Telco Project - README

Table of Contents:

Project Description

Project Goals

Hypothesis/Questions

Data Dictionary

Planning

Instruction To Replicate

Takeaways

Project Description:

Back to 'Table of Contents'

Using the 'telco' dataset from a SQL database, figure out what are causes to customer churn within the company.

Project Goals:

Back to 'Table of Contents'

Implement Data Science Pipeline

Acquire 'Telco' Dataset

Prepare 'Telco' Dataset

Explore 'Telco' Dataset

Model 'Telco' Dataset

Deliver 'Telco' Dataset

Hypothesis/Questions:

Back to 'Table of Contents'

Hypothesis:

Given that 'Telco' is a company providing a cellular and internet service to consumers(People), I believe that being able to identify what is deemed as valuable to the consumer weighed by it's cost is the primary cause to customer churn.

Questions:

What does telco offer

What is the price of everything

Are there specific services that retain better

Are there specific services that retain worse

Are there demographics of individuals

What sort of advertisements are posted

Are there customer comments for those who churned

Are there customer comments for those who joined

Are there timestamps for joining and leaving

Are there customer satisfaction reports

Are there reports of faulty services/equipment

What is the situation locally(placement, access, patterns, crime, common knowledge/beliefs/morals/ethics of community)

Data Dictionary:

Back to 'Table of Contents'

Field Name	Data Type	Data Format	Description	Example
gender	object	str	Defines gender of customer	'Male'
senior_citizen	int	1, 0	Defines customer as being of senior-citizen status or not	1
partner	object	str	Defines customer as having a partner or not	'Yes'
dependents	object	str	Defines customer as having dependents or not	'Yes'
tenure	int	#	Defines customer's tenure in months	12
phone_service	object	str	Defines customer as having phone service or not	'Yes'
multiple_lines	object	str	Defines customer as having multiple lines or not	'Yes'
online_security	object	str	Defines customer as having online security or not	'Yes'
online_backup	object	str	Defines customer as having online backup or not	'Yes'
device_protection	object	str	Defines customer as having device protection or not	'Yes'
tech_support	object	str	Defines customer as having tech support or not	'Yes'
streaming_tv	object	str	Defines customer as having tv streaming service or not	'Yes'
streaming_movies	object	str	Defines customer as having movie streaming service or not	'Yes'
paperless_billing	object	str	Defines customer as having paperless billing or not	'Yes'
monthly_charges	float	00.00	Defines customer's monthly charges in USD($)	37.48
total_charges	float	00.00	Defines customer's total charges with telco company in USD($)	7862.44
churn	object	str	Defines customer churn status within the telco company	'Yes'
churn_month	datetime	YYYY-MM-DD	Defines customer's date of churn	2022-01-31
signup_date	datetime	YYYY-MM-DD	Defines customer's date of signing-up with the company	2021-04-21
contract_type	object	str	Defines customer's contract type	'Month-to-month'
internet_service_type	object	str	Defines customer's internet service type	'DSL'
payment_type	object	str	Defines customer's payment type	'Mailed Check'
sign_year	object	str	Defines customer's signup year	2021
sign_month	object	str	Defines customer's signup month	12
sign_day	object	str	Defines customer's signup day	21
sign_dayofweek	object	str	Defines customer's signup day-of-week	0 (Monday)
value_per_total_services	float	float	Defines the mean cost per service for the customer in USD	12.87
value_per_total_extra_services	float	float	Defines the mean cost per extra service for the customer in USD	12.87
DUMMY COLUMNS	uint	0, 1	Dummy columns generated from object columns	0

Planning:

Back to 'Table of Contents'

Planning

This good ol’ thingy

Acquire

env.file

SQL query

acquire.py

acquire.ipynb

Prepare

Remove unwanted cols

Aggregate cols

Confirm veracity of cols

prepare.py

prepare.ipynb

Explore

Determine best cols

Do hypothesis null/alternate

Stat check best cols

explore.py

explore.ipynb

Modeling

Decision Tree

Best model

Random Forest

Best model

KNN

Best model

Linear Regression

Best model

TOP 3 COMPARISON

modeling.py

modeling.ipynb

Delivery

Final notebook (.ipynb)

Readme.md

Instructions To Replicate:

Back to 'Table of Contents'

1. Clone this repo
2. Create 'env.py' file that connects to SQL
3. Run desired .ipynb
   
   
   

Takeaways:

Back to 'Table of Contents'

Key Findings:

• Less Likely to Churn
  • 'partner' = 'Yes'
  • 'dependents' = 'Yes'
  • 'online_security' = 'Yes', 'No internet service'
  • 'online_backup = 'No internet service'
  • 'device_protection' = 'Yes', 'No internet service'
  • 'tech_support' = 'Yes', 'No internet service'
  • 'streaming_tv' = 'No internet service'
  • 'streaming_movies' = 'No internet service'
  • 'contract_type' = 'One year', 'Two year'
  • 'total_services' = '1', '6', '7', '8', '9'
  • 'total_extra_services' = '0', '3', '4', '5', '6'
• AVOID
  • 'payment_type' = 'Electronic check'
  • 'contract_type' = 'Month-to-month'
  • 'total_services'= '3', '4'
  • 'total_extra_services' = '1', '2'
    
    

Recommendations:

• Focus advertisement efforts towards families and couples and avoid older population
• Market offers as single services or bundled services with at least 3 additional services
• For customers, that want internet services
  • Get them to go with DSL services
  • Minimum add-ons of 3
• For paying customers:
  • Do not allow monthly contracts
  • Do not allow electronic check forms of payment
    
    

Takeaways:

• Target audience:
  • Couples
  • Families/Dependents
  • Non-senior citizen
• Services Emphasis:
  • Internet Services:
    • No internet or DSL
    • If internet, then bundle at least 3 add-ons
• Payments:
  • DO NOT ALLOW ELECTRONIC CHECK
• Contracts:
  • DO NOT ALLOW MONTH CONTRACT
    
    

If I Had More Time:

• For customer service:
  • Why does everyone only signup on the 21st of any given month/year
  • Why did everyone churn on 31JAN2022
  • What happened in 2021
  • Assess customer comments/sentiment:
    • Are our services adequate for their price
    • Are our services performing properly or breaking
    • Why did the customer signup/churn
• For marketing:
  • What are the prices of all services/bundles we provide
  • What advertisements are we running
• Further exploration:
  • Why is month-to-month contract popular
  • Why is tenure getting shorter, but monthly charges dropping
  • Are extra services necessary or a burden
  • Compare local competition if any
  • Establish local 'common knowledge'/'norm'