, 23 Sep 2025 11:18:32 +0000

Introduction

It is a blue print of a data base.

Types of schemas

1.Star schema
2.Snowflake schema