#RintisKarirImpian

Data Science Methodology (CRISP-DM)

& Business Use Case In Marketing

Internal
 LEARNING GOALS

In this section, we will cover :

● Understanding data science methodology
● Understanding data science usecases

#RintisKarirImpian
Internal
 CRISP-DM

• Cross-industry standard process for data mining (CRISP-DM)

• A comprehensive data mining methodology and framework

• Breaks down the life cycle of a data mining project into 6 phases

• Over 200 members of the CRISP-DM

• DM Vendors - SPSS, NCR, IBM, SAS, SGI, Data Distilleries, Syllogic, Magnify, ..

• System Suppliers / consultants - Cap Gemini, ICL Retail, Deloitte & Touche, …

• End Users - BT, ABB, Lloyds Bank, AirTouch, Experian, ...

#RintisKarirImpian
Internal
 WHY CRISP – DM ?

• The data mining process must be reliable and repeatable by

people with little data mining skills

• CRISP-DM provides a uniform framework for

• Guidelines

• Experience documentation

• CRISP-DM is flexible to account for differences

• Different business/agency problems

#RintisKarirImpian
Internal • Different data
 TOP DATA MINING METHODOLOGY

#RintisKarirImpian
Internal
 CRISP–DM : 6 PHASES

Phases … Explanation …

• Understanding business objectives and

Business Understanding requirements
• Define data mining problem / pain point

• Initial data collection and familiarization

Data Understanding
• Identify data quality issues

• Record and attribute selection

Data Preparation
• Data cleansing

Modeling • Run the data mining tools

Evaluation • Determine if results meet business objectives

• Put the resulting models into practice

Deployment
• Set up for continuous mining of the data

#RintisKarirImpian
Internal
 CRISP-DM : 6 PHASES AND TASKS

Business Data
Data
Understandin Understandin Modeling Evaluation Deployment
Preparation
g g

Determine Select
Collect Evaluate Plan
Business Select Data Modeling
Initial Data Results Deployment
Objectives Technique
Plan
Monitoring
Assess Describe Generate Review
Clean Data &
Situation Data Test Design Process
Maintenanc
e
Determine
Explore Construct Determine Produce
Data Mining Build Model
Data Data Next Steps Final Report
Goals

Produce Verify Data Integrate Assess Review

Project Plan Quality Data Model Project

Format
Data
#RintisKarirImpian
Internal
 BUSINESS UNDERSTANDING

Task … Description …

Determine business • Describe your primary

objectives objective from a business
perspective

Assess situation • Inventory of resources, risks, cost &

benefits

Determine data • Business success criteria

mining goals • Data mining success criteria

• Estimated timeline and budget

Develop Project plan • Initial assessment of tools and
techniques

#RintisKarirImpian
Internal
 BUSINESS UNDERSTANDING

• Difficult

• Often, you have to enter a new field

• You have to explain data science limitations to

non-experts
• No, performance will not be 100%

• We need much more data to train an accurate

model
#RintisKarirImpian
• For tomorrow, it is impossible
Internal
 BUSINESS UNDERSTANDING – PROJECT PLAN

Oct Nov Dec

Activity 27. 04. 11. 18. 25. 01. 08. 15. 22. 29. 06. 13. 20. 27.

Data
Understanding

Data
Preparation

Modelin
g

Evaluatio
n

Deployme
nt

10/15/202
0
#RintisKarirImpian
Internal
 BUSINESS UNDERSTANDING – DO AND DON’TS

• Have a lot of patient for vaguely defined problems

• Learn to concretize or even reduce the scope of the initial idea

• Data Sample
• Real life case
• Quantitative success metrics
• Don’t waste you time on ill-defined, unrealistic projects
#RintisKarirImpian
Internal
 DATA UNDERSTANDING

Task … Description …
• Initial data collection report
Collect initial data • Acquire the data (of access to the data) listed
in the project resources.

Describe data • Describe the data, including the format of the

data, the quantity of data

Explore data • Using querying, visualization, and reporting

techniques.

• Is the data complete (does it cover all the

Verify data quality cases required)?
• Are there missing values in the data?

#RintisKarirImpian
Internal
 DATA UNDERSTANDING – SPOT OUTLIER

#RintisKarirImpian
Internal
 DATA UNDERSTANDING – DATA EXPLORATION

#RintisKarirImpian
Internal
 DATA UNDERSTANDING – DO AND DON’TS

• Don’t economize on this phase

• Don’t trust data quality estimates provided by your customer

• Verify as far as you can, if you data is correct, complete, coherent,

deduplicated, representative, independent, up-to-date, stationary

• Investigate what sort of processing was applied to the raw data

• Understand anomalies and outliers

#RintisKarirImpian
Internal
 DATA PREPARATION
Task … Description …
• Reconsider data selection criteria and
decide which dataset will be used
Select data
• Explain why certain data was included or
excluded

Clean Data • Correct, remove or ignore noise.

• Derived attributes.
Construct data • How can missing attributes be
constructed or imputed?

• Integrate sources and store result (new

Integrate Data
tables and records).

