Session 6 Building a Successful Analytics Team Ashim Avinash Sahu - - PDF document

SOA Predictive Analytics Seminar – Malaysia

27 Aug. 2018 | Kuala Lumpur, Malaysia

Session 6 Building a Successful Analytics Team

Ashim Avinash Sahu

BUILDING A SUCCESSFUL ANALYTICS TEAM

Ashim Sahu

WHAT’S THE AGENDA ?

01 02 03 04

What does an analytics team do ? The 8 step process Describing the 8 Step process Some Do’s and Don'ts

WHAT’S THE “IDEA” IN INSURANCE?

Marketing

01

• How can I improve my leads quality while

keeping costs down ?

• Can I do targeted customer campaigns to

improve my brand awareness ?

Customer Experience

02

• Why are my customers dropping their policies ?
• Where should I set up my branches ?
• Do customers have a preferred channel ?

Sales

03

• Who and where should I recruit ?
• What can I do to maintain to retain my agents

performance ?

Pricing

04

• Am I predicting the right risk against my customer’s

health ?

• Are my assumptions of customer’s health still valid ?

Product Development

05

• What are the features in my product do my

customer like ?

• Does the new dynamic world create new insurance

needs for my customers ?

Claims

06

• Am I paying the right amount of claims for the

treatment / damage ?

• Are my providers overcharging me ?

Template downloaded from https://www.showeet.com/

HOW DOES AN ANALYTICS TEAM SOLVE THIS ?

The job of any Analytics team is to answer business questions – by extracting the right solutions from the right data

Icons downloaded from https://www.flaticon.com/packs/data-analytics

Data Engineering Data Sciences and Analytical Solutioning Insights and Strategy Experimentation, Testing and Implementation

THE 8 STEP SETUP AND SCALE UP PROCESS

Identify the Low Hanging Fruits

• Defining easy to implement business

problems which can generate significant business value.

Get the right stakeholders

• The Low Hanging fruits will define
• this. Generally Senior Stakeholders.

Hire the right people

• Should be designed to cater to the

needs of the first hires and have low setup time.

Scale up

• Take the next step by identifying the

newer business problems and the resources needed.

• Re-iterate the cycle.

Market it !

• Market your results.
• Start exploring newer business

problems with the broader

• rganization.

Solve it and Test it !

• The real work. Re-iterate till it
• works. If possible test many times.

Integrate team with the business

• The best solutions come together with

the business understanding. Inter team collaboration becomes critical.

01 02 03 04 05 06 07 08

• The first few hires are the most

critical since they will define the medium term direction as well as team perception.

Invest in the right technology

FOCUS ON LOW HANGING FRUITS FIRST

The first year defines the team

• Data will not be freely available and questionable quality.
• Low on resources.
• The technology would be basic.
• Only work on analytical problems which have are valuable

and not difficult to implement.

https://hatrabbits.com/

STAKEHOLDER MANAGEMENT AND INTEGRATION TO BUSINESS BECOMES THE NEXT KEY STEP

They define the problem

• Affected directly by the problem, so

they know what they want as an output.

Introduce you to the working level

• Will

get you integrated with the people who know how things are run, the knowledge of which is vital for the success of any analytical project.

Support you in the boardroom

• Will fight the battles for you in the

boardroom to get you the initial investment and fend off detractors.

Implement your solutions

• Will find the resources to test your

hypothesis and implement your proposals and be your biggest advocates.

Engage the right Senior Stakeholders

LETS PUT THE CLASSICAL CUSTOMER LAPSE PROBLEM THROUGH THE ANALYTICAL PROCESS

Develop Solution

Define

Get Data Size of Price Test

Defining Business Problem

• Can I predict policies which have the

highest risk of lapsing and improve my lapse rates?

What Data do I need ?

• Customer Policy and Attached Agent

historical buying/selling behavior

• Call Centre and claims behavior
• External data enrichment ?

What’s the right solution ?