• Rearranging attributes
Format Data • Reordering records
• Reformatted within-value

#RintisKarirImpian
Internal
 DATA PREPARATION – REMOVE OUTLIER

Data understanding and preparation will usually consume half of more of your
project time !

#RintisKarirImpian
Internal
 DATA PREPARATION – DO AND DON’TS

• Automate this phase as far as possible

• When merging multiple sources, track provenance of your data

• Use workflow tools to help you with the above

• Prepare your customer that data understanding, and preparation

take considerable amount of time

#RintisKarirImpian
Internal
 MODELING

Task … Description …

Select modelling • Document the actual modelling technique that is to

technique be used.

Generate test • Describe the intended plan for training, testing, and
design evaluating the models. 

Built model • Run the modelling tool on the prepared dataset to

create one or more models.

• Interpret the models according to your domain

Assess model knowledge, your data mining success criteria and
your desired test design. 

#RintisKarirImpian
Internal
 MODELLING – TOOLING SELECTION

• Should I use general purpose language ?

Breadth = performance, lots of general purpose libraries
and tooling, easy creation of webs services

• Should I use data analysis language ?

Depth = easy data manipulation, latest models and
statistical technique available

• Where your model will be deployed ?

•
#RintisKarirImpian
Can I afford a prototype ?
Internal
 MODELLING – DO AND DON’TS

•Be creative with your features (feature engineering)

•Allocate time for hyperparameter optimization

•When possible, peak inside your model and consult it with domain expert
• assess feature importance

• run your model on simulated data

• Develop your model with deployment conditions in mind

#RintisKarirImpian
Internal
 EVALUATION

Task … Description …

• Assessment of data mining results

Review results
• Approved models

• Summarize the process review and highlight

Review Process activities that have been missed and those
that should be repeated.

Determine next • List of possible actions 

steps • Decision 

#RintisKarirImpian
Internal
 EVALUATION – PERFORMANCE OF MODEL

#RintisKarirImpian
Internal
 EVALUATION – DO AND DON’TS

•Work with performance criteria dictated by your customer’s business model

•Assess not only performance, but also practical aspects, related to

deployment, for example :

•training and prediction speed

•robustness and maintainability (tooling, dependence on another
subsystems, library)
#RintisKarirImpian
Internal
 DEPLOYMENT

Task … Description …

Plan Deployment • Summaries your deployment strategy

Plan Monitoring • Summaries the monitoring and

maintenance strategy

Produce Final • This is the final written report of the data

Report mining engagement

• Assess what went right and what went

Review project wrong, what was done well and what
needs to be improved.

#RintisKarirImpian
Internal
 TOP 5 BUSINESS CASE IN MARKETING
Business Case What is …

• Grouping a customer into groups of individuals that are similar

Customer Segmentation
specific ways relevant to marketing

Product • A filtering system that seeks to predict and show that items that
Recommendation
a user would like to purchase

• Data mining technique used by retailers to increase sales by

Market Basket Analysis
better understanding customer purchasing pattern

• Machine learning technique to identify the likelihood of future

Predictive Analytics
outcomes based on historical data

• Help you learn more about who your customers are and how
Sentiment Analysis
you can better engage with them. 

#RintisKarirImpian
Internal
 CUSTOMER SEGMENTATION

Companies… Strategy and success story…

• Segmentation based on user experience survey results

• Objective of survey is to see what each of their customers

expected out of their company

• Serve various industries and different buyer personas

• Segmentation based on :

• Buyer persona: chef, restaurant owner, f&b manager

• Type of business: restaurant, hotel, restaurant chain, food

production.

#RintisKarirImpian • Conversion rate from lead is increase 3x

Internal
 PREDICTIVE ANALYTICS

Companies… Strategy and success story…

• Helps customers enjoy their mobile life

• They developed a predictive model based on profiling data of

customers who used their mobile phones at winter sports

hotspots in the past.

• More targeted campaign

• The result has been up to 380% higher campaign response rates

#RintisKarirImpian
Internal
 PRODUCT RECOMMENDATION

Companies… Strategy and success story…

• Over 80% of what people watch comes from recommendation

• Recommendations are driven by machine learning

• With the increasing demand for more personalized playlists, Spotify launched

Discover Weekly in 2015

• According to data, Discover Weekly listeners stream music on Spotify more than

double the amount of users who don’t listen Discover Weekly.

#RintisKarirImpian
Internal
 MARKET BASKET ANALYSIS

Companies… Strategy and success story…

• Present products that are most likely being bought

together.
• It’s led to a 10% to 15% increase in conversion rate.

• It identifies items bought together and shows

relevant recommendations to customers.

#RintisKarirImpian
Internal
 SENTIMENT ANALYTICS

Companies… Strategy and success story…

• The insights gained have been a demographic +

psychographic profiling of Twitter users, to define
segments and better target online marketing
campaigns for their customers

• Monitor opinions through social media for marketing

campaign:
• potential reputation issue
• check and reassure customers that products are
safe

#RintisKarirImpian
Internal