• Build multiple supervised learning models

and select the model with best prediction power

How much value will this create?

• If we try to target the top decile of

customers at the highest risk, how will this improve my persistency ?

How to Test and

• perationalize this ?
• How should I design a pilot

campaign using the right sample size ?

• How do I scale up and
• perationalize if the pilot is

successful?

LET’S TWEAK THE BUSINESS PROBLEM – NO MORE A LOW HANGING FRUIT

Defining Business Problem

• Can I understand why my policies

are lapsing and how can I reduce it?

What Data do I need ?

• Customer Policy and Attached Agent

historical buying/selling behavior

• Call Centre and claims behavior
• External data enrichment ?

What’s the right solution ?

• Build a robust Regression model for

customer micro-segments and understand which variables have the most impact after partialling out

How do I interpret this?

• Can I affect to reduce the impact of the

predictors through business strategy ?

• Can I leverage on the predictive model ?

How to Test and

• perationalize this ?
• How should I design a pilot

campaign using the right sample size ?

• How do I scale up and
• perationalize if the pilot is

successful?

Develop Solution

Define

Get Data Size of Price Test

THE RECRUITMENT NEEDS TO BE CAREFULLY THOUGHT OF

Data Engineers

• ETL, SQL Experts
• Mostly Software
• riented

Data Scientists

• Statisticians / Quants
• Machine Learning

Engineers

Insights and Strategy

• Management Consultants
• Visualization Experts

Testing and Implementation

• Lean Innovation

practitioners

• Campaign Managers
• Project Managers

Maturity Scaling Up Inception

Working level under the experts

• Junior level hires with some experience in one of the fields below

to be trained and molded based on established processes

Senior Experts hiring

• Experienced professionals hired to fine tune and operationalize the first

solutions created. Find newer problems to work on.

• Experts in1 or 2 of the skills listed.

Analytics Breadth and Business understanding

• First Hires. Help define the goal and strategy of the team.
• Solve the first analytical problems.
• Have at least 3 of the 4 skill sets listed below.

Time Breadth of skills per hire

TECHNOLOGY INVESTMENT NEEDS TO BE STRUCTURED AND FLEXIBLE

Data Engineers

• ETL, SQL Experts
• Mostly Software
• riented

Data Scientists

• Statisticians / Quants
• Machine Learning

Engineers

Insights and Strategy

• Management

Consultants

• Visualization Experts

Testing and Implementation

• Lean Innovation

practitioners

• Campaign Managers
• Project Managers

Enterprise Server Versions

Inception Scale Up Maturity

Workstations Enterprise Servers Big Data / Real Time Analytics / GPUs

Time

THE PROCESS OF SOLVING THE PROBLEM … IS HIGHLY ITERATIVE !

Develop Solution

Define

Get Data and Explore Size of Price Test

Refine Business Understanding Improve Solution Improve Solution Get More Data Refine Business Understanding

Value Creation

SHARE, SHARE AND SHARE. SCALE UP !

Step 01

Step 02

Step 03

Share Successes across multiple forums

Share your success across multiple business verticals. Both bottoms up and top down sharing are essential.

Get the detractors on your side

Value Creation is the biggest winner in the board room. Good opportunity to start getting

• ther senior stakeholders on board.

Plan to solve new business problems

Starting devising plans for solving newer and tougher business problems.

Scale Up !

Upgrade and scale up people, technology and value creation targets. Prepare for your next 8 Step cycle.

SOME OFTEN FORGOTTEN DO’S AND DON’TS

Hold Analytics accountable for promised value Hire both internal and external people Do Do Invest in Data Quality – Garbage In = Garbage out Do Keep it simple in the beginning Do Forget to upskill your analytics teams Believe analytics is an immediate magic wand Don’t Don’t Forget to Test before deploying Don’t Give in to temptation to use analytics for brownie points Don’t

Do’s Don'ts

THANK YOU!

Do you have any questions?

Ashim Sahu AshimAvinash.Sahu@aia.